thesis.lyx 437 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687688689690691692693694695696697698699700701702703704705706707708709710711712713714715716717718719720721722723724725726727728729730731732733734735736737738739740741742743744745746747748749750751752753754755756757758759760761762763764765766767768769770771772773774775776777778779780781782783784785786787788789790791792793794795796797798799800801802803804805806807808809810811812813814815816817818819820821822823824825826827828829830831832833834835836837838839840841842843844845846847848849850851852853854855856857858859860861862863864865866867868869870871872873874875876877878879880881882883884885886887888889890891892893894895896897898899900901902903904905906907908909910911912913914915916917918919920921922923924925926927928929930931932933934935936937938939940941942943944945946947948949950951952953954955956957958959960961962963964965966967968969970971972973974975976977978979980981982983984985986987988989990991992993994995996997998999100010011002100310041005100610071008100910101011101210131014101510161017101810191020102110221023102410251026102710281029103010311032103310341035103610371038103910401041104210431044104510461047104810491050105110521053105410551056105710581059106010611062106310641065106610671068106910701071107210731074107510761077107810791080108110821083108410851086108710881089109010911092109310941095109610971098109911001101110211031104110511061107110811091110111111121113111411151116111711181119112011211122112311241125112611271128112911301131113211331134113511361137113811391140114111421143114411451146114711481149115011511152115311541155115611571158115911601161116211631164116511661167116811691170117111721173117411751176117711781179118011811182118311841185118611871188118911901191119211931194119511961197119811991200120112021203120412051206120712081209121012111212121312141215121612171218121912201221122212231224122512261227122812291230123112321233123412351236123712381239124012411242124312441245124612471248124912501251125212531254125512561257125812591260126112621263126412651266126712681269127012711272127312741275127612771278127912801281128212831284128512861287128812891290129112921293129412951296129712981299130013011302130313041305130613071308130913101311131213131314131513161317131813191320132113221323132413251326132713281329133013311332133313341335133613371338133913401341134213431344134513461347134813491350135113521353135413551356135713581359136013611362136313641365136613671368136913701371137213731374137513761377137813791380138113821383138413851386138713881389139013911392139313941395139613971398139914001401140214031404140514061407140814091410141114121413141414151416141714181419142014211422142314241425142614271428142914301431143214331434143514361437143814391440144114421443144414451446144714481449145014511452145314541455145614571458145914601461146214631464146514661467146814691470147114721473147414751476147714781479148014811482148314841485148614871488148914901491149214931494149514961497149814991500150115021503150415051506150715081509151015111512151315141515151615171518151915201521152215231524152515261527152815291530153115321533153415351536153715381539154015411542154315441545154615471548154915501551155215531554155515561557155815591560156115621563156415651566156715681569157015711572157315741575157615771578157915801581158215831584158515861587158815891590159115921593159415951596159715981599160016011602160316041605160616071608160916101611161216131614161516161617161816191620162116221623162416251626162716281629163016311632163316341635163616371638163916401641164216431644164516461647164816491650165116521653165416551656165716581659166016611662166316641665166616671668166916701671167216731674167516761677167816791680168116821683168416851686168716881689169016911692169316941695169616971698169917001701170217031704170517061707170817091710171117121713171417151716171717181719172017211722172317241725172617271728172917301731173217331734173517361737173817391740174117421743174417451746174717481749175017511752175317541755175617571758175917601761176217631764176517661767176817691770177117721773177417751776177717781779178017811782178317841785178617871788178917901791179217931794179517961797179817991800180118021803180418051806180718081809181018111812181318141815181618171818181918201821182218231824182518261827182818291830183118321833183418351836183718381839184018411842184318441845184618471848184918501851185218531854185518561857185818591860186118621863186418651866186718681869187018711872187318741875187618771878187918801881188218831884188518861887188818891890189118921893189418951896189718981899190019011902190319041905190619071908190919101911191219131914191519161917191819191920192119221923192419251926192719281929193019311932193319341935193619371938193919401941194219431944194519461947194819491950195119521953195419551956195719581959196019611962196319641965196619671968196919701971197219731974197519761977197819791980198119821983198419851986198719881989199019911992199319941995199619971998199920002001200220032004200520062007200820092010201120122013201420152016201720182019202020212022202320242025202620272028202920302031203220332034203520362037203820392040204120422043204420452046204720482049205020512052205320542055205620572058205920602061206220632064206520662067206820692070207120722073207420752076207720782079208020812082208320842085208620872088208920902091209220932094209520962097209820992100210121022103210421052106210721082109211021112112211321142115211621172118211921202121212221232124212521262127212821292130213121322133213421352136213721382139214021412142214321442145214621472148214921502151215221532154215521562157215821592160216121622163216421652166216721682169217021712172217321742175217621772178217921802181218221832184218521862187218821892190219121922193219421952196219721982199220022012202220322042205220622072208220922102211221222132214221522162217221822192220222122222223222422252226222722282229223022312232223322342235223622372238223922402241224222432244224522462247224822492250225122522253225422552256225722582259226022612262226322642265226622672268226922702271227222732274227522762277227822792280228122822283228422852286228722882289229022912292229322942295229622972298229923002301230223032304230523062307230823092310231123122313231423152316231723182319232023212322232323242325232623272328232923302331233223332334233523362337233823392340234123422343234423452346234723482349235023512352235323542355235623572358235923602361236223632364236523662367236823692370237123722373237423752376237723782379238023812382238323842385238623872388238923902391239223932394239523962397239823992400240124022403240424052406240724082409241024112412241324142415241624172418241924202421242224232424242524262427242824292430243124322433243424352436243724382439244024412442244324442445244624472448244924502451245224532454245524562457245824592460246124622463246424652466246724682469247024712472247324742475247624772478247924802481248224832484248524862487248824892490249124922493249424952496249724982499250025012502250325042505250625072508250925102511251225132514251525162517251825192520252125222523252425252526252725282529253025312532253325342535253625372538253925402541254225432544254525462547254825492550255125522553255425552556255725582559256025612562256325642565256625672568256925702571257225732574257525762577257825792580258125822583258425852586258725882589259025912592259325942595259625972598259926002601260226032604260526062607260826092610261126122613261426152616261726182619262026212622262326242625262626272628262926302631263226332634263526362637263826392640264126422643264426452646264726482649265026512652265326542655265626572658265926602661266226632664266526662667266826692670267126722673267426752676267726782679268026812682268326842685268626872688268926902691269226932694269526962697269826992700270127022703270427052706270727082709271027112712271327142715271627172718271927202721272227232724272527262727272827292730273127322733273427352736273727382739274027412742274327442745274627472748274927502751275227532754275527562757275827592760276127622763276427652766276727682769277027712772277327742775277627772778277927802781278227832784278527862787278827892790279127922793279427952796279727982799280028012802280328042805280628072808280928102811281228132814281528162817281828192820282128222823282428252826282728282829283028312832283328342835283628372838283928402841284228432844284528462847284828492850285128522853285428552856285728582859286028612862286328642865286628672868286928702871287228732874287528762877287828792880288128822883288428852886288728882889289028912892289328942895289628972898289929002901290229032904290529062907290829092910291129122913291429152916291729182919292029212922292329242925292629272928292929302931293229332934293529362937293829392940294129422943294429452946294729482949295029512952295329542955295629572958295929602961296229632964296529662967296829692970297129722973297429752976297729782979298029812982298329842985298629872988298929902991299229932994299529962997299829993000300130023003300430053006300730083009301030113012301330143015301630173018301930203021302230233024302530263027302830293030303130323033303430353036303730383039304030413042304330443045304630473048304930503051305230533054305530563057305830593060306130623063306430653066306730683069307030713072307330743075307630773078307930803081308230833084308530863087308830893090309130923093309430953096309730983099310031013102310331043105310631073108310931103111311231133114311531163117311831193120312131223123312431253126312731283129313031313132313331343135313631373138313931403141314231433144314531463147314831493150315131523153315431553156315731583159316031613162316331643165316631673168316931703171317231733174317531763177317831793180318131823183318431853186318731883189319031913192319331943195319631973198319932003201320232033204320532063207320832093210321132123213321432153216321732183219322032213222322332243225322632273228322932303231323232333234323532363237323832393240324132423243324432453246324732483249325032513252325332543255325632573258325932603261326232633264326532663267326832693270327132723273327432753276327732783279328032813282328332843285328632873288328932903291329232933294329532963297329832993300330133023303330433053306330733083309331033113312331333143315331633173318331933203321332233233324332533263327332833293330333133323333333433353336333733383339334033413342334333443345334633473348334933503351335233533354335533563357335833593360336133623363336433653366336733683369337033713372337333743375337633773378337933803381338233833384338533863387338833893390339133923393339433953396339733983399340034013402340334043405340634073408340934103411341234133414341534163417341834193420342134223423342434253426342734283429343034313432343334343435343634373438343934403441344234433444344534463447344834493450345134523453345434553456345734583459346034613462346334643465346634673468346934703471347234733474347534763477347834793480348134823483348434853486348734883489349034913492349334943495349634973498349935003501350235033504350535063507350835093510351135123513351435153516351735183519352035213522352335243525352635273528352935303531353235333534353535363537353835393540354135423543354435453546354735483549355035513552355335543555355635573558355935603561356235633564356535663567356835693570357135723573357435753576357735783579358035813582358335843585358635873588358935903591359235933594359535963597359835993600360136023603360436053606360736083609361036113612361336143615361636173618361936203621362236233624362536263627362836293630363136323633363436353636363736383639364036413642364336443645364636473648364936503651365236533654365536563657365836593660366136623663366436653666366736683669367036713672367336743675367636773678367936803681368236833684368536863687368836893690369136923693369436953696369736983699370037013702370337043705370637073708370937103711371237133714371537163717371837193720372137223723372437253726372737283729373037313732373337343735373637373738373937403741374237433744374537463747374837493750375137523753375437553756375737583759376037613762376337643765376637673768376937703771377237733774377537763777377837793780378137823783378437853786378737883789379037913792379337943795379637973798379938003801380238033804380538063807380838093810381138123813381438153816381738183819382038213822382338243825382638273828382938303831383238333834383538363837383838393840384138423843384438453846384738483849385038513852385338543855385638573858385938603861386238633864386538663867386838693870387138723873387438753876387738783879388038813882388338843885388638873888388938903891389238933894389538963897389838993900390139023903390439053906390739083909391039113912391339143915391639173918391939203921392239233924392539263927392839293930393139323933393439353936393739383939394039413942394339443945394639473948394939503951395239533954395539563957395839593960396139623963396439653966396739683969397039713972397339743975397639773978397939803981398239833984398539863987398839893990399139923993399439953996399739983999400040014002400340044005400640074008400940104011401240134014401540164017401840194020402140224023402440254026402740284029403040314032403340344035403640374038403940404041404240434044404540464047404840494050405140524053405440554056405740584059406040614062406340644065406640674068406940704071407240734074407540764077407840794080408140824083408440854086408740884089409040914092409340944095409640974098409941004101410241034104410541064107410841094110411141124113411441154116411741184119412041214122412341244125412641274128412941304131413241334134413541364137413841394140414141424143414441454146414741484149415041514152415341544155415641574158415941604161416241634164416541664167416841694170417141724173417441754176417741784179418041814182418341844185418641874188418941904191419241934194419541964197419841994200420142024203420442054206420742084209421042114212421342144215421642174218421942204221422242234224422542264227422842294230423142324233423442354236423742384239424042414242424342444245424642474248424942504251425242534254425542564257425842594260426142624263426442654266426742684269427042714272427342744275427642774278427942804281428242834284428542864287428842894290429142924293429442954296429742984299430043014302430343044305430643074308430943104311431243134314431543164317431843194320432143224323432443254326432743284329433043314332433343344335433643374338433943404341434243434344434543464347434843494350435143524353435443554356435743584359436043614362436343644365436643674368436943704371437243734374437543764377437843794380438143824383438443854386438743884389439043914392439343944395439643974398439944004401440244034404440544064407440844094410441144124413441444154416441744184419442044214422442344244425442644274428442944304431443244334434443544364437443844394440444144424443444444454446444744484449445044514452445344544455445644574458445944604461446244634464446544664467446844694470447144724473447444754476447744784479448044814482448344844485448644874488448944904491449244934494449544964497449844994500450145024503450445054506450745084509451045114512451345144515451645174518451945204521452245234524452545264527452845294530453145324533453445354536453745384539454045414542454345444545454645474548454945504551455245534554455545564557455845594560456145624563456445654566456745684569457045714572457345744575457645774578457945804581458245834584458545864587458845894590459145924593459445954596459745984599460046014602460346044605460646074608460946104611461246134614461546164617461846194620462146224623462446254626462746284629463046314632463346344635463646374638463946404641464246434644464546464647464846494650465146524653465446554656465746584659466046614662466346644665466646674668466946704671467246734674467546764677467846794680468146824683468446854686468746884689469046914692469346944695469646974698469947004701470247034704470547064707470847094710471147124713471447154716471747184719472047214722472347244725472647274728472947304731473247334734473547364737473847394740474147424743474447454746474747484749475047514752475347544755475647574758475947604761476247634764476547664767476847694770477147724773477447754776477747784779478047814782478347844785478647874788478947904791479247934794479547964797479847994800480148024803480448054806480748084809481048114812481348144815481648174818481948204821482248234824482548264827482848294830483148324833483448354836483748384839484048414842484348444845484648474848484948504851485248534854485548564857485848594860486148624863486448654866486748684869487048714872487348744875487648774878487948804881488248834884488548864887488848894890489148924893489448954896489748984899490049014902490349044905490649074908490949104911491249134914491549164917491849194920492149224923492449254926492749284929493049314932493349344935493649374938493949404941494249434944494549464947494849494950495149524953495449554956495749584959496049614962496349644965496649674968496949704971497249734974497549764977497849794980498149824983498449854986498749884989499049914992499349944995499649974998499950005001500250035004500550065007500850095010501150125013501450155016501750185019502050215022502350245025502650275028502950305031503250335034503550365037503850395040504150425043504450455046504750485049505050515052505350545055505650575058505950605061506250635064506550665067506850695070507150725073507450755076507750785079508050815082508350845085508650875088508950905091509250935094509550965097509850995100510151025103510451055106510751085109511051115112511351145115511651175118511951205121512251235124512551265127512851295130513151325133513451355136513751385139514051415142514351445145514651475148514951505151515251535154515551565157515851595160516151625163516451655166516751685169517051715172517351745175517651775178517951805181518251835184518551865187518851895190519151925193519451955196519751985199520052015202520352045205520652075208520952105211521252135214521552165217521852195220522152225223522452255226522752285229523052315232523352345235523652375238523952405241524252435244524552465247524852495250525152525253525452555256525752585259526052615262526352645265526652675268526952705271527252735274527552765277527852795280528152825283528452855286528752885289529052915292529352945295529652975298529953005301530253035304530553065307530853095310531153125313531453155316531753185319532053215322532353245325532653275328532953305331533253335334533553365337533853395340534153425343534453455346534753485349535053515352535353545355535653575358535953605361536253635364536553665367536853695370537153725373537453755376537753785379538053815382538353845385538653875388538953905391539253935394539553965397539853995400540154025403540454055406540754085409541054115412541354145415541654175418541954205421542254235424542554265427542854295430543154325433543454355436543754385439544054415442544354445445544654475448544954505451545254535454545554565457545854595460546154625463546454655466546754685469547054715472547354745475547654775478547954805481548254835484548554865487548854895490549154925493549454955496549754985499550055015502550355045505550655075508550955105511551255135514551555165517551855195520552155225523552455255526552755285529553055315532553355345535553655375538553955405541554255435544554555465547554855495550555155525553555455555556555755585559556055615562556355645565556655675568556955705571557255735574557555765577557855795580558155825583558455855586558755885589559055915592559355945595559655975598559956005601560256035604560556065607560856095610561156125613561456155616561756185619562056215622562356245625562656275628562956305631563256335634563556365637563856395640564156425643564456455646564756485649565056515652565356545655565656575658565956605661566256635664566556665667566856695670567156725673567456755676567756785679568056815682568356845685568656875688568956905691569256935694569556965697569856995700570157025703570457055706570757085709571057115712571357145715571657175718571957205721572257235724572557265727572857295730573157325733573457355736573757385739574057415742574357445745574657475748574957505751575257535754575557565757575857595760576157625763576457655766576757685769577057715772577357745775577657775778577957805781578257835784578557865787578857895790579157925793579457955796579757985799580058015802580358045805580658075808580958105811581258135814581558165817581858195820582158225823582458255826582758285829583058315832583358345835583658375838583958405841584258435844584558465847584858495850585158525853585458555856585758585859586058615862586358645865586658675868586958705871587258735874587558765877587858795880588158825883588458855886588758885889589058915892589358945895589658975898589959005901590259035904590559065907590859095910591159125913591459155916591759185919592059215922592359245925592659275928592959305931593259335934593559365937593859395940594159425943594459455946594759485949595059515952595359545955595659575958595959605961596259635964596559665967596859695970597159725973597459755976597759785979598059815982598359845985598659875988598959905991599259935994599559965997599859996000600160026003600460056006600760086009601060116012601360146015601660176018601960206021602260236024602560266027602860296030603160326033603460356036603760386039604060416042604360446045604660476048604960506051605260536054605560566057605860596060606160626063606460656066606760686069607060716072607360746075607660776078607960806081608260836084608560866087608860896090609160926093609460956096609760986099610061016102610361046105610661076108610961106111611261136114611561166117611861196120612161226123612461256126612761286129613061316132613361346135613661376138613961406141614261436144614561466147614861496150615161526153615461556156615761586159616061616162616361646165616661676168616961706171617261736174617561766177617861796180618161826183618461856186618761886189619061916192619361946195619661976198619962006201620262036204620562066207620862096210621162126213621462156216621762186219622062216222622362246225622662276228622962306231623262336234623562366237623862396240624162426243624462456246624762486249625062516252625362546255625662576258625962606261626262636264626562666267626862696270627162726273627462756276627762786279628062816282628362846285628662876288628962906291629262936294629562966297629862996300630163026303630463056306630763086309631063116312631363146315631663176318631963206321632263236324632563266327632863296330633163326333633463356336633763386339634063416342634363446345634663476348634963506351635263536354635563566357635863596360636163626363636463656366636763686369637063716372637363746375637663776378637963806381638263836384638563866387638863896390639163926393639463956396639763986399640064016402640364046405640664076408640964106411641264136414641564166417641864196420642164226423642464256426642764286429643064316432643364346435643664376438643964406441644264436444644564466447644864496450645164526453645464556456645764586459646064616462646364646465646664676468646964706471647264736474647564766477647864796480648164826483648464856486648764886489649064916492649364946495649664976498649965006501650265036504650565066507650865096510651165126513651465156516651765186519652065216522652365246525652665276528652965306531653265336534653565366537653865396540654165426543654465456546654765486549655065516552655365546555655665576558655965606561656265636564656565666567656865696570657165726573657465756576657765786579658065816582658365846585658665876588658965906591659265936594659565966597659865996600660166026603660466056606660766086609661066116612661366146615661666176618661966206621662266236624662566266627662866296630663166326633663466356636663766386639664066416642664366446645664666476648664966506651665266536654665566566657665866596660666166626663666466656666666766686669667066716672667366746675667666776678667966806681668266836684668566866687668866896690669166926693669466956696669766986699670067016702670367046705670667076708670967106711671267136714671567166717671867196720672167226723672467256726672767286729673067316732673367346735673667376738673967406741674267436744674567466747674867496750675167526753675467556756675767586759676067616762676367646765676667676768676967706771677267736774677567766777677867796780678167826783678467856786678767886789679067916792679367946795679667976798679968006801680268036804680568066807680868096810681168126813681468156816681768186819682068216822682368246825682668276828682968306831683268336834683568366837683868396840684168426843684468456846684768486849685068516852685368546855685668576858685968606861686268636864686568666867686868696870687168726873687468756876687768786879688068816882688368846885688668876888688968906891689268936894689568966897689868996900690169026903690469056906690769086909691069116912691369146915691669176918691969206921692269236924692569266927692869296930693169326933693469356936693769386939694069416942694369446945694669476948694969506951695269536954695569566957695869596960696169626963696469656966696769686969697069716972697369746975697669776978697969806981698269836984698569866987698869896990699169926993699469956996699769986999700070017002700370047005700670077008700970107011701270137014701570167017701870197020702170227023702470257026702770287029703070317032703370347035703670377038703970407041704270437044704570467047704870497050705170527053705470557056705770587059706070617062706370647065706670677068706970707071707270737074707570767077707870797080708170827083708470857086708770887089709070917092709370947095709670977098709971007101710271037104710571067107710871097110711171127113711471157116711771187119712071217122712371247125712671277128712971307131713271337134713571367137713871397140714171427143714471457146714771487149715071517152715371547155715671577158715971607161716271637164716571667167716871697170717171727173717471757176717771787179718071817182718371847185718671877188718971907191719271937194719571967197719871997200720172027203720472057206720772087209721072117212721372147215721672177218721972207221722272237224722572267227722872297230723172327233723472357236723772387239724072417242724372447245724672477248724972507251725272537254725572567257725872597260726172627263726472657266726772687269727072717272727372747275727672777278727972807281728272837284728572867287728872897290729172927293729472957296729772987299730073017302730373047305730673077308730973107311731273137314731573167317731873197320732173227323732473257326732773287329733073317332733373347335733673377338733973407341734273437344734573467347734873497350735173527353735473557356735773587359736073617362736373647365736673677368736973707371737273737374737573767377737873797380738173827383738473857386738773887389739073917392739373947395739673977398739974007401740274037404740574067407740874097410741174127413741474157416741774187419742074217422742374247425742674277428742974307431743274337434743574367437743874397440744174427443744474457446744774487449745074517452745374547455745674577458745974607461746274637464746574667467746874697470747174727473747474757476747774787479748074817482748374847485748674877488748974907491749274937494749574967497749874997500750175027503750475057506750775087509751075117512751375147515751675177518751975207521752275237524752575267527752875297530753175327533753475357536753775387539754075417542754375447545754675477548754975507551755275537554755575567557755875597560756175627563756475657566756775687569757075717572757375747575757675777578757975807581758275837584758575867587758875897590759175927593759475957596759775987599760076017602760376047605760676077608760976107611761276137614761576167617761876197620762176227623762476257626762776287629763076317632763376347635763676377638763976407641764276437644764576467647764876497650765176527653765476557656765776587659766076617662766376647665766676677668766976707671767276737674767576767677767876797680768176827683768476857686768776887689769076917692769376947695769676977698769977007701770277037704770577067707770877097710771177127713771477157716771777187719772077217722772377247725772677277728772977307731773277337734773577367737773877397740774177427743774477457746774777487749775077517752775377547755775677577758775977607761776277637764776577667767776877697770777177727773777477757776777777787779778077817782778377847785778677877788778977907791779277937794779577967797779877997800780178027803780478057806780778087809781078117812781378147815781678177818781978207821782278237824782578267827782878297830783178327833783478357836783778387839784078417842784378447845784678477848784978507851785278537854785578567857785878597860786178627863786478657866786778687869787078717872787378747875787678777878787978807881788278837884788578867887788878897890789178927893789478957896789778987899790079017902790379047905790679077908790979107911791279137914791579167917791879197920792179227923792479257926792779287929793079317932793379347935793679377938793979407941794279437944794579467947794879497950795179527953795479557956795779587959796079617962796379647965796679677968796979707971797279737974797579767977797879797980798179827983798479857986798779887989799079917992799379947995799679977998799980008001800280038004800580068007800880098010801180128013801480158016801780188019802080218022802380248025802680278028802980308031803280338034803580368037803880398040804180428043804480458046804780488049805080518052805380548055805680578058805980608061806280638064806580668067806880698070807180728073807480758076807780788079808080818082808380848085808680878088808980908091809280938094809580968097809880998100810181028103810481058106810781088109811081118112811381148115811681178118811981208121812281238124812581268127812881298130813181328133813481358136813781388139814081418142814381448145814681478148814981508151815281538154815581568157815881598160816181628163816481658166816781688169817081718172817381748175817681778178817981808181818281838184818581868187818881898190819181928193819481958196819781988199820082018202820382048205820682078208820982108211821282138214821582168217821882198220822182228223822482258226822782288229823082318232823382348235823682378238823982408241824282438244824582468247824882498250825182528253825482558256825782588259826082618262826382648265826682678268826982708271827282738274827582768277827882798280828182828283828482858286828782888289829082918292829382948295829682978298829983008301830283038304830583068307830883098310831183128313831483158316831783188319832083218322832383248325832683278328832983308331833283338334833583368337833883398340834183428343834483458346834783488349835083518352835383548355835683578358835983608361836283638364836583668367836883698370837183728373837483758376837783788379838083818382838383848385838683878388838983908391839283938394839583968397839883998400840184028403840484058406840784088409841084118412841384148415841684178418841984208421842284238424842584268427842884298430843184328433843484358436843784388439844084418442844384448445844684478448844984508451845284538454845584568457845884598460846184628463846484658466846784688469847084718472847384748475847684778478847984808481848284838484848584868487848884898490849184928493849484958496849784988499850085018502850385048505850685078508850985108511851285138514851585168517851885198520852185228523852485258526852785288529853085318532853385348535853685378538853985408541854285438544854585468547854885498550855185528553855485558556855785588559856085618562856385648565856685678568856985708571857285738574857585768577857885798580858185828583858485858586858785888589859085918592859385948595859685978598859986008601860286038604860586068607860886098610861186128613861486158616861786188619862086218622862386248625862686278628862986308631863286338634863586368637863886398640864186428643864486458646864786488649865086518652865386548655865686578658865986608661866286638664866586668667866886698670867186728673867486758676867786788679868086818682868386848685868686878688868986908691869286938694869586968697869886998700870187028703870487058706870787088709871087118712871387148715871687178718871987208721872287238724872587268727872887298730873187328733873487358736873787388739874087418742874387448745874687478748874987508751875287538754875587568757875887598760876187628763876487658766876787688769877087718772877387748775877687778778877987808781878287838784878587868787878887898790879187928793879487958796879787988799880088018802880388048805880688078808880988108811881288138814881588168817881888198820882188228823882488258826882788288829883088318832883388348835883688378838883988408841884288438844884588468847884888498850885188528853885488558856885788588859886088618862886388648865886688678868886988708871887288738874887588768877887888798880888188828883888488858886888788888889889088918892889388948895889688978898889989008901890289038904890589068907890889098910891189128913891489158916891789188919892089218922892389248925892689278928892989308931893289338934893589368937893889398940894189428943894489458946894789488949895089518952895389548955895689578958895989608961896289638964896589668967896889698970897189728973897489758976897789788979898089818982898389848985898689878988898989908991899289938994899589968997899889999000900190029003900490059006900790089009901090119012901390149015901690179018901990209021902290239024902590269027902890299030903190329033903490359036903790389039904090419042904390449045904690479048904990509051905290539054905590569057905890599060906190629063906490659066906790689069907090719072907390749075907690779078907990809081908290839084908590869087908890899090909190929093909490959096909790989099910091019102910391049105910691079108910991109111911291139114911591169117911891199120912191229123912491259126912791289129913091319132913391349135913691379138913991409141914291439144914591469147914891499150915191529153915491559156915791589159916091619162916391649165916691679168916991709171917291739174917591769177917891799180918191829183918491859186918791889189919091919192919391949195919691979198919992009201920292039204920592069207920892099210921192129213921492159216921792189219922092219222922392249225922692279228922992309231923292339234923592369237923892399240924192429243924492459246924792489249925092519252925392549255925692579258925992609261926292639264926592669267926892699270927192729273927492759276927792789279928092819282928392849285928692879288928992909291929292939294929592969297929892999300930193029303930493059306930793089309931093119312931393149315931693179318931993209321932293239324932593269327932893299330933193329333933493359336933793389339934093419342934393449345934693479348934993509351935293539354935593569357935893599360936193629363936493659366936793689369937093719372937393749375937693779378937993809381938293839384938593869387938893899390939193929393939493959396939793989399940094019402940394049405940694079408940994109411941294139414941594169417941894199420942194229423942494259426942794289429943094319432943394349435943694379438943994409441944294439444944594469447944894499450945194529453945494559456945794589459946094619462946394649465946694679468946994709471947294739474947594769477947894799480948194829483948494859486948794889489949094919492949394949495949694979498949995009501950295039504950595069507950895099510951195129513951495159516951795189519952095219522952395249525952695279528952995309531953295339534953595369537953895399540954195429543954495459546954795489549955095519552955395549555955695579558955995609561956295639564956595669567956895699570957195729573957495759576957795789579958095819582958395849585958695879588958995909591959295939594959595969597959895999600960196029603960496059606960796089609961096119612961396149615961696179618961996209621962296239624962596269627962896299630963196329633963496359636963796389639964096419642964396449645964696479648964996509651965296539654965596569657965896599660966196629663966496659666966796689669967096719672967396749675967696779678967996809681968296839684968596869687968896899690969196929693969496959696969796989699970097019702970397049705970697079708970997109711971297139714971597169717971897199720972197229723972497259726972797289729973097319732973397349735973697379738973997409741974297439744974597469747974897499750975197529753975497559756975797589759976097619762976397649765976697679768976997709771977297739774977597769777977897799780978197829783978497859786978797889789979097919792979397949795979697979798979998009801980298039804980598069807980898099810981198129813981498159816981798189819982098219822982398249825982698279828982998309831983298339834983598369837983898399840984198429843984498459846984798489849985098519852985398549855985698579858985998609861986298639864986598669867986898699870987198729873987498759876987798789879988098819882988398849885988698879888988998909891989298939894989598969897989898999900990199029903990499059906990799089909991099119912991399149915991699179918991999209921992299239924992599269927992899299930993199329933993499359936993799389939994099419942994399449945994699479948994999509951995299539954995599569957995899599960996199629963996499659966996799689969997099719972997399749975997699779978997999809981998299839984998599869987998899899990999199929993999499959996999799989999100001000110002100031000410005100061000710008100091001010011100121001310014100151001610017100181001910020100211002210023100241002510026100271002810029100301003110032100331003410035100361003710038100391004010041100421004310044100451004610047100481004910050100511005210053100541005510056100571005810059100601006110062100631006410065100661006710068100691007010071100721007310074100751007610077100781007910080100811008210083100841008510086100871008810089100901009110092100931009410095100961009710098100991010010101101021010310104101051010610107101081010910110101111011210113101141011510116101171011810119101201012110122101231012410125101261012710128101291013010131101321013310134101351013610137101381013910140101411014210143101441014510146101471014810149101501015110152101531015410155101561015710158101591016010161101621016310164101651016610167101681016910170101711017210173101741017510176101771017810179101801018110182101831018410185101861018710188101891019010191101921019310194101951019610197101981019910200102011020210203102041020510206102071020810209102101021110212102131021410215102161021710218102191022010221102221022310224102251022610227102281022910230102311023210233102341023510236102371023810239102401024110242102431024410245102461024710248102491025010251102521025310254102551025610257102581025910260102611026210263102641026510266102671026810269102701027110272102731027410275102761027710278102791028010281102821028310284102851028610287102881028910290102911029210293102941029510296102971029810299103001030110302103031030410305103061030710308103091031010311103121031310314103151031610317103181031910320103211032210323103241032510326103271032810329103301033110332103331033410335103361033710338103391034010341103421034310344103451034610347103481034910350103511035210353103541035510356103571035810359103601036110362103631036410365103661036710368103691037010371103721037310374103751037610377103781037910380103811038210383103841038510386103871038810389103901039110392103931039410395103961039710398103991040010401104021040310404104051040610407104081040910410104111041210413104141041510416104171041810419104201042110422104231042410425104261042710428104291043010431104321043310434104351043610437104381043910440104411044210443104441044510446104471044810449104501045110452104531045410455104561045710458104591046010461104621046310464104651046610467104681046910470104711047210473104741047510476104771047810479104801048110482104831048410485104861048710488104891049010491104921049310494104951049610497104981049910500105011050210503105041050510506105071050810509105101051110512105131051410515105161051710518105191052010521105221052310524105251052610527105281052910530105311053210533105341053510536105371053810539105401054110542105431054410545105461054710548105491055010551105521055310554105551055610557105581055910560105611056210563105641056510566105671056810569105701057110572105731057410575105761057710578105791058010581105821058310584105851058610587105881058910590105911059210593105941059510596105971059810599106001060110602106031060410605106061060710608106091061010611106121061310614106151061610617106181061910620106211062210623106241062510626106271062810629106301063110632106331063410635106361063710638106391064010641106421064310644106451064610647106481064910650106511065210653106541065510656106571065810659106601066110662106631066410665106661066710668106691067010671106721067310674106751067610677106781067910680106811068210683106841068510686106871068810689106901069110692106931069410695106961069710698106991070010701107021070310704107051070610707107081070910710107111071210713107141071510716107171071810719107201072110722107231072410725107261072710728107291073010731107321073310734107351073610737107381073910740107411074210743107441074510746107471074810749107501075110752107531075410755107561075710758107591076010761107621076310764107651076610767107681076910770107711077210773107741077510776107771077810779107801078110782107831078410785107861078710788107891079010791107921079310794107951079610797107981079910800108011080210803108041080510806108071080810809108101081110812108131081410815108161081710818108191082010821108221082310824108251082610827108281082910830108311083210833108341083510836108371083810839108401084110842108431084410845108461084710848108491085010851108521085310854108551085610857108581085910860108611086210863108641086510866108671086810869108701087110872108731087410875108761087710878108791088010881108821088310884108851088610887108881088910890108911089210893108941089510896108971089810899109001090110902109031090410905109061090710908109091091010911109121091310914109151091610917109181091910920109211092210923109241092510926109271092810929109301093110932109331093410935109361093710938109391094010941109421094310944109451094610947109481094910950109511095210953109541095510956109571095810959109601096110962109631096410965109661096710968109691097010971109721097310974109751097610977109781097910980109811098210983109841098510986109871098810989109901099110992109931099410995109961099710998109991100011001110021100311004110051100611007110081100911010110111101211013110141101511016110171101811019110201102111022110231102411025110261102711028110291103011031110321103311034110351103611037110381103911040110411104211043110441104511046110471104811049110501105111052110531105411055110561105711058110591106011061110621106311064110651106611067110681106911070110711107211073110741107511076110771107811079110801108111082110831108411085110861108711088110891109011091110921109311094110951109611097110981109911100111011110211103111041110511106111071110811109111101111111112111131111411115111161111711118111191112011121111221112311124111251112611127111281112911130111311113211133111341113511136111371113811139111401114111142111431114411145111461114711148111491115011151111521115311154111551115611157111581115911160111611116211163111641116511166111671116811169111701117111172111731117411175111761117711178111791118011181111821118311184111851118611187111881118911190111911119211193111941119511196111971119811199112001120111202112031120411205112061120711208112091121011211112121121311214112151121611217112181121911220112211122211223112241122511226112271122811229112301123111232112331123411235112361123711238112391124011241112421124311244112451124611247112481124911250112511125211253112541125511256112571125811259112601126111262112631126411265112661126711268112691127011271112721127311274112751127611277112781127911280112811128211283112841128511286112871128811289112901129111292112931129411295112961129711298112991130011301113021130311304113051130611307113081130911310113111131211313113141131511316113171131811319113201132111322113231132411325113261132711328113291133011331113321133311334113351133611337113381133911340113411134211343113441134511346113471134811349113501135111352113531135411355113561135711358113591136011361113621136311364113651136611367113681136911370113711137211373113741137511376113771137811379113801138111382113831138411385113861138711388113891139011391113921139311394113951139611397113981139911400114011140211403114041140511406114071140811409114101141111412114131141411415114161141711418114191142011421114221142311424114251142611427114281142911430114311143211433114341143511436114371143811439114401144111442114431144411445114461144711448114491145011451114521145311454114551145611457114581145911460114611146211463114641146511466114671146811469114701147111472114731147411475114761147711478114791148011481114821148311484114851148611487114881148911490114911149211493114941149511496114971149811499115001150111502115031150411505115061150711508115091151011511115121151311514115151151611517115181151911520115211152211523115241152511526115271152811529115301153111532115331153411535115361153711538115391154011541115421154311544115451154611547115481154911550115511155211553115541155511556115571155811559115601156111562115631156411565115661156711568115691157011571115721157311574115751157611577115781157911580115811158211583115841158511586115871158811589115901159111592115931159411595115961159711598115991160011601116021160311604116051160611607116081160911610116111161211613116141161511616116171161811619116201162111622116231162411625116261162711628116291163011631116321163311634116351163611637116381163911640116411164211643116441164511646116471164811649116501165111652116531165411655116561165711658116591166011661116621166311664116651166611667116681166911670116711167211673116741167511676116771167811679116801168111682116831168411685116861168711688116891169011691116921169311694116951169611697116981169911700117011170211703117041170511706117071170811709117101171111712117131171411715117161171711718117191172011721117221172311724117251172611727117281172911730117311173211733117341173511736117371173811739117401174111742117431174411745117461174711748117491175011751117521175311754117551175611757117581175911760117611176211763117641176511766117671176811769117701177111772117731177411775117761177711778117791178011781117821178311784117851178611787117881178911790117911179211793117941179511796117971179811799118001180111802118031180411805118061180711808118091181011811118121181311814118151181611817118181181911820118211182211823118241182511826118271182811829118301183111832118331183411835118361183711838118391184011841118421184311844118451184611847118481184911850118511185211853118541185511856118571185811859118601186111862118631186411865118661186711868118691187011871118721187311874118751187611877118781187911880118811188211883118841188511886118871188811889118901189111892118931189411895118961189711898118991190011901119021190311904119051190611907119081190911910119111191211913119141191511916119171191811919119201192111922119231192411925119261192711928119291193011931119321193311934119351193611937119381193911940119411194211943119441194511946119471194811949119501195111952119531195411955119561195711958119591196011961119621196311964119651196611967119681196911970119711197211973119741197511976119771197811979119801198111982119831198411985119861198711988119891199011991119921199311994119951199611997119981199912000120011200212003120041200512006120071200812009120101201112012120131201412015120161201712018120191202012021120221202312024120251202612027120281202912030120311203212033120341203512036120371203812039120401204112042120431204412045120461204712048120491205012051120521205312054120551205612057120581205912060120611206212063120641206512066120671206812069120701207112072120731207412075120761207712078120791208012081120821208312084120851208612087120881208912090120911209212093120941209512096120971209812099121001210112102121031210412105121061210712108121091211012111121121211312114121151211612117121181211912120121211212212123121241212512126121271212812129121301213112132121331213412135121361213712138121391214012141121421214312144121451214612147121481214912150121511215212153121541215512156121571215812159121601216112162121631216412165121661216712168121691217012171121721217312174121751217612177121781217912180121811218212183121841218512186121871218812189121901219112192121931219412195121961219712198121991220012201122021220312204122051220612207122081220912210122111221212213122141221512216122171221812219122201222112222122231222412225122261222712228122291223012231122321223312234122351223612237122381223912240122411224212243122441224512246122471224812249122501225112252122531225412255122561225712258122591226012261122621226312264122651226612267122681226912270122711227212273122741227512276122771227812279122801228112282122831228412285122861228712288122891229012291122921229312294122951229612297122981229912300123011230212303123041230512306123071230812309123101231112312123131231412315123161231712318123191232012321123221232312324123251232612327123281232912330123311233212333123341233512336123371233812339123401234112342123431234412345123461234712348123491235012351123521235312354123551235612357123581235912360123611236212363123641236512366123671236812369123701237112372123731237412375123761237712378123791238012381123821238312384123851238612387123881238912390123911239212393123941239512396123971239812399124001240112402124031240412405124061240712408124091241012411124121241312414124151241612417124181241912420124211242212423124241242512426124271242812429124301243112432124331243412435124361243712438124391244012441124421244312444124451244612447124481244912450124511245212453124541245512456124571245812459124601246112462124631246412465124661246712468124691247012471124721247312474124751247612477124781247912480124811248212483124841248512486124871248812489124901249112492124931249412495124961249712498124991250012501125021250312504125051250612507125081250912510125111251212513125141251512516125171251812519125201252112522125231252412525125261252712528125291253012531125321253312534125351253612537125381253912540125411254212543125441254512546125471254812549125501255112552125531255412555125561255712558125591256012561125621256312564125651256612567125681256912570125711257212573125741257512576125771257812579125801258112582125831258412585125861258712588125891259012591125921259312594125951259612597125981259912600126011260212603126041260512606126071260812609126101261112612126131261412615126161261712618126191262012621126221262312624126251262612627126281262912630126311263212633126341263512636126371263812639126401264112642126431264412645126461264712648126491265012651126521265312654126551265612657126581265912660126611266212663126641266512666126671266812669126701267112672126731267412675126761267712678126791268012681126821268312684126851268612687126881268912690126911269212693126941269512696126971269812699127001270112702127031270412705127061270712708127091271012711127121271312714127151271612717127181271912720127211272212723127241272512726127271272812729127301273112732127331273412735127361273712738127391274012741127421274312744127451274612747127481274912750127511275212753127541275512756127571275812759127601276112762127631276412765127661276712768127691277012771127721277312774127751277612777127781277912780127811278212783127841278512786127871278812789127901279112792127931279412795127961279712798127991280012801128021280312804128051280612807128081280912810128111281212813128141281512816128171281812819128201282112822128231282412825128261282712828128291283012831128321283312834128351283612837128381283912840128411284212843128441284512846128471284812849128501285112852128531285412855128561285712858128591286012861128621286312864128651286612867128681286912870128711287212873128741287512876128771287812879128801288112882128831288412885128861288712888128891289012891128921289312894128951289612897128981289912900129011290212903129041290512906129071290812909129101291112912129131291412915129161291712918129191292012921129221292312924129251292612927129281292912930129311293212933129341293512936129371293812939129401294112942129431294412945129461294712948129491295012951129521295312954129551295612957129581295912960129611296212963129641296512966129671296812969129701297112972129731297412975129761297712978129791298012981129821298312984129851298612987129881298912990129911299212993129941299512996129971299812999130001300113002130031300413005130061300713008130091301013011130121301313014130151301613017130181301913020130211302213023130241302513026130271302813029130301303113032130331303413035130361303713038130391304013041130421304313044130451304613047130481304913050130511305213053130541305513056130571305813059130601306113062130631306413065130661306713068130691307013071130721307313074130751307613077130781307913080130811308213083130841308513086130871308813089130901309113092130931309413095130961309713098130991310013101131021310313104131051310613107131081310913110131111311213113131141311513116131171311813119131201312113122131231312413125131261312713128131291313013131131321313313134131351313613137131381313913140131411314213143131441314513146131471314813149131501315113152131531315413155131561315713158131591316013161131621316313164131651316613167131681316913170131711317213173131741317513176131771317813179131801318113182131831318413185131861318713188131891319013191131921319313194131951319613197131981319913200132011320213203132041320513206132071320813209132101321113212132131321413215132161321713218132191322013221132221322313224132251322613227132281322913230132311323213233132341323513236132371323813239132401324113242132431324413245132461324713248132491325013251132521325313254132551325613257132581325913260132611326213263132641326513266132671326813269132701327113272132731327413275132761327713278132791328013281132821328313284132851328613287132881328913290132911329213293132941329513296132971329813299133001330113302133031330413305133061330713308133091331013311133121331313314133151331613317133181331913320133211332213323133241332513326133271332813329133301333113332133331333413335133361333713338133391334013341133421334313344133451334613347133481334913350133511335213353133541335513356133571335813359133601336113362133631336413365133661336713368133691337013371133721337313374133751337613377133781337913380133811338213383133841338513386133871338813389133901339113392133931339413395133961339713398133991340013401134021340313404134051340613407134081340913410134111341213413134141341513416134171341813419134201342113422134231342413425134261342713428134291343013431134321343313434134351343613437134381343913440134411344213443134441344513446134471344813449134501345113452134531345413455134561345713458134591346013461134621346313464134651346613467134681346913470134711347213473134741347513476134771347813479134801348113482134831348413485134861348713488134891349013491134921349313494134951349613497134981349913500135011350213503135041350513506135071350813509135101351113512135131351413515135161351713518135191352013521135221352313524135251352613527135281352913530135311353213533135341353513536135371353813539135401354113542135431354413545135461354713548135491355013551135521355313554135551355613557135581355913560135611356213563135641356513566135671356813569135701357113572135731357413575135761357713578135791358013581135821358313584135851358613587135881358913590135911359213593135941359513596135971359813599136001360113602136031360413605136061360713608136091361013611136121361313614136151361613617136181361913620136211362213623136241362513626136271362813629136301363113632136331363413635136361363713638136391364013641136421364313644136451364613647136481364913650136511365213653136541365513656136571365813659136601366113662136631366413665136661366713668136691367013671136721367313674136751367613677136781367913680136811368213683136841368513686136871368813689136901369113692136931369413695136961369713698136991370013701137021370313704137051370613707137081370913710137111371213713137141371513716137171371813719137201372113722137231372413725137261372713728137291373013731137321373313734137351373613737137381373913740137411374213743137441374513746137471374813749137501375113752137531375413755137561375713758137591376013761137621376313764137651376613767137681376913770137711377213773137741377513776137771377813779137801378113782137831378413785137861378713788137891379013791137921379313794137951379613797137981379913800138011380213803138041380513806138071380813809138101381113812138131381413815138161381713818138191382013821138221382313824138251382613827138281382913830138311383213833138341383513836138371383813839138401384113842138431384413845138461384713848138491385013851138521385313854138551385613857138581385913860138611386213863138641386513866138671386813869138701387113872138731387413875138761387713878138791388013881138821388313884138851388613887138881388913890138911389213893138941389513896138971389813899139001390113902139031390413905139061390713908139091391013911139121391313914139151391613917139181391913920139211392213923139241392513926139271392813929139301393113932139331393413935139361393713938139391394013941139421394313944139451394613947139481394913950139511395213953139541395513956139571395813959139601396113962139631396413965139661396713968139691397013971139721397313974139751397613977139781397913980139811398213983139841398513986139871398813989139901399113992139931399413995139961399713998139991400014001140021400314004140051400614007140081400914010140111401214013140141401514016140171401814019140201402114022140231402414025140261402714028140291403014031140321403314034140351403614037140381403914040140411404214043140441404514046140471404814049140501405114052140531405414055140561405714058140591406014061140621406314064140651406614067140681406914070140711407214073140741407514076140771407814079140801408114082140831408414085140861408714088140891409014091140921409314094140951409614097140981409914100141011410214103141041410514106141071410814109141101411114112141131411414115141161411714118141191412014121141221412314124141251412614127141281412914130141311413214133141341413514136141371413814139141401414114142141431414414145141461414714148141491415014151141521415314154141551415614157141581415914160141611416214163141641416514166141671416814169141701417114172141731417414175141761417714178141791418014181141821418314184141851418614187141881418914190141911419214193141941419514196141971419814199142001420114202142031420414205142061420714208142091421014211142121421314214142151421614217142181421914220142211422214223142241422514226142271422814229142301423114232142331423414235142361423714238142391424014241142421424314244142451424614247142481424914250142511425214253142541425514256142571425814259142601426114262142631426414265142661426714268142691427014271142721427314274142751427614277142781427914280142811428214283142841428514286142871428814289142901429114292142931429414295142961429714298142991430014301143021430314304143051430614307143081430914310143111431214313143141431514316143171431814319143201432114322143231432414325143261432714328143291433014331143321433314334143351433614337143381433914340143411434214343143441434514346143471434814349143501435114352143531435414355143561435714358143591436014361143621436314364143651436614367143681436914370143711437214373143741437514376143771437814379143801438114382143831438414385143861438714388143891439014391143921439314394143951439614397143981439914400144011440214403144041440514406144071440814409144101441114412144131441414415144161441714418144191442014421144221442314424144251442614427144281442914430144311443214433144341443514436144371443814439144401444114442144431444414445144461444714448144491445014451144521445314454144551445614457144581445914460144611446214463144641446514466144671446814469144701447114472144731447414475144761447714478144791448014481144821448314484144851448614487144881448914490144911449214493144941449514496144971449814499145001450114502145031450414505145061450714508145091451014511145121451314514145151451614517145181451914520145211452214523145241452514526145271452814529145301453114532145331453414535145361453714538145391454014541145421454314544145451454614547145481454914550145511455214553145541455514556145571455814559145601456114562145631456414565145661456714568145691457014571145721457314574145751457614577145781457914580145811458214583145841458514586145871458814589145901459114592145931459414595145961459714598145991460014601146021460314604146051460614607146081460914610146111461214613146141461514616146171461814619146201462114622146231462414625146261462714628146291463014631146321463314634146351463614637146381463914640146411464214643146441464514646146471464814649146501465114652146531465414655146561465714658146591466014661146621466314664146651466614667146681466914670146711467214673146741467514676146771467814679146801468114682146831468414685146861468714688146891469014691146921469314694146951469614697146981469914700147011470214703147041470514706147071470814709147101471114712147131471414715147161471714718147191472014721147221472314724147251472614727147281472914730147311473214733147341473514736147371473814739147401474114742147431474414745147461474714748147491475014751147521475314754147551475614757147581475914760147611476214763147641476514766147671476814769147701477114772147731477414775147761477714778147791478014781147821478314784147851478614787147881478914790147911479214793147941479514796147971479814799148001480114802148031480414805148061480714808148091481014811148121481314814148151481614817148181481914820148211482214823148241482514826148271482814829148301483114832148331483414835148361483714838148391484014841148421484314844148451484614847148481484914850148511485214853148541485514856148571485814859148601486114862148631486414865148661486714868148691487014871148721487314874148751487614877148781487914880148811488214883148841488514886148871488814889148901489114892148931489414895148961489714898148991490014901149021490314904149051490614907149081490914910149111491214913149141491514916149171491814919149201492114922149231492414925149261492714928149291493014931149321493314934149351493614937149381493914940149411494214943149441494514946149471494814949149501495114952149531495414955149561495714958149591496014961149621496314964149651496614967149681496914970149711497214973149741497514976149771497814979149801498114982149831498414985149861498714988149891499014991149921499314994149951499614997149981499915000150011500215003150041500515006150071500815009150101501115012150131501415015150161501715018150191502015021150221502315024150251502615027150281502915030150311503215033150341503515036150371503815039150401504115042150431504415045150461504715048150491505015051150521505315054150551505615057150581505915060150611506215063150641506515066150671506815069150701507115072150731507415075150761507715078150791508015081150821508315084150851508615087150881508915090150911509215093150941509515096150971509815099151001510115102151031510415105151061510715108151091511015111151121511315114151151511615117151181511915120151211512215123151241512515126151271512815129151301513115132151331513415135151361513715138151391514015141151421514315144151451514615147151481514915150151511515215153151541515515156151571515815159151601516115162151631516415165151661516715168151691517015171151721517315174151751517615177151781517915180151811518215183151841518515186151871518815189151901519115192151931519415195151961519715198151991520015201152021520315204152051520615207152081520915210152111521215213152141521515216152171521815219152201522115222152231522415225152261522715228152291523015231152321523315234152351523615237152381523915240152411524215243152441524515246152471524815249152501525115252152531525415255152561525715258152591526015261152621526315264152651526615267152681526915270152711527215273152741527515276152771527815279152801528115282152831528415285152861528715288152891529015291152921529315294152951529615297152981529915300153011530215303153041530515306153071530815309153101531115312153131531415315153161531715318153191532015321153221532315324153251532615327153281532915330153311533215333153341533515336153371533815339153401534115342153431534415345153461534715348153491535015351153521535315354153551535615357153581535915360153611536215363153641536515366153671536815369153701537115372153731537415375153761537715378153791538015381153821538315384153851538615387153881538915390153911539215393153941539515396153971539815399154001540115402154031540415405154061540715408154091541015411154121541315414154151541615417154181541915420154211542215423154241542515426154271542815429154301543115432154331543415435154361543715438154391544015441154421544315444154451544615447154481544915450154511545215453154541545515456154571545815459154601546115462154631546415465154661546715468154691547015471154721547315474154751547615477154781547915480154811548215483154841548515486154871548815489154901549115492154931549415495154961549715498154991550015501155021550315504155051550615507155081550915510155111551215513155141551515516155171551815519155201552115522155231552415525155261552715528155291553015531155321553315534155351553615537155381553915540155411554215543155441554515546155471554815549155501555115552155531555415555155561555715558155591556015561155621556315564155651556615567155681556915570155711557215573155741557515576155771557815579155801558115582155831558415585155861558715588155891559015591155921559315594155951559615597155981559915600156011560215603156041560515606156071560815609156101561115612156131561415615156161561715618156191562015621156221562315624156251562615627156281562915630156311563215633156341563515636156371563815639156401564115642156431564415645156461564715648156491565015651156521565315654156551565615657156581565915660156611566215663156641566515666156671566815669156701567115672156731567415675156761567715678156791568015681156821568315684156851568615687156881568915690156911569215693156941569515696156971569815699157001570115702157031570415705157061570715708157091571015711157121571315714157151571615717157181571915720157211572215723157241572515726157271572815729157301573115732157331573415735157361573715738157391574015741157421574315744157451574615747157481574915750157511575215753157541575515756157571575815759157601576115762157631576415765157661576715768157691577015771157721577315774157751577615777157781577915780157811578215783157841578515786157871578815789157901579115792157931579415795157961579715798157991580015801158021580315804158051580615807158081580915810158111581215813158141581515816158171581815819158201582115822158231582415825158261582715828158291583015831158321583315834158351583615837158381583915840158411584215843158441584515846158471584815849158501585115852158531585415855158561585715858158591586015861158621586315864158651586615867158681586915870158711587215873158741587515876158771587815879158801588115882158831588415885158861588715888158891589015891158921589315894158951589615897158981589915900159011590215903159041590515906159071590815909159101591115912159131591415915159161591715918159191592015921159221592315924159251592615927159281592915930159311593215933159341593515936159371593815939159401594115942159431594415945159461594715948159491595015951159521595315954159551595615957159581595915960159611596215963159641596515966159671596815969159701597115972159731597415975159761597715978159791598015981159821598315984159851598615987159881598915990159911599215993159941599515996159971599815999160001600116002160031600416005160061600716008160091601016011160121601316014160151601616017160181601916020160211602216023160241602516026160271602816029160301603116032160331603416035160361603716038160391604016041160421604316044160451604616047160481604916050160511605216053160541605516056160571605816059160601606116062160631606416065160661606716068160691607016071160721607316074160751607616077160781607916080160811608216083160841608516086160871608816089160901609116092160931609416095160961609716098160991610016101161021610316104161051610616107161081610916110161111611216113161141611516116161171611816119161201612116122161231612416125161261612716128161291613016131161321613316134161351613616137161381613916140161411614216143161441614516146161471614816149161501615116152161531615416155161561615716158161591616016161161621616316164161651616616167161681616916170161711617216173161741617516176161771617816179161801618116182161831618416185161861618716188161891619016191161921619316194161951619616197161981619916200162011620216203162041620516206162071620816209162101621116212162131621416215162161621716218162191622016221162221622316224162251622616227162281622916230162311623216233162341623516236162371623816239162401624116242162431624416245162461624716248162491625016251162521625316254162551625616257162581625916260162611626216263162641626516266162671626816269162701627116272162731627416275162761627716278162791628016281162821628316284162851628616287162881628916290162911629216293162941629516296162971629816299163001630116302163031630416305163061630716308163091631016311163121631316314163151631616317163181631916320163211632216323163241632516326163271632816329163301633116332163331633416335163361633716338163391634016341163421634316344163451634616347163481634916350163511635216353163541635516356163571635816359163601636116362163631636416365163661636716368163691637016371163721637316374163751637616377163781637916380163811638216383163841638516386163871638816389163901639116392163931639416395163961639716398163991640016401164021640316404164051640616407164081640916410164111641216413164141641516416164171641816419164201642116422164231642416425164261642716428164291643016431164321643316434164351643616437164381643916440164411644216443164441644516446164471644816449164501645116452164531645416455164561645716458164591646016461164621646316464164651646616467164681646916470164711647216473164741647516476164771647816479164801648116482164831648416485164861648716488164891649016491164921649316494164951649616497164981649916500165011650216503165041650516506165071650816509165101651116512165131651416515165161651716518165191652016521165221652316524165251652616527165281652916530165311653216533165341653516536165371653816539165401654116542165431654416545165461654716548165491655016551165521655316554165551655616557165581655916560165611656216563165641656516566165671656816569165701657116572165731657416575165761657716578165791658016581165821658316584165851658616587165881658916590165911659216593165941659516596165971659816599166001660116602166031660416605166061660716608166091661016611166121661316614166151661616617166181661916620166211662216623166241662516626166271662816629166301663116632166331663416635166361663716638166391664016641166421664316644166451664616647166481664916650166511665216653166541665516656166571665816659166601666116662166631666416665166661666716668166691667016671166721667316674166751667616677166781667916680166811668216683166841668516686166871668816689166901669116692166931669416695166961669716698166991670016701167021670316704167051670616707167081670916710167111671216713167141671516716167171671816719167201672116722167231672416725167261672716728167291673016731167321673316734167351673616737167381673916740167411674216743167441674516746167471674816749167501675116752167531675416755167561675716758167591676016761167621676316764167651676616767167681676916770167711677216773167741677516776167771677816779167801678116782167831678416785167861678716788167891679016791167921679316794167951679616797167981679916800168011680216803168041680516806168071680816809168101681116812168131681416815168161681716818168191682016821168221682316824168251682616827168281682916830168311683216833168341683516836168371683816839168401684116842168431684416845168461684716848168491685016851168521685316854168551685616857168581685916860168611686216863168641686516866168671686816869168701687116872168731687416875168761687716878168791688016881168821688316884168851688616887168881688916890168911689216893168941689516896168971689816899169001690116902169031690416905169061690716908169091691016911169121691316914169151691616917169181691916920169211692216923169241692516926169271692816929169301693116932169331693416935169361693716938169391694016941169421694316944169451694616947169481694916950169511695216953169541695516956169571695816959169601696116962169631696416965169661696716968169691697016971169721697316974169751697616977169781697916980169811698216983169841698516986169871698816989169901699116992169931699416995169961699716998169991700017001170021700317004170051700617007170081700917010170111701217013170141701517016170171701817019170201702117022170231702417025170261702717028170291703017031170321703317034170351703617037170381703917040170411704217043170441704517046170471704817049170501705117052170531705417055170561705717058170591706017061170621706317064170651706617067170681706917070170711707217073170741707517076170771707817079170801708117082170831708417085170861708717088170891709017091170921709317094170951709617097170981709917100171011710217103171041710517106171071710817109171101711117112171131711417115171161711717118171191712017121171221712317124171251712617127171281712917130171311713217133171341713517136171371713817139171401714117142171431714417145171461714717148171491715017151171521715317154171551715617157171581715917160171611716217163171641716517166171671716817169171701717117172171731717417175171761717717178171791718017181171821718317184171851718617187171881718917190171911719217193171941719517196171971719817199172001720117202172031720417205172061720717208172091721017211172121721317214172151721617217172181721917220172211722217223172241722517226172271722817229172301723117232172331723417235172361723717238172391724017241172421724317244172451724617247172481724917250172511725217253172541725517256172571725817259172601726117262172631726417265172661726717268172691727017271172721727317274172751727617277172781727917280172811728217283172841728517286172871728817289172901729117292172931729417295172961729717298172991730017301173021730317304173051730617307173081730917310173111731217313173141731517316173171731817319173201732117322173231732417325173261732717328173291733017331173321733317334173351733617337173381733917340173411734217343173441734517346173471734817349173501735117352173531735417355173561735717358173591736017361173621736317364173651736617367173681736917370173711737217373173741737517376173771737817379173801738117382173831738417385173861738717388173891739017391173921739317394173951739617397173981739917400174011740217403174041740517406174071740817409174101741117412174131741417415174161741717418174191742017421174221742317424174251742617427174281742917430174311743217433174341743517436174371743817439174401744117442174431744417445174461744717448174491745017451174521745317454174551745617457174581745917460174611746217463174641746517466174671746817469174701747117472174731747417475174761747717478174791748017481174821748317484174851748617487174881748917490174911749217493174941749517496174971749817499175001750117502175031750417505175061750717508175091751017511175121751317514175151751617517175181751917520175211752217523175241752517526175271752817529175301753117532175331753417535175361753717538175391754017541175421754317544175451754617547175481754917550175511755217553175541755517556175571755817559175601756117562175631756417565175661756717568175691757017571175721757317574175751757617577175781757917580175811758217583175841758517586175871758817589175901759117592175931759417595175961759717598175991760017601176021760317604176051760617607176081760917610176111761217613176141761517616176171761817619176201762117622176231762417625176261762717628176291763017631176321763317634176351763617637176381763917640176411764217643176441764517646176471764817649176501765117652176531765417655176561765717658176591766017661176621766317664176651766617667176681766917670176711767217673176741767517676176771767817679176801768117682176831768417685176861768717688176891769017691176921769317694176951769617697176981769917700177011770217703177041770517706177071770817709177101771117712177131771417715177161771717718177191772017721177221772317724177251772617727177281772917730177311773217733177341773517736177371773817739177401774117742177431774417745177461774717748177491775017751177521775317754177551775617757177581775917760177611776217763177641776517766177671776817769177701777117772177731777417775177761777717778177791778017781177821778317784177851778617787177881778917790177911779217793177941779517796177971779817799178001780117802178031780417805178061780717808178091781017811178121781317814178151781617817178181781917820178211782217823178241782517826178271782817829178301783117832178331783417835178361783717838178391784017841178421784317844178451784617847178481784917850178511785217853178541785517856178571785817859178601786117862178631786417865178661786717868178691787017871178721787317874178751787617877178781787917880178811788217883178841788517886178871788817889178901789117892178931789417895178961789717898178991790017901179021790317904179051790617907179081790917910179111791217913179141791517916179171791817919179201792117922179231792417925179261792717928179291793017931179321793317934179351793617937179381793917940179411794217943179441794517946179471794817949179501795117952179531795417955179561795717958179591796017961179621796317964179651796617967179681796917970179711797217973179741797517976179771797817979179801798117982179831798417985179861798717988179891799017991179921799317994179951799617997179981799918000180011800218003180041800518006180071800818009180101801118012180131801418015180161801718018180191802018021180221802318024180251802618027180281802918030180311803218033180341803518036180371803818039180401804118042180431804418045180461804718048180491805018051180521805318054180551805618057180581805918060180611806218063180641806518066180671806818069180701807118072180731807418075180761807718078180791808018081180821808318084180851808618087180881808918090180911809218093180941809518096180971809818099181001810118102181031810418105181061810718108181091811018111181121811318114181151811618117181181811918120181211812218123181241812518126181271812818129181301813118132181331813418135181361813718138181391814018141181421814318144181451814618147181481814918150181511815218153181541815518156181571815818159181601816118162181631816418165181661816718168181691817018171181721817318174181751817618177181781817918180181811818218183181841818518186181871818818189181901819118192181931819418195181961819718198181991820018201182021820318204182051820618207182081820918210182111821218213182141821518216182171821818219182201822118222182231822418225182261822718228182291823018231182321823318234182351823618237182381823918240182411824218243182441824518246182471824818249182501825118252182531825418255182561825718258182591826018261182621826318264182651826618267182681826918270182711827218273182741827518276182771827818279182801828118282182831828418285182861828718288182891829018291182921829318294182951829618297182981829918300183011830218303183041830518306183071830818309183101831118312183131831418315183161831718318183191832018321183221832318324183251832618327183281832918330183311833218333183341833518336183371833818339183401834118342183431834418345183461834718348183491835018351183521835318354183551835618357183581835918360183611836218363183641836518366183671836818369183701837118372183731837418375183761837718378183791838018381183821838318384183851838618387183881838918390183911839218393183941839518396183971839818399184001840118402184031840418405184061840718408184091841018411184121841318414184151841618417184181841918420184211842218423184241842518426184271842818429184301843118432184331843418435184361843718438184391844018441184421844318444184451844618447184481844918450184511845218453184541845518456184571845818459184601846118462184631846418465184661846718468184691847018471184721847318474184751847618477184781847918480184811848218483184841848518486184871848818489184901849118492184931849418495184961849718498184991850018501185021850318504185051850618507185081850918510185111851218513185141851518516185171851818519185201852118522185231852418525185261852718528185291853018531185321853318534185351853618537185381853918540185411854218543185441854518546185471854818549185501855118552185531855418555185561855718558185591856018561185621856318564185651856618567185681856918570185711857218573185741857518576185771857818579185801858118582185831858418585185861858718588185891859018591185921859318594185951859618597185981859918600186011860218603186041860518606186071860818609186101861118612186131861418615186161861718618186191862018621186221862318624186251862618627186281862918630186311863218633186341863518636186371863818639186401864118642186431864418645186461864718648186491865018651186521865318654186551865618657186581865918660186611866218663186641866518666186671866818669186701867118672186731867418675186761867718678186791868018681186821868318684186851868618687186881868918690186911869218693186941869518696186971869818699187001870118702187031870418705187061870718708187091871018711187121871318714187151871618717187181871918720187211872218723187241872518726187271872818729187301873118732187331873418735187361873718738187391874018741187421874318744187451874618747187481874918750187511875218753187541875518756187571875818759187601876118762187631876418765187661876718768187691877018771187721877318774187751877618777187781877918780187811878218783187841878518786187871878818789187901879118792187931879418795187961879718798187991880018801188021880318804188051880618807188081880918810188111881218813188141881518816188171881818819188201882118822188231882418825188261882718828188291883018831188321883318834188351883618837188381883918840188411884218843188441884518846188471884818849188501885118852188531885418855188561885718858188591886018861188621886318864188651886618867188681886918870188711887218873188741887518876188771887818879188801888118882188831888418885188861888718888188891889018891188921889318894188951889618897188981889918900189011890218903189041890518906189071890818909189101891118912189131891418915189161891718918189191892018921189221892318924189251892618927189281892918930189311893218933189341893518936189371893818939189401894118942189431894418945189461894718948189491895018951189521895318954189551895618957189581895918960189611896218963189641896518966189671896818969189701897118972189731897418975189761897718978189791898018981189821898318984189851898618987189881898918990189911899218993189941899518996189971899818999190001900119002190031900419005190061900719008190091901019011190121901319014190151901619017190181901919020190211902219023190241902519026190271902819029190301903119032190331903419035190361903719038190391904019041190421904319044190451904619047190481904919050190511905219053190541905519056190571905819059190601906119062190631906419065190661906719068190691907019071190721907319074190751907619077190781907919080190811908219083190841908519086190871908819089190901909119092190931909419095190961909719098190991910019101191021910319104191051910619107191081910919110191111911219113191141911519116191171911819119191201912119122191231912419125191261912719128191291913019131191321913319134191351913619137191381913919140191411914219143191441914519146191471914819149191501915119152191531915419155191561915719158191591916019161191621916319164191651916619167191681916919170191711917219173191741917519176191771917819179191801918119182191831918419185191861918719188191891919019191191921919319194191951919619197191981919919200192011920219203192041920519206192071920819209192101921119212192131921419215192161921719218192191922019221192221922319224192251922619227192281922919230192311923219233192341923519236192371923819239192401924119242192431924419245192461924719248192491925019251192521925319254192551925619257192581925919260192611926219263192641926519266192671926819269192701927119272192731927419275192761927719278192791928019281192821928319284192851928619287192881928919290192911929219293192941929519296192971929819299193001930119302193031930419305193061930719308193091931019311193121931319314193151931619317193181931919320193211932219323193241932519326193271932819329193301933119332193331933419335193361933719338193391934019341193421934319344193451934619347193481934919350193511935219353193541935519356193571935819359193601936119362193631936419365193661936719368193691937019371193721937319374193751937619377193781937919380193811938219383193841938519386193871938819389193901939119392193931939419395193961939719398193991940019401194021940319404194051940619407194081940919410194111941219413194141941519416194171941819419194201942119422194231942419425194261942719428194291943019431194321943319434194351943619437194381943919440194411944219443194441944519446194471944819449194501945119452194531945419455194561945719458194591946019461194621946319464194651946619467194681946919470194711947219473194741947519476194771947819479194801948119482194831948419485194861948719488194891949019491194921949319494194951949619497194981949919500195011950219503195041950519506195071950819509195101951119512195131951419515195161951719518195191952019521195221952319524195251952619527195281952919530195311953219533195341953519536195371953819539195401954119542195431954419545195461954719548195491955019551195521955319554195551955619557195581955919560195611956219563195641956519566195671956819569195701957119572195731957419575195761957719578195791958019581195821958319584195851958619587195881958919590195911959219593195941959519596195971959819599196001960119602196031960419605196061960719608196091961019611196121961319614196151961619617196181961919620196211962219623196241962519626196271962819629196301963119632196331963419635196361963719638196391964019641196421964319644196451964619647196481964919650196511965219653196541965519656196571965819659196601966119662196631966419665196661966719668196691967019671196721967319674196751967619677196781967919680196811968219683196841968519686196871968819689196901969119692196931969419695196961969719698196991970019701197021970319704197051970619707197081970919710197111971219713197141971519716197171971819719197201972119722197231972419725197261972719728197291973019731197321973319734197351973619737197381973919740197411974219743197441974519746197471974819749197501975119752197531975419755197561975719758197591976019761197621976319764197651976619767197681976919770197711977219773197741977519776197771977819779197801978119782197831978419785197861978719788197891979019791197921979319794197951979619797197981979919800198011980219803198041980519806198071980819809198101981119812198131981419815198161981719818198191982019821198221982319824198251982619827198281982919830198311983219833198341983519836198371983819839198401984119842198431984419845198461984719848198491985019851198521985319854198551985619857198581985919860198611986219863198641986519866198671986819869198701987119872198731987419875198761987719878198791988019881198821988319884198851988619887198881988919890198911989219893198941989519896198971989819899199001990119902199031990419905199061990719908199091991019911199121991319914199151991619917199181991919920199211992219923199241992519926199271992819929199301993119932199331993419935199361993719938199391994019941199421994319944199451994619947199481994919950199511995219953199541995519956199571995819959199601996119962199631996419965199661996719968199691997019971199721997319974199751997619977199781997919980199811998219983199841998519986199871998819989199901999119992199931999419995199961999719998199992000020001200022000320004200052000620007200082000920010200112001220013200142001520016200172001820019200202002120022200232002420025200262002720028200292003020031200322003320034200352003620037200382003920040200412004220043200442004520046200472004820049200502005120052200532005420055200562005720058200592006020061200622006320064200652006620067200682006920070200712007220073200742007520076200772007820079200802008120082200832008420085200862008720088200892009020091200922009320094200952009620097200982009920100201012010220103201042010520106201072010820109201102011120112201132011420115201162011720118201192012020121201222012320124201252012620127201282012920130201312013220133201342013520136201372013820139201402014120142201432014420145201462014720148201492015020151201522015320154201552015620157201582015920160201612016220163201642016520166201672016820169201702017120172201732017420175201762017720178201792018020181201822018320184201852018620187201882018920190201912019220193201942019520196201972019820199202002020120202202032020420205202062020720208202092021020211202122021320214202152021620217202182021920220202212022220223202242022520226202272022820229202302023120232202332023420235202362023720238202392024020241202422024320244202452024620247202482024920250202512025220253202542025520256202572025820259202602026120262202632026420265202662026720268202692027020271202722027320274202752027620277202782027920280202812028220283202842028520286202872028820289202902029120292202932029420295202962029720298202992030020301203022030320304203052030620307203082030920310203112031220313203142031520316203172031820319203202032120322203232032420325203262032720328203292033020331203322033320334203352033620337203382033920340203412034220343203442034520346203472034820349203502035120352203532035420355203562035720358203592036020361203622036320364203652036620367203682036920370203712037220373203742037520376203772037820379203802038120382203832038420385203862038720388203892039020391203922039320394203952039620397203982039920400204012040220403204042040520406204072040820409204102041120412204132041420415204162041720418204192042020421204222042320424204252042620427204282042920430204312043220433204342043520436204372043820439204402044120442204432044420445204462044720448204492045020451204522045320454204552045620457204582045920460204612046220463204642046520466204672046820469204702047120472204732047420475204762047720478204792048020481204822048320484204852048620487204882048920490204912049220493204942049520496204972049820499205002050120502205032050420505205062050720508205092051020511205122051320514205152051620517205182051920520205212052220523205242052520526205272052820529205302053120532205332053420535205362053720538205392054020541205422054320544205452054620547205482054920550205512055220553205542055520556205572055820559205602056120562205632056420565205662056720568205692057020571205722057320574205752057620577205782057920580205812058220583205842058520586205872058820589205902059120592205932059420595205962059720598205992060020601206022060320604206052060620607206082060920610206112061220613206142061520616206172061820619206202062120622206232062420625206262062720628206292063020631206322063320634206352063620637206382063920640206412064220643206442064520646206472064820649206502065120652206532065420655206562065720658206592066020661206622066320664206652066620667206682066920670206712067220673206742067520676206772067820679206802068120682206832068420685206862068720688206892069020691206922069320694206952069620697206982069920700207012070220703207042070520706207072070820709207102071120712207132071420715207162071720718207192072020721207222072320724207252072620727207282072920730207312073220733207342073520736207372073820739207402074120742207432074420745207462074720748207492075020751207522075320754207552075620757207582075920760207612076220763207642076520766207672076820769207702077120772207732077420775207762077720778207792078020781207822078320784207852078620787207882078920790207912079220793207942079520796207972079820799208002080120802208032080420805208062080720808208092081020811208122081320814208152081620817208182081920820208212082220823208242082520826208272082820829208302083120832208332083420835208362083720838208392084020841208422084320844208452084620847208482084920850208512085220853208542085520856208572085820859208602086120862208632086420865208662086720868208692087020871208722087320874208752087620877208782087920880208812088220883208842088520886208872088820889208902089120892208932089420895208962089720898208992090020901209022090320904209052090620907209082090920910209112091220913209142091520916209172091820919209202092120922209232092420925209262092720928209292093020931209322093320934209352093620937209382093920940209412094220943209442094520946209472094820949209502095120952209532095420955209562095720958209592096020961209622096320964209652096620967209682096920970209712097220973209742097520976209772097820979209802098120982209832098420985209862098720988209892099020991209922099320994209952099620997209982099921000210012100221003210042100521006210072100821009210102101121012210132101421015210162101721018210192102021021210222102321024210252102621027210282102921030210312103221033210342103521036210372103821039210402104121042210432104421045210462104721048210492105021051210522105321054210552105621057210582105921060210612106221063210642106521066210672106821069210702107121072210732107421075210762107721078210792108021081210822108321084210852108621087210882108921090210912109221093210942109521096210972109821099211002110121102211032110421105211062110721108211092111021111211122111321114211152111621117211182111921120211212112221123211242112521126211272112821129211302113121132211332113421135211362113721138211392114021141211422114321144211452114621147211482114921150211512115221153211542115521156211572115821159211602116121162211632116421165211662116721168211692117021171211722117321174211752117621177211782117921180211812118221183211842118521186211872118821189211902119121192211932119421195211962119721198211992120021201212022120321204212052120621207212082120921210212112121221213212142121521216212172121821219212202122121222212232122421225212262122721228212292123021231212322123321234212352123621237212382123921240212412124221243212442124521246212472124821249212502125121252212532125421255212562125721258212592126021261212622126321264212652126621267212682126921270212712127221273212742127521276212772127821279212802128121282212832128421285212862128721288212892129021291212922129321294212952129621297212982129921300213012130221303213042130521306213072130821309213102131121312213132131421315213162131721318213192132021321213222132321324213252132621327213282132921330213312133221333213342133521336213372133821339213402134121342213432134421345213462134721348213492135021351213522135321354213552135621357213582135921360213612136221363213642136521366213672136821369213702137121372213732137421375213762137721378213792138021381213822138321384213852138621387213882138921390213912139221393213942139521396213972139821399214002140121402214032140421405214062140721408214092141021411214122141321414214152141621417214182141921420214212142221423214242142521426214272142821429214302143121432214332143421435214362143721438214392144021441214422144321444214452144621447214482144921450214512145221453214542145521456214572145821459214602146121462214632146421465214662146721468214692147021471214722147321474214752147621477214782147921480214812148221483214842148521486214872148821489214902149121492214932149421495214962149721498214992150021501215022150321504215052150621507215082150921510215112151221513215142151521516215172151821519215202152121522215232152421525215262152721528215292153021531215322153321534215352153621537215382153921540215412154221543215442154521546215472154821549215502155121552215532155421555215562155721558215592156021561215622156321564215652156621567215682156921570215712157221573215742157521576215772157821579215802158121582215832158421585215862158721588215892159021591215922159321594215952159621597215982159921600216012160221603216042160521606216072160821609216102161121612216132161421615216162161721618216192162021621216222162321624216252162621627216282162921630216312163221633216342163521636216372163821639216402164121642216432164421645216462164721648216492165021651216522165321654216552165621657216582165921660216612166221663216642166521666216672166821669216702167121672216732167421675216762167721678216792168021681216822168321684216852168621687216882168921690216912169221693216942169521696216972169821699217002170121702217032170421705217062170721708217092171021711217122171321714217152171621717217182171921720217212172221723217242172521726217272172821729217302173121732217332173421735217362173721738217392174021741217422174321744217452174621747217482174921750217512175221753217542175521756217572175821759217602176121762217632176421765217662176721768217692177021771217722177321774217752177621777217782177921780217812178221783217842178521786217872178821789217902179121792217932179421795217962179721798217992180021801218022180321804218052180621807218082180921810218112181221813218142181521816218172181821819218202182121822218232182421825218262182721828218292183021831218322183321834218352183621837218382183921840218412184221843218442184521846218472184821849218502185121852218532185421855218562185721858218592186021861218622186321864218652186621867218682186921870218712187221873218742187521876218772187821879218802188121882218832188421885218862188721888218892189021891218922189321894218952189621897218982189921900219012190221903219042190521906219072190821909219102191121912219132191421915219162191721918219192192021921219222192321924219252192621927219282192921930219312193221933219342193521936219372193821939219402194121942219432194421945219462194721948219492195021951219522195321954219552195621957219582195921960219612196221963219642196521966219672196821969219702197121972219732197421975219762197721978219792198021981219822198321984219852198621987219882198921990219912199221993219942199521996219972199821999220002200122002220032200422005220062200722008220092201022011220122201322014220152201622017220182201922020220212202222023220242202522026220272202822029220302203122032220332203422035220362203722038220392204022041220422204322044220452204622047220482204922050220512205222053220542205522056220572205822059220602206122062220632206422065220662206722068220692207022071220722207322074220752207622077220782207922080220812208222083220842208522086220872208822089220902209122092220932209422095220962209722098220992210022101221022210322104221052210622107221082210922110221112211222113221142211522116221172211822119221202212122122221232212422125221262212722128221292213022131221322213322134221352213622137221382213922140221412214222143221442214522146221472214822149221502215122152221532215422155221562215722158221592216022161221622216322164221652216622167221682216922170221712217222173221742217522176221772217822179221802218122182221832218422185221862218722188221892219022191221922219322194221952219622197221982219922200222012220222203222042220522206222072220822209222102221122212222132221422215222162221722218222192222022221222222222322224222252222622227222282222922230222312223222233222342223522236222372223822239222402224122242222432224422245222462224722248222492225022251222522225322254222552225622257222582225922260222612226222263222642226522266222672226822269222702227122272222732227422275222762227722278222792228022281222822228322284222852228622287222882228922290222912229222293222942229522296222972229822299223002230122302223032230422305223062230722308223092231022311223122231322314223152231622317223182231922320223212232222323223242232522326223272232822329223302233122332223332233422335223362233722338223392234022341223422234322344223452234622347223482234922350223512235222353223542235522356223572235822359223602236122362223632236422365223662236722368223692237022371223722237322374223752237622377223782237922380223812238222383223842238522386223872238822389223902239122392223932239422395223962239722398223992240022401224022240322404224052240622407224082240922410224112241222413224142241522416224172241822419224202242122422224232242422425224262242722428224292243022431224322243322434224352243622437224382243922440224412244222443224442244522446224472244822449224502245122452224532245422455224562245722458224592246022461224622246322464224652246622467224682246922470224712247222473224742247522476224772247822479224802248122482224832248422485224862248722488224892249022491224922249322494224952249622497224982249922500225012250222503225042250522506225072250822509225102251122512225132251422515225162251722518225192252022521225222252322524225252252622527225282252922530225312253222533225342253522536225372253822539225402254122542225432254422545225462254722548225492255022551225522255322554225552255622557225582255922560225612256222563225642256522566225672256822569225702257122572225732257422575225762257722578225792258022581225822258322584225852258622587225882258922590225912259222593225942259522596225972259822599226002260122602226032260422605226062260722608226092261022611226122261322614226152261622617226182261922620226212262222623226242262522626226272262822629226302263122632226332263422635226362263722638226392264022641226422264322644226452264622647226482264922650226512265222653226542265522656226572265822659226602266122662226632266422665226662266722668226692267022671226722267322674226752267622677226782267922680226812268222683226842268522686226872268822689226902269122692226932269422695226962269722698226992270022701227022270322704227052270622707227082270922710227112271222713227142271522716227172271822719227202272122722227232272422725227262272722728227292273022731227322273322734227352273622737227382273922740227412274222743227442274522746227472274822749227502275122752227532275422755227562275722758227592276022761227622276322764227652276622767227682276922770227712277222773227742277522776227772277822779227802278122782227832278422785227862278722788227892279022791227922279322794227952279622797227982279922800228012280222803228042280522806228072280822809228102281122812228132281422815228162281722818228192282022821228222282322824228252282622827228282282922830228312283222833228342283522836228372283822839228402284122842228432284422845228462284722848228492285022851228522285322854228552285622857228582285922860228612286222863228642286522866228672286822869228702287122872228732287422875228762287722878228792288022881228822288322884228852288622887228882288922890228912289222893228942289522896228972289822899229002290122902229032290422905229062290722908229092291022911229122291322914229152291622917229182291922920229212292222923229242292522926229272292822929229302293122932229332293422935229362293722938229392294022941229422294322944229452294622947229482294922950229512295222953229542295522956229572295822959229602296122962229632296422965229662296722968229692297022971229722297322974229752297622977229782297922980229812298222983229842298522986229872298822989229902299122992229932299422995229962299722998229992300023001230022300323004230052300623007230082300923010230112301223013230142301523016230172301823019230202302123022230232302423025230262302723028230292303023031230322303323034230352303623037230382303923040230412304223043230442304523046230472304823049230502305123052230532305423055230562305723058230592306023061230622306323064230652306623067230682306923070230712307223073230742307523076230772307823079230802308123082230832308423085230862308723088230892309023091230922309323094230952309623097230982309923100231012310223103231042310523106231072310823109231102311123112231132311423115231162311723118231192312023121231222312323124231252312623127231282312923130231312313223133231342313523136231372313823139231402314123142231432314423145231462314723148231492315023151231522315323154231552315623157231582315923160231612316223163231642316523166231672316823169231702317123172231732317423175231762317723178231792318023181231822318323184231852318623187231882318923190231912319223193231942319523196231972319823199232002320123202232032320423205232062320723208232092321023211232122321323214232152321623217232182321923220232212322223223232242322523226232272322823229232302323123232232332323423235232362323723238232392324023241232422324323244232452324623247232482324923250232512325223253232542325523256232572325823259232602326123262232632326423265232662326723268232692327023271232722327323274232752327623277232782327923280232812328223283232842328523286232872328823289232902329123292232932329423295232962329723298232992330023301233022330323304233052330623307233082330923310233112331223313233142331523316233172331823319233202332123322233232332423325233262332723328233292333023331233322333323334233352333623337233382333923340233412334223343233442334523346233472334823349
  1. #LyX 2.3 created this file. For more info see http://www.lyx.org/
  2. \lyxformat 544
  3. \begin_document
  4. \begin_header
  5. \save_transient_properties true
  6. \origin unavailable
  7. \textclass extbook
  8. \begin_preamble
  9. % List all used files in log output
  10. \listfiles
  11. %% Add TOC, List of Figures, etc. to TOC
  12. \usepackage{tocbibind}
  13. % Add a DRAFT watermark
  14. \usepackage{draftwatermark}
  15. \usepackage{accsupp}
  16. \SetWatermarkLightness{0.97}
  17. \SetWatermarkScale{1}
  18. % Make watermark not copyable (in Adobe Reader)
  19. \SetWatermarkText{\BeginAccSupp{method=escape,ActualText={}}DRAFT\EndAccSupp{}}
  20. % Set up required header format
  21. \usepackage{fancyhdr}
  22. \pagestyle{fancy}
  23. \renewcommand{\headrulewidth}{0pt}
  24. \rhead{}
  25. \lhead{}
  26. \chead{}
  27. \rfoot{}
  28. \lfoot{}
  29. % Make page number not copyable (in Adobe Reader)
  30. \cfoot{\BeginAccSupp{method=escape,ActualText={}}\thepage\EndAccSupp{}} % Page number bottom center
  31. % Allow FloatBarrier command
  32. \usepackage{placeins}
  33. % Allow landscape pages
  34. \usepackage{pdflscape}
  35. % Allow doing things after the end of the current page
  36. % (to avoid landscape figures breaking up text)
  37. \usepackage{afterpage}
  38. % Consider: force floats after placement in text
  39. % https://tex.stackexchange.com/questions/15706/force-floats-to-be-typeset-after-their-occurrence-in-the-source-text
  40. % This one breaks subfigs so it's disabled
  41. % https://tex.stackexchange.com/questions/65680/automatically-bold-first-sentence-of-a-floats-caption
  42. \usepackage[automake=immediate,nonumberlist,nohypertypes={abbreviation}]{glossaries-extra}
  43. \setabbreviationstyle{long-short}
  44. \loadglsentries{abbrevs.tex}
  45. \makeglossaries
  46. % arara: xelatex
  47. % arara: biber
  48. % arara: makeglossaries
  49. % arara: xelatex
  50. \end_preamble
  51. \use_default_options true
  52. \begin_modules
  53. todonotes
  54. logicalmkup
  55. \end_modules
  56. \maintain_unincluded_children false
  57. \begin_local_layout
  58. Format 66
  59. InsetLayout "Flex:Glossary Term"
  60. LyxType custom
  61. LabelString gls
  62. LatexType command
  63. LatexName gls*
  64. InToc true
  65. CustomPars false
  66. End
  67. InsetLayout "Flex:Glossary Term (Capital)"
  68. LyxType custom
  69. LabelString Gls
  70. LatexType command
  71. LatexName Gls*
  72. InToc true
  73. CustomPars false
  74. End
  75. InsetLayout "Flex:Glossary Term (pl)"
  76. LyxType custom
  77. LabelString glspl
  78. LatexType command
  79. LatexName glspl*
  80. InToc true
  81. CustomPars false
  82. End
  83. InsetLayout "Flex:Glossary Term (Capital, pl)"
  84. LyxType custom
  85. LabelString Glspl
  86. LatexType command
  87. LatexName Glspl*
  88. InToc true
  89. CustomPars false
  90. End
  91. InsetLayout "Flex:Glossary Term (glstext)"
  92. LyxType custom
  93. LabelString glstext
  94. LatexType command
  95. LatexName glstext*
  96. InToc true
  97. CustomPars false
  98. End
  99. InsetLayout "Flex:Glossary Term (Glstext)"
  100. LyxType custom
  101. LabelString Glstext
  102. LatexType command
  103. LatexName Glstext*
  104. InToc true
  105. CustomPars false
  106. End
  107. InsetLayout "Flex:Glossary Term (glsfirst)"
  108. LyxType custom
  109. LabelString glsfirst
  110. LatexType command
  111. LatexName glsfirst*
  112. InToc true
  113. CustomPars false
  114. End
  115. InsetLayout "Flex:Glossary Term (Glsfirst)"
  116. LyxType custom
  117. LabelString Glsfirst
  118. LatexType command
  119. LatexName Glsfirst*
  120. InToc true
  121. CustomPars false
  122. End
  123. InsetLayout "Flex:Glossary Term (glsdesc)"
  124. LyxType custom
  125. LabelString glsdesc
  126. LatexType command
  127. LatexName glsdesc*
  128. InToc true
  129. CustomPars false
  130. End
  131. InsetLayout "Flex:Glossary Term (Glsdesc)"
  132. LyxType custom
  133. LabelString Glsdesc
  134. LatexType command
  135. LatexName Glsdesc*
  136. InToc true
  137. CustomPars false
  138. End
  139. \end_local_layout
  140. \language english
  141. \language_package default
  142. \inputencoding utf8
  143. \fontencoding default
  144. \font_roman "default" "default"
  145. \font_sans "default" "default"
  146. \font_typewriter "default" "default"
  147. \font_math "auto" "auto"
  148. \font_default_family default
  149. \use_non_tex_fonts false
  150. \font_sc false
  151. \font_osf false
  152. \font_sf_scale 100 100
  153. \font_tt_scale 100 100
  154. \use_microtype false
  155. \use_dash_ligatures true
  156. \graphics default
  157. \default_output_format pdf4
  158. \output_sync 0
  159. \bibtex_command biber
  160. \index_command default
  161. \paperfontsize 12
  162. \spacing double
  163. \use_hyperref true
  164. \pdf_author "Ryan C. Thompson"
  165. \pdf_bookmarks true
  166. \pdf_bookmarksnumbered true
  167. \pdf_bookmarksopen true
  168. \pdf_bookmarksopenlevel 1
  169. \pdf_breaklinks true
  170. \pdf_pdfborder true
  171. \pdf_colorlinks false
  172. \pdf_backref false
  173. \pdf_pdfusetitle true
  174. \papersize letterpaper
  175. \use_geometry true
  176. \use_package amsmath 1
  177. \use_package amssymb 1
  178. \use_package cancel 1
  179. \use_package esint 1
  180. \use_package mathdots 1
  181. \use_package mathtools 1
  182. \use_package mhchem 1
  183. \use_package stackrel 1
  184. \use_package stmaryrd 1
  185. \use_package undertilde 1
  186. \cite_engine biblatex
  187. \cite_engine_type numerical
  188. \biblio_style plain
  189. \biblio_options sorting=none
  190. \biblatex_bibstyle numeric
  191. \biblatex_citestyle numeric
  192. \use_bibtopic false
  193. \use_indices false
  194. \paperorientation portrait
  195. \suppress_date false
  196. \justification true
  197. \use_refstyle 1
  198. \use_minted 0
  199. \index Index
  200. \shortcut idx
  201. \color #008000
  202. \end_index
  203. \leftmargin 1.5in
  204. \topmargin 1in
  205. \rightmargin 1in
  206. \bottommargin 1in
  207. \secnumdepth 3
  208. \tocdepth 3
  209. \paragraph_separation indent
  210. \paragraph_indentation default
  211. \is_math_indent 0
  212. \math_numbering_side default
  213. \quotes_style english
  214. \dynamic_quotes 0
  215. \papercolumns 1
  216. \papersides 1
  217. \paperpagestyle default
  218. \tracking_changes false
  219. \output_changes false
  220. \html_math_output 0
  221. \html_css_as_file 0
  222. \html_be_strict false
  223. \end_header
  224. \begin_body
  225. \begin_layout Standard
  226. \begin_inset ERT
  227. status collapsed
  228. \begin_layout Plain Layout
  229. \backslash
  230. pdfbookmark{Title page}{title}
  231. \end_layout
  232. \end_inset
  233. \end_layout
  234. \begin_layout Title
  235. Bioinformatic analysis of complex, high-throughput genomic and epigenomic
  236. data in the context of immunology and transplant rejection
  237. \end_layout
  238. \begin_layout Author
  239. A thesis presented
  240. \begin_inset Newline newline
  241. \end_inset
  242. by
  243. \begin_inset Newline newline
  244. \end_inset
  245. Ryan C.
  246. Thompson
  247. \begin_inset Newline newline
  248. \end_inset
  249. to
  250. \begin_inset Newline newline
  251. \end_inset
  252. The Scripps Research Institute Graduate Program
  253. \begin_inset Newline newline
  254. \end_inset
  255. in partial fulfillment of the requirements for the degree of
  256. \begin_inset Newline newline
  257. \end_inset
  258. Doctor of Philosophy in the subject of Biology
  259. \begin_inset Newline newline
  260. \end_inset
  261. for
  262. \begin_inset Newline newline
  263. \end_inset
  264. The Scripps Research Institute
  265. \begin_inset Newline newline
  266. \end_inset
  267. La Jolla, California
  268. \end_layout
  269. \begin_layout Date
  270. October 2019
  271. \end_layout
  272. \begin_layout Standard
  273. \begin_inset Note Note
  274. status open
  275. \begin_layout Plain Layout
  276. To remove TODOs and watermark: Add
  277. \begin_inset Quotes eld
  278. \end_inset
  279. final
  280. \begin_inset Quotes erd
  281. \end_inset
  282. to the document class custom options.
  283. \end_layout
  284. \end_inset
  285. \end_layout
  286. \begin_layout Standard
  287. \begin_inset ERT
  288. status open
  289. \begin_layout Plain Layout
  290. \backslash
  291. frontmatter
  292. \end_layout
  293. \end_inset
  294. \begin_inset Note Note
  295. status open
  296. \begin_layout Plain Layout
  297. Use roman numeral page numbers
  298. \end_layout
  299. \end_inset
  300. \end_layout
  301. \begin_layout Standard
  302. \begin_inset Newpage newpage
  303. \end_inset
  304. \end_layout
  305. \begin_layout Standard
  306. \align center
  307. \begin_inset ERT
  308. status collapsed
  309. \begin_layout Plain Layout
  310. \backslash
  311. phantomsection
  312. \end_layout
  313. \begin_layout Plain Layout
  314. \backslash
  315. addcontentsline{toc}{chapter}{Copyright notice}
  316. \end_layout
  317. \end_inset
  318. \end_layout
  319. \begin_layout Standard
  320. \align center
  321. \begin_inset ERT
  322. status collapsed
  323. \begin_layout Plain Layout
  324. \backslash
  325. vspace*{
  326. \backslash
  327. stretch{1}}
  328. \end_layout
  329. \end_inset
  330. \end_layout
  331. \begin_layout Standard
  332. \align center
  333. © 2019 by Ryan C.
  334. Thompson
  335. \end_layout
  336. \begin_layout Standard
  337. \align center
  338. All rights reserved.
  339. \end_layout
  340. \begin_layout Standard
  341. \align center
  342. \begin_inset ERT
  343. status collapsed
  344. \begin_layout Plain Layout
  345. \backslash
  346. vspace*{
  347. \backslash
  348. stretch{2}}
  349. \end_layout
  350. \end_inset
  351. \end_layout
  352. \begin_layout Standard
  353. \begin_inset Newpage newpage
  354. \end_inset
  355. \end_layout
  356. \begin_layout Standard
  357. \align center
  358. \begin_inset ERT
  359. status collapsed
  360. \begin_layout Plain Layout
  361. \backslash
  362. phantomsection
  363. \end_layout
  364. \begin_layout Plain Layout
  365. \backslash
  366. addcontentsline{toc}{chapter}{Thesis acceptance form}
  367. \end_layout
  368. \end_inset
  369. \end_layout
  370. \begin_layout Standard
  371. \align center
  372. [Thesis acceptance form]
  373. \end_layout
  374. \begin_layout Standard
  375. \begin_inset Newpage newpage
  376. \end_inset
  377. \end_layout
  378. \begin_layout Standard
  379. \align center
  380. \begin_inset ERT
  381. status collapsed
  382. \begin_layout Plain Layout
  383. \backslash
  384. phantomsection
  385. \end_layout
  386. \begin_layout Plain Layout
  387. \backslash
  388. addcontentsline{toc}{chapter}{Dedication}
  389. \end_layout
  390. \end_inset
  391. \end_layout
  392. \begin_layout Standard
  393. \align center
  394. \begin_inset ERT
  395. status collapsed
  396. \begin_layout Plain Layout
  397. \backslash
  398. vspace*{
  399. \backslash
  400. stretch{1}}
  401. \end_layout
  402. \end_inset
  403. \end_layout
  404. \begin_layout Standard
  405. \align center
  406. [Dedication]
  407. \end_layout
  408. \begin_layout Standard
  409. \align center
  410. \begin_inset ERT
  411. status collapsed
  412. \begin_layout Plain Layout
  413. \backslash
  414. vspace*{
  415. \backslash
  416. stretch{2}}
  417. \end_layout
  418. \end_inset
  419. \end_layout
  420. \begin_layout Standard
  421. \begin_inset Newpage newpage
  422. \end_inset
  423. \end_layout
  424. \begin_layout Standard
  425. \align center
  426. \begin_inset ERT
  427. status collapsed
  428. \begin_layout Plain Layout
  429. \backslash
  430. phantomsection
  431. \end_layout
  432. \begin_layout Plain Layout
  433. \backslash
  434. addcontentsline{toc}{chapter}{Acknowledgements}
  435. \end_layout
  436. \end_inset
  437. \end_layout
  438. \begin_layout Section*
  439. \begin_inset ERT
  440. status collapsed
  441. \begin_layout Plain Layout
  442. \backslash
  443. hspace*{
  444. \backslash
  445. stretch{1}}
  446. \end_layout
  447. \end_inset
  448. Acknowledgements
  449. \begin_inset ERT
  450. status collapsed
  451. \begin_layout Plain Layout
  452. \backslash
  453. hspace*{
  454. \backslash
  455. stretch{1}}
  456. \end_layout
  457. \end_inset
  458. \end_layout
  459. \begin_layout Standard
  460. [Acknowledgements]
  461. \end_layout
  462. \begin_layout Standard
  463. \begin_inset Newpage newpage
  464. \end_inset
  465. \end_layout
  466. \begin_layout Standard
  467. \begin_inset CommandInset toc
  468. LatexCommand tableofcontents
  469. \end_inset
  470. \end_layout
  471. \begin_layout Standard
  472. \begin_inset FloatList table
  473. \end_inset
  474. \end_layout
  475. \begin_layout Standard
  476. \begin_inset FloatList figure
  477. \end_inset
  478. \end_layout
  479. \begin_layout Standard
  480. \begin_inset Note Note
  481. status open
  482. \begin_layout Plain Layout
  483. To create a new abbreviation:
  484. \end_layout
  485. \begin_layout Enumerate
  486. Add an entry to abbrevs.tex
  487. \end_layout
  488. \begin_layout Enumerate
  489. Wrap every occurrence of the term in Insert -> Custom Insets -> Glossary
  490. Term (use appropriate variants for caiptal, plural, etc.), using Edit ->
  491. Find & Replace (Advanced).
  492. Skip section headers and float captions.
  493. \end_layout
  494. \begin_layout Plain Layout
  495. \begin_inset CommandInset href
  496. LatexCommand href
  497. target "https://ctan.org/pkg/glossaries?lang=en"
  498. literal "false"
  499. \end_inset
  500. \begin_inset CommandInset href
  501. LatexCommand href
  502. target "https://ctan.org/pkg/glossaries-extra"
  503. literal "false"
  504. \end_inset
  505. \end_layout
  506. \end_inset
  507. \end_layout
  508. \begin_layout Standard
  509. \begin_inset ERT
  510. status collapsed
  511. \begin_layout Plain Layout
  512. \backslash
  513. renewcommand*{
  514. \backslash
  515. glossaryname}{List of Abbreviations}%
  516. \end_layout
  517. \begin_layout Plain Layout
  518. \backslash
  519. printglossaries
  520. \end_layout
  521. \end_inset
  522. \end_layout
  523. \begin_layout List of TODOs
  524. \end_layout
  525. \begin_layout Chapter*
  526. Abstract
  527. \end_layout
  528. \begin_layout Standard
  529. \begin_inset Note Note
  530. status open
  531. \begin_layout Plain Layout
  532. It is included as an integral part of the thesis and should immediately
  533. precede the introduction.
  534. \end_layout
  535. \begin_layout Plain Layout
  536. Preparing your Abstract.
  537. Your abstract (a succinct description of your work) is limited to 350 words.
  538. UMI will shorten it if they must; please do not exceed the limit.
  539. \end_layout
  540. \begin_layout Itemize
  541. Include pertinent place names, names of persons (in full), and other proper
  542. nouns.
  543. These are useful in automated retrieval.
  544. \end_layout
  545. \begin_layout Itemize
  546. Display symbols, as well as foreign words and phrases, clearly and accurately.
  547. Include transliterations for characters other than Roman and Greek letters
  548. and Arabic numerals.
  549. Include accents and diacritical marks.
  550. \end_layout
  551. \begin_layout Itemize
  552. Do not include graphs, charts, tables, or illustrations in your abstract.
  553. \end_layout
  554. \end_inset
  555. \end_layout
  556. \begin_layout Standard
  557. \begin_inset Flex TODO Note (inline)
  558. status open
  559. \begin_layout Plain Layout
  560. Obviously the abstract gets written last.
  561. \end_layout
  562. \end_inset
  563. \end_layout
  564. \begin_layout Standard
  565. \begin_inset Note Note
  566. status collapsed
  567. \begin_layout Chapter*
  568. Notes to draft readers
  569. \end_layout
  570. \begin_layout Plain Layout
  571. Thank you so much for agreeing to read my thesis and give me feedback on
  572. it.
  573. What you are currently reading is a rough draft, in need of many revisions.
  574. You can always find the latest version at
  575. \begin_inset CommandInset href
  576. LatexCommand href
  577. target "https://mneme.dedyn.io/~ryan/Thesis/thesis.pdf"
  578. literal "false"
  579. \end_inset
  580. .
  581. the PDF at this link is updated periodically with my latest revisions,
  582. but you can just download the current version and give me feedback on that.
  583. Don't worry about keeping up with the updates.
  584. \end_layout
  585. \begin_layout Plain Layout
  586. As for what feedback I'm looking for, first of all, don't waste your time
  587. marking spelling mistakes and such.
  588. I haven't run a spell checker on it yet, so let me worry about that.
  589. Also, I'm aware that many abbreviations are not properly introduced the
  590. first time they are used, so don't worry about that either.
  591. However, if you see any glaring formatting issues, such as a figure being
  592. too large and getting cut off at the edge of the page, please note them.
  593. In addition, if any of the text in the figures is too small, please note
  594. that as well.
  595. \end_layout
  596. \begin_layout Plain Layout
  597. Beyond that, what I'm mainly interested in is feedback on the content.
  598. For example: does the introduction flow logically, and does it provide
  599. enough background to understand the other chapters? Does each chapter make
  600. it clear what work and analyses I have done? Do the figures clearly communicate
  601. the results I'm trying to show? Do you feel that the claims in the results
  602. and discussion sections are well-supported? There's no need to suggest
  603. improvements; just note areas that you feel need improvement.
  604. Additionally, if you notice any un-cited claims in any chapter, please
  605. flag them for my attention.
  606. Similarly, if you discover any factual errors, please note them as well.
  607. \end_layout
  608. \begin_layout Plain Layout
  609. You can provide your feedback in whatever way is most convenient to you.
  610. You could mark up this PDF with highlights and notes, then send it back
  611. to me.
  612. Or you could collect your comments in a separate text file and send that
  613. to me, or whatever else you like.
  614. However, if you send me your feedback in a separate document, please note
  615. a section/figure/table number for each comment, and
  616. \emph on
  617. also
  618. \emph default
  619. send me the exact PDF that you read so I can reference it while reading
  620. your comments, since as mentioned above, the current version I'm working
  621. on will have changed by that point (which might include shuffling sections
  622. and figures around, changing their numbers).
  623. One last thing: you'll see a bunch of text in orange boxes throughout the
  624. PDF.
  625. These are notes to myself about things that need to be fixed later, so
  626. if you see a problem noted in an orange box, that means I'm already aware
  627. of it, and there's no need to comment on it.
  628. \end_layout
  629. \begin_layout Plain Layout
  630. My thesis is due Thursday, October 10th, so in order to be useful to me,
  631. I'll need your feedback at least several days before that, ideally by Monday,
  632. October 7th.
  633. If you have limited time and are unable to get through the whole thesis,
  634. please focus your efforts on Chapters 1 and 2, since those are the roughest
  635. and most in need of revision.
  636. Chapter 3 is fairly short and straightforward, and Chapter 4 is an adaptation
  637. of a paper that's already been through a few rounds of revision, so they
  638. should be a lot tighter.
  639. If you can't spare any time between now and then, or if something unexpected
  640. comes up, I understand.
  641. Just let me know.
  642. \end_layout
  643. \begin_layout Plain Layout
  644. Thanks again for your help, and happy reading!
  645. \end_layout
  646. \end_inset
  647. \end_layout
  648. \begin_layout Standard
  649. \begin_inset ERT
  650. status open
  651. \begin_layout Plain Layout
  652. \backslash
  653. mainmatter
  654. \end_layout
  655. \end_inset
  656. \begin_inset Note Note
  657. status open
  658. \begin_layout Plain Layout
  659. Switch from roman numerals to arabic for page numbers.
  660. \end_layout
  661. \end_inset
  662. \end_layout
  663. \begin_layout Chapter
  664. Introduction
  665. \end_layout
  666. \begin_layout Standard
  667. \begin_inset ERT
  668. status collapsed
  669. \begin_layout Plain Layout
  670. \backslash
  671. glsresetall
  672. \end_layout
  673. \end_inset
  674. \begin_inset Note Note
  675. status collapsed
  676. \begin_layout Plain Layout
  677. Reintroduce all abbreviations
  678. \end_layout
  679. \end_inset
  680. \end_layout
  681. \begin_layout Section
  682. \begin_inset CommandInset label
  683. LatexCommand label
  684. name "sec:Biological-motivation"
  685. \end_inset
  686. Biological motivation
  687. \end_layout
  688. \begin_layout Standard
  689. \begin_inset Flex TODO Note (inline)
  690. status open
  691. \begin_layout Plain Layout
  692. Find some figures to include even if permission is not obtained.
  693. Try to obtain permission, and if it cannot be obtained, remove/replace
  694. them later.
  695. \end_layout
  696. \end_inset
  697. \end_layout
  698. \begin_layout Standard
  699. \begin_inset Flex TODO Note (inline)
  700. status open
  701. \begin_layout Plain Layout
  702. Rethink the subsection organization after the intro is written.
  703. \end_layout
  704. \end_inset
  705. \end_layout
  706. \begin_layout Subsection
  707. Rejection is the major long-term threat to organ and tissue allografts
  708. \end_layout
  709. \begin_layout Standard
  710. Organ and tissue transplants are a life-saving treatment for people who
  711. have lost the function of an important organ.
  712. In some cases, it is possible to transplant a patient's own tissue from
  713. one area of their body to another, referred to as an autograft.
  714. This is common for tissues that are distributed throughout many areas of
  715. the body, such as skin and bone.
  716. However, in cases of organ failure, there is no functional self tissue
  717. remaining, and a transplant from another person – a donor – is required.
  718. This is referred to as an allograft
  719. \begin_inset CommandInset citation
  720. LatexCommand cite
  721. key "Valenzuela2017"
  722. literal "false"
  723. \end_inset
  724. .
  725. \end_layout
  726. \begin_layout Standard
  727. \begin_inset Flex TODO Note (inline)
  728. status open
  729. \begin_layout Plain Layout
  730. How much mechanistic detail is needed here? My work doesn't really go into
  731. specific rejection mechanisms, so I think it's best to keep it basic.
  732. \end_layout
  733. \end_inset
  734. \end_layout
  735. \begin_layout Standard
  736. Because an allograft comes from a donor of the same species who is genetically
  737. distinct from the recipient (with rare exceptions), genetic variants in
  738. protein-coding regions affect the polypeptide sequences encoded by the
  739. affected genes, resulting in protein products in the allograft that differ
  740. from the equivalent proteins produced by the graft recipient's own tissue.
  741. As a result, without intervention, the recipient's immune system will eventuall
  742. y identify the graft as foreign tissue and begin attacking it.
  743. This is called an alloimmune response, and if left unchecked, it eventually
  744. results in failure and death of the graft, a process referred to as transplant
  745. rejection
  746. \begin_inset CommandInset citation
  747. LatexCommand cite
  748. key "Murphy2012"
  749. literal "false"
  750. \end_inset
  751. .
  752. Rejection is the primary obstacle to long-term health and survival of an
  753. allograft
  754. \begin_inset CommandInset citation
  755. LatexCommand cite
  756. key "Valenzuela2017"
  757. literal "false"
  758. \end_inset
  759. .
  760. Like any adaptive immune response, an alloimmune response generally occurs
  761. via two broad mechanisms: cellular immunity, in which CD8
  762. \begin_inset Formula $^{+}$
  763. \end_inset
  764. T-cells recognizing graft-specific antigens induce apoptosis in the graft
  765. cells; and humoral immunity, in which B-cells produce antibodies that bind
  766. to graft proteins and direct an immune response against the graft
  767. \begin_inset CommandInset citation
  768. LatexCommand cite
  769. key "Murphy2012"
  770. literal "false"
  771. \end_inset
  772. .
  773. In either case, alloimmunity and rejection show most of the typical hallmarks
  774. of an adaptive immune response, in particular mediation by CD4
  775. \begin_inset Formula $^{+}$
  776. \end_inset
  777. T-cells and formation of immune memory.
  778. \end_layout
  779. \begin_layout Subsection
  780. Diagnosis and treatment of allograft rejection is a major challenge
  781. \end_layout
  782. \begin_layout Standard
  783. To prevent rejection, allograft recipients are treated with immune suppressive
  784. drugs
  785. \begin_inset CommandInset citation
  786. LatexCommand cite
  787. key "Kowalski2003,Murphy2012"
  788. literal "false"
  789. \end_inset
  790. .
  791. The goal is to achieve sufficient suppression of the immune system to prevent
  792. rejection of the graft without compromising the ability of the immune system
  793. to raise a normal response against infection.
  794. As such, a delicate balance must be struck: insufficient immune suppression
  795. may lead to rejection and ultimately loss of the graft; excessive suppression
  796. leaves the patient vulnerable to life-threatening opportunistic infections
  797. \begin_inset CommandInset citation
  798. LatexCommand cite
  799. key "Murphy2012"
  800. literal "false"
  801. \end_inset
  802. .
  803. Because every patient's matabolism is different, achieving this delicate
  804. balance requires drug dosage to be tailored for each patient.
  805. Furthermore, dosage must be tuned over time, as the immune system's activity
  806. varies over time and in response to external stimuli with no fixed pattern.
  807. In order to properly adjust the dosage of immune suppression drugs, it
  808. is necessary to monitor the health of the transplant and increase the dosage
  809. if evidence of rejection or alloimmune activity is observed.
  810. \end_layout
  811. \begin_layout Standard
  812. However, diagnosis of rejection is a significant challenge.
  813. Early diagnosis is essential in order to step up immune suppression before
  814. the immune system damages the graft beyond recovery
  815. \begin_inset CommandInset citation
  816. LatexCommand cite
  817. key "Israeli2007"
  818. literal "false"
  819. \end_inset
  820. .
  821. The current gold standard test for graft rejection is a tissue biopsy,
  822. examined for visible signs of rejection by a trained histologist
  823. \begin_inset CommandInset citation
  824. LatexCommand cite
  825. key "Kurian2014"
  826. literal "false"
  827. \end_inset
  828. .
  829. When a patient shows symptoms of possible rejection, a
  830. \begin_inset Quotes eld
  831. \end_inset
  832. for cause
  833. \begin_inset Quotes erd
  834. \end_inset
  835. biopsy is performed to confirm the diagnosis, and immune suppression is
  836. adjusted as necessary.
  837. However, in many cases, the early stages of rejection are asymptomatic,
  838. known as
  839. \begin_inset Quotes eld
  840. \end_inset
  841. sub-clinical
  842. \begin_inset Quotes erd
  843. \end_inset
  844. rejection.
  845. In light of this, is is now common to perform
  846. \begin_inset Quotes eld
  847. \end_inset
  848. protocol biopsies
  849. \begin_inset Quotes erd
  850. \end_inset
  851. at specific times after transplantation of a graft, even if no symptoms
  852. of rejection are apparent, in addition to
  853. \begin_inset Quotes eld
  854. \end_inset
  855. for cause
  856. \begin_inset Quotes erd
  857. \end_inset
  858. biopsies
  859. \begin_inset CommandInset citation
  860. LatexCommand cite
  861. key "Salomon2002,Wilkinson2006,Patel2018,Zachariah2018"
  862. literal "false"
  863. \end_inset
  864. .
  865. \end_layout
  866. \begin_layout Standard
  867. However, biopsies have a number of downsides that limit their effectiveness
  868. as a diagnostic tool.
  869. First, the need for manual inspection by a histologist means that diagnosis
  870. is subject to the biases of the particular histologist examining the biopsy
  871. \begin_inset CommandInset citation
  872. LatexCommand cite
  873. key "Kurian2014"
  874. literal "false"
  875. \end_inset
  876. .
  877. In marginal cases, two different histologists may give two different diagnoses
  878. to the same biopsy.
  879. Second, a biopsy can only evaluate if rejection is occurring in the section
  880. of the graft from which the tissue was extracted.
  881. If rejection is localized to one section of the graft and the tissue is
  882. extracted from a different section, a false negative diagnosis may result.
  883. Most importantly, extraction of tissue from a graft is invasive and is
  884. treated as an injury by the body, which results in inflammation that in
  885. turn promotes increased immune system activity.
  886. Hence, the invasiveness of biopsies severely limits the frequency with
  887. which they can safely be performed
  888. \begin_inset CommandInset citation
  889. LatexCommand cite
  890. key "Patel2018"
  891. literal "false"
  892. \end_inset
  893. .
  894. Typically, protocol biopsies are not scheduled more than about once per
  895. month
  896. \begin_inset CommandInset citation
  897. LatexCommand cite
  898. key "Wilkinson2006"
  899. literal "false"
  900. \end_inset
  901. .
  902. A less invasive diagnostic test for rejection would bring manifold benefits.
  903. Such a test would enable more frequent testing and therefore earlier detection
  904. of rejection events.
  905. In addition, having a larger pool of historical data for a given patient
  906. would make it easier to evaluate when a given test is outside the normal
  907. parameters for that specific patient, rather than relying on normal ranges
  908. for the population as a whole.
  909. Lastly, the accumulated data from more frequent tests would be a boon to
  910. the transplant research community.
  911. Beyond simply providing more data overall, the better time granularity
  912. of the tests will enable studying the progression of a rejection event
  913. on the scale of days to weeks, rather than months.
  914. \end_layout
  915. \begin_layout Subsection
  916. Memory cells are resistant to immune suppression
  917. \end_layout
  918. \begin_layout Standard
  919. One of the defining features of the adaptive immune system is immune memory:
  920. the ability of the immune system to recognize a previously encountered
  921. foreign antigen and respond more quickly and more strongly to that antigen
  922. in subsequent encounters
  923. \begin_inset CommandInset citation
  924. LatexCommand cite
  925. key "Murphy2012"
  926. literal "false"
  927. \end_inset
  928. .
  929. When the immune system first encounters a new antigen, the T-cells that
  930. respond are known as naïve cells – T-cells that have never detected their
  931. target antigens before.
  932. Once activated by their specific antigen presented by an antigen-presenting
  933. cell in the proper co-stimulatory context, naïve cells differentiate into
  934. effector cells that carry out their respective functions in targeting and
  935. destroying the source of the foreign antigen.
  936. The
  937. \begin_inset Flex Glossary Term
  938. status open
  939. \begin_layout Plain Layout
  940. TCR
  941. \end_layout
  942. \end_inset
  943. is cell-surface protein complex produced by T-cells that is responsible
  944. for recognizing the T-cell's specific antigen, presented on a
  945. \begin_inset Flex Glossary Term
  946. status open
  947. \begin_layout Plain Layout
  948. MHC
  949. \end_layout
  950. \end_inset
  951. , the cell-surface protein complex used by an
  952. \begin_inset Flex Glossary Term
  953. status open
  954. \begin_layout Plain Layout
  955. APC
  956. \end_layout
  957. \end_inset
  958. to present antigens to the T-cell.
  959. However, a naïve T-cell that recognizes its antigen also requires a co-stimulat
  960. ory signal, delivered through other interactions between
  961. \begin_inset Flex Glossary Term
  962. status open
  963. \begin_layout Plain Layout
  964. APC
  965. \end_layout
  966. \end_inset
  967. surface proteins and T-cell surface proteins such as CD28.
  968. Without proper co-stimulation, a T-cell that recognizes its antigen either
  969. dies or enters an unresponsive state known as anergy, in which the T-cell
  970. becomes much more resistant to subsequent activation even with proper co-stimul
  971. ation.
  972. The dependency of activation on co-stimulation is an important feature
  973. of naïve lymphocytes that limits
  974. \begin_inset Quotes eld
  975. \end_inset
  976. false positive
  977. \begin_inset Quotes erd
  978. \end_inset
  979. immune responses against self antigens, because
  980. \begin_inset Flex Glossary Term (pl)
  981. status open
  982. \begin_layout Plain Layout
  983. APC
  984. \end_layout
  985. \end_inset
  986. usually only express the proper co-stimulation after the innate immune
  987. system detects signs of an active infection, such as the presence of common
  988. bacterial cell components or inflamed tissue.
  989. \end_layout
  990. \begin_layout Standard
  991. After the foreign antigen is cleared, most effector cells die since they
  992. are no longer needed, but some differentiate into memory cells and remain
  993. alive indefinitely.
  994. Like naïve cells, memory cells respond to detection of their specific antigen
  995. by differentiating into effector cells, ready to fight an infection
  996. \begin_inset CommandInset citation
  997. LatexCommand cite
  998. key "Murphy2012"
  999. literal "false"
  1000. \end_inset
  1001. .
  1002. However, the memory response to antigen is qualitatively different: memory
  1003. cells are more sensitive to detection of their antigen, and a lower concentrati
  1004. on of antigen is suffiicient to activate them
  1005. \begin_inset CommandInset citation
  1006. LatexCommand cite
  1007. key "Rogers2000,London2000,Berard2002"
  1008. literal "false"
  1009. \end_inset
  1010. .
  1011. In addition, memory cells are much less dependent on co-stimulation for
  1012. activation: they can activate without certain co-stimulatory signals that
  1013. are required by naïve cells, and the signals they do require are only required
  1014. at lower levels in order to cause activation
  1015. \begin_inset CommandInset citation
  1016. LatexCommand cite
  1017. key "London2000"
  1018. literal "false"
  1019. \end_inset
  1020. .
  1021. Furthermore, mechanisms that induce tolerance (non-response to antigen)
  1022. in naïve cells are much less effective on memory cells
  1023. \begin_inset CommandInset citation
  1024. LatexCommand cite
  1025. key "London2000"
  1026. literal "false"
  1027. \end_inset
  1028. .
  1029. Lastly, once activated, memory cells proliferate and differentiate into
  1030. effector cells more quickly than naïve cells do
  1031. \begin_inset CommandInset citation
  1032. LatexCommand cite
  1033. key "Berard2002"
  1034. literal "false"
  1035. \end_inset
  1036. .
  1037. In combination, these changes in lymphocyte behavior upon differentiation
  1038. into memory cells account for the much quicker and stronger response of
  1039. the immune system to subsequent exposure to a previously-encountered antigen.
  1040. \end_layout
  1041. \begin_layout Standard
  1042. In the context of a pathogenic infection, immune memory is a major advantage,
  1043. allowing an organism to rapidly fight off a previously encountered pathogen
  1044. much more quickly and effectively than the first time it was encountered
  1045. \begin_inset CommandInset citation
  1046. LatexCommand cite
  1047. key "Murphy2012"
  1048. literal "false"
  1049. \end_inset
  1050. .
  1051. However, if effector cells that recognize an antigen from an allograft
  1052. are allowed to differentiate into memory cells, preventing rejection of
  1053. the graft becomes much more difficult.
  1054. Many immune suppression drugs work by interfering with the co-stimulation
  1055. that naïve cells require in order to mount an immune response.
  1056. Since memory cells do not require the same degree of co-stimulation, these
  1057. drugs are not effective at suppressing an immune response that is mediated
  1058. by memory cells.
  1059. Secondly, because memory cells are able to mount a stronger and faster
  1060. response to an antigen, all else being equal stronger immune suppression
  1061. is required to prevent an immune response mediated by memory cells.
  1062. \end_layout
  1063. \begin_layout Standard
  1064. However, immune suppression affects the entire immune system, not just cells
  1065. recognizing a specific antigen, so increasing the dosage of immune suppression
  1066. drugs also increases the risk of complications from a compromised immune
  1067. system, such as opportunistic infections
  1068. \begin_inset CommandInset citation
  1069. LatexCommand cite
  1070. key "Murphy2012"
  1071. literal "false"
  1072. \end_inset
  1073. .
  1074. While the differences in cell surface markers between naïve and memory
  1075. cells have been fairly well characterized, the internal regulatory mechanisms
  1076. that allow memory cells to respond more quickly and without co-stimulation
  1077. are still poorly understood.
  1078. In order to develop methods of immune suppression that either prevent the
  1079. formation of memory cells or work more effectively against memory cells,
  1080. a more complete understanding of the mechanisms of immune memory formation
  1081. and regulation is required.
  1082. \end_layout
  1083. \begin_layout Subsection
  1084. Infusion of allogenic mesenchymal stem cells modulates the alloimmune response
  1085. \end_layout
  1086. \begin_layout Standard
  1087. One promising experimental treatment for transplant rejection involves the
  1088. infusion of allogenic
  1089. \begin_inset Flex Glossary Term (pl)
  1090. status open
  1091. \begin_layout Plain Layout
  1092. MSC
  1093. \end_layout
  1094. \end_inset
  1095. .
  1096. \begin_inset Flex Glossary Term (pl)
  1097. status open
  1098. \begin_layout Plain Layout
  1099. MSC
  1100. \end_layout
  1101. \end_inset
  1102. have been shown to have immune modulatory effects, both in general and
  1103. specifically in the case of immune responses against allografts
  1104. \begin_inset CommandInset citation
  1105. LatexCommand cite
  1106. key "LeBlanc2003,Aggarwal2005,Bartholomew2009,Berman2010"
  1107. literal "false"
  1108. \end_inset
  1109. .
  1110. Furthermore, allogenic
  1111. \begin_inset Flex Glossary Term (pl)
  1112. status open
  1113. \begin_layout Plain Layout
  1114. MSC
  1115. \end_layout
  1116. \end_inset
  1117. themselves are immune-evasive and are rejected by the recipient's immune
  1118. system more slowly than most allogenic tissues
  1119. \begin_inset CommandInset citation
  1120. LatexCommand cite
  1121. key "Ankrum2014,Berglund2017"
  1122. literal "false"
  1123. \end_inset
  1124. .
  1125. In addition, treating
  1126. \begin_inset Flex Glossary Term (pl)
  1127. status open
  1128. \begin_layout Plain Layout
  1129. MSC
  1130. \end_layout
  1131. \end_inset
  1132. in culture with
  1133. \begin_inset Flex Glossary Term
  1134. status open
  1135. \begin_layout Plain Layout
  1136. IFNg
  1137. \end_layout
  1138. \end_inset
  1139. is shown to enhance their immunosuppressive properties and homogenize their
  1140. cellulat phenotype, making them more amenable to development into a well-contro
  1141. lled treatment
  1142. \begin_inset CommandInset citation
  1143. LatexCommand cite
  1144. key "Majumdar2003,Ryan2007"
  1145. literal "false"
  1146. \end_inset
  1147. .
  1148. The mechanisms by which
  1149. \begin_inset Flex Glossary Term (pl)
  1150. status open
  1151. \begin_layout Plain Layout
  1152. MSC
  1153. \end_layout
  1154. \end_inset
  1155. modulate the immune system are still poorly understood.
  1156. Despite this, there is signifcant interest in using
  1157. \begin_inset Flex Glossary Term
  1158. status open
  1159. \begin_layout Plain Layout
  1160. IFNg
  1161. \end_layout
  1162. \end_inset
  1163. -activated
  1164. \begin_inset Flex Glossary Term
  1165. status open
  1166. \begin_layout Plain Layout
  1167. MSC
  1168. \end_layout
  1169. \end_inset
  1170. infusion as a supplementary immune suppressive treatment for allograft
  1171. transplantation.
  1172. \end_layout
  1173. \begin_layout Standard
  1174. Note that despite the name, none of the above properties of
  1175. \begin_inset Flex Glossary Term (pl)
  1176. status open
  1177. \begin_layout Plain Layout
  1178. MSC
  1179. \end_layout
  1180. \end_inset
  1181. are believed to involve their ability as stem cells to differentiate into
  1182. multiple different mature cell types, but rather the intercellular signals
  1183. they produce
  1184. \begin_inset CommandInset citation
  1185. LatexCommand cite
  1186. key "Ankrum2014"
  1187. literal "false"
  1188. \end_inset
  1189. .
  1190. \end_layout
  1191. \begin_layout Section
  1192. \begin_inset CommandInset label
  1193. LatexCommand label
  1194. name "sec:Overview-of-bioinformatic"
  1195. \end_inset
  1196. Overview of bioinformatic analysis methods
  1197. \end_layout
  1198. \begin_layout Standard
  1199. \begin_inset Flex TODO Note (inline)
  1200. status open
  1201. \begin_layout Plain Layout
  1202. Also cite somewhere: R, Bioconductor
  1203. \end_layout
  1204. \end_inset
  1205. \end_layout
  1206. \begin_layout Itemize
  1207. Powerful methods for assaying gene expression and epigenetics across entire
  1208. genomes
  1209. \end_layout
  1210. \begin_layout Itemize
  1211. Proper analysis requires finding and exploiting systematic genome-wide trends
  1212. \end_layout
  1213. \begin_layout Standard
  1214. The studies presented in this work all involve the analysis of high-throughput
  1215. genomic and epigenomic assay data.
  1216. These data present many unique analysis challenges, and a wide array of
  1217. software tools are available to analyze them.
  1218. This section presents an overview of the most important methods and tools
  1219. used throughout the following analyses, including what problems they solve,
  1220. what assumptions they make, and a basic description of how they work.
  1221. \end_layout
  1222. \begin_layout Subsection
  1223. \begin_inset Flex Code
  1224. status open
  1225. \begin_layout Plain Layout
  1226. Limma
  1227. \end_layout
  1228. \end_inset
  1229. : The standard linear modeling framework for genomics
  1230. \end_layout
  1231. \begin_layout Standard
  1232. Linear models are a generalization of the
  1233. \begin_inset Formula $t$
  1234. \end_inset
  1235. -test and ANOVA to arbitrarily complex experimental designs
  1236. \begin_inset CommandInset citation
  1237. LatexCommand cite
  1238. key "chambers:1992"
  1239. literal "false"
  1240. \end_inset
  1241. .
  1242. In a typical linear model, there is one dependent variable observation
  1243. per sample and a large number of samples.
  1244. For example, in a linear model of height as a function of age and sex,
  1245. there is one height measurement per person.
  1246. However, when analyzing genomic data, each sample consists of observations
  1247. of thousands of dependent variables.
  1248. For example, in a
  1249. \begin_inset Flex Glossary Term
  1250. status open
  1251. \begin_layout Plain Layout
  1252. RNA-seq
  1253. \end_layout
  1254. \end_inset
  1255. experiment, the dependent variables may be the count of
  1256. \begin_inset Flex Glossary Term
  1257. status open
  1258. \begin_layout Plain Layout
  1259. RNA-seq
  1260. \end_layout
  1261. \end_inset
  1262. reads for each annotated gene, and there are tens of thousands of genes
  1263. in the human genome.
  1264. Since many assays measure other things than gene expression, the abstract
  1265. term
  1266. \begin_inset Quotes eld
  1267. \end_inset
  1268. feature
  1269. \begin_inset Quotes erd
  1270. \end_inset
  1271. is used to refer to each dependent variable being measured, which may include
  1272. any genomic element, such as genes, promoters, peaks, enhancers, exons,
  1273. etc.
  1274. \end_layout
  1275. \begin_layout Standard
  1276. The simplest approach to analyzing such data would be to fit the same model
  1277. independently to each feature.
  1278. However, this is undesirable for most genomics data sets.
  1279. Genomics assays like
  1280. \begin_inset Flex Glossary Term
  1281. status open
  1282. \begin_layout Plain Layout
  1283. HTS
  1284. \end_layout
  1285. \end_inset
  1286. are expensive, and often the process of generating the samples is also
  1287. quite expensive and time-consuming.
  1288. This expense limits the sample sizes typically employed in genomics experiments
  1289. , so a typical genomic data set has far more features being measured than
  1290. observations (samples) per feature.
  1291. As a result, the statistical power of the linear model for each individual
  1292. feature is likewise limited by the small number of samples.
  1293. However, because thousands of features from the same set of samples are
  1294. analyzed together, there is an opportunity to improve the statistical power
  1295. of the analysis by exploiting shared patterns of variation across features.
  1296. This is the core feature of
  1297. \begin_inset Flex Code
  1298. status open
  1299. \begin_layout Plain Layout
  1300. limma
  1301. \end_layout
  1302. \end_inset
  1303. , a linear modeling framework designed for genomic data.
  1304. \begin_inset Flex Code
  1305. status open
  1306. \begin_layout Plain Layout
  1307. Limma
  1308. \end_layout
  1309. \end_inset
  1310. is typically used to analyze expression microarray data, and more recently
  1311. \begin_inset Flex Glossary Term
  1312. status open
  1313. \begin_layout Plain Layout
  1314. RNA-seq
  1315. \end_layout
  1316. \end_inset
  1317. data, but it can also be used to analyze any other data for which linear
  1318. modeling is appropriate.
  1319. \end_layout
  1320. \begin_layout Standard
  1321. The central challenge when fitting a linear model is to estimate the variance
  1322. of the data accurately.
  1323. Out of all parameters required to evaluate statistical significance of
  1324. an effect, the variance is the most difficult to estimate when sample sizes
  1325. are small.
  1326. A single shared variance could be estimated for all of the features together,
  1327. and this estimate would be very stable, in contrast to the individual feature
  1328. variance estimates.
  1329. However, this would require the assumption that all features have equal
  1330. variance, which is known to be false for most genomic data sets (for example,
  1331. some genes' expression is known to be more variable than others').
  1332. \begin_inset Flex Code
  1333. status open
  1334. \begin_layout Plain Layout
  1335. Limma
  1336. \end_layout
  1337. \end_inset
  1338. offers a compromise between these two extremes by using a method called
  1339. empirical Bayes moderation to
  1340. \begin_inset Quotes eld
  1341. \end_inset
  1342. squeeze
  1343. \begin_inset Quotes erd
  1344. \end_inset
  1345. the distribution of estimated variances toward a single common value that
  1346. represents the variance of an average feature in the data (Figure
  1347. \begin_inset CommandInset ref
  1348. LatexCommand ref
  1349. reference "fig:ebayes-example"
  1350. plural "false"
  1351. caps "false"
  1352. noprefix "false"
  1353. \end_inset
  1354. )
  1355. \begin_inset CommandInset citation
  1356. LatexCommand cite
  1357. key "Smyth2004"
  1358. literal "false"
  1359. \end_inset
  1360. .
  1361. While the individual feature variance estimates are not stable, the common
  1362. variance estimate for the entire data set is quite stable, so using a combinati
  1363. on of the two yields a variance estimate for each feature with greater precision
  1364. than the individual feature variances.
  1365. The trade-off for this improvement is that squeezing each estimated variance
  1366. toward the common value introduces some bias – the variance will be underestima
  1367. ted for features with high variance and overestimated for features with
  1368. low variance.
  1369. Essentially,
  1370. \begin_inset Flex Code
  1371. status open
  1372. \begin_layout Plain Layout
  1373. limma
  1374. \end_layout
  1375. \end_inset
  1376. assumes that extreme variances are less common than variances close to
  1377. the common value.
  1378. The squeezed variance estimates from this empirical Bayes procedure are
  1379. shown empirically to yield greater statistical power than either the individual
  1380. feature variances or the single common value.
  1381. \end_layout
  1382. \begin_layout Standard
  1383. \begin_inset Float figure
  1384. wide false
  1385. sideways false
  1386. status collapsed
  1387. \begin_layout Plain Layout
  1388. \align center
  1389. \begin_inset Graphics
  1390. filename graphics/Intro/eBayes-CROP-RASTER.png
  1391. lyxscale 25
  1392. width 100col%
  1393. groupId colwidth-raster
  1394. \end_inset
  1395. \end_layout
  1396. \begin_layout Plain Layout
  1397. \begin_inset Caption Standard
  1398. \begin_layout Plain Layout
  1399. \begin_inset Argument 1
  1400. status collapsed
  1401. \begin_layout Plain Layout
  1402. Example of empirical Bayes squeezing of per-gene variances.
  1403. \end_layout
  1404. \end_inset
  1405. \begin_inset CommandInset label
  1406. LatexCommand label
  1407. name "fig:ebayes-example"
  1408. \end_inset
  1409. \series bold
  1410. Example of empirical Bayes squeezing of per-gene variances.
  1411. \series default
  1412. A smooth trend line (red) is fitted to the individual gene variances (light
  1413. blue) as a function of average gene abundance (logCPM).
  1414. Then the individual gene variances are
  1415. \begin_inset Quotes eld
  1416. \end_inset
  1417. squeezed
  1418. \begin_inset Quotes erd
  1419. \end_inset
  1420. toward the trend (dark blue).
  1421. \end_layout
  1422. \end_inset
  1423. \end_layout
  1424. \begin_layout Plain Layout
  1425. \end_layout
  1426. \end_inset
  1427. \end_layout
  1428. \begin_layout Standard
  1429. On top of this core framework,
  1430. \begin_inset Flex Code
  1431. status open
  1432. \begin_layout Plain Layout
  1433. limma
  1434. \end_layout
  1435. \end_inset
  1436. also implements many other enhancements that, further relax the assumptions
  1437. of the model and extend the scope of what kinds of data it can analyze.
  1438. Instead of squeezing toward a single common variance value,
  1439. \begin_inset Flex Code
  1440. status open
  1441. \begin_layout Plain Layout
  1442. limma
  1443. \end_layout
  1444. \end_inset
  1445. can model the common variance as a function of a covariate, such as average
  1446. expression
  1447. \begin_inset CommandInset citation
  1448. LatexCommand cite
  1449. key "Law2014"
  1450. literal "false"
  1451. \end_inset
  1452. .
  1453. This is essential for
  1454. \begin_inset Flex Glossary Term
  1455. status open
  1456. \begin_layout Plain Layout
  1457. RNA-seq
  1458. \end_layout
  1459. \end_inset
  1460. data, where higher gene counts yield more precise expression measurements
  1461. and therefore smaller variances than low-count genes.
  1462. While linear models typically assume that all samples have equal variance,
  1463. \begin_inset Flex Code
  1464. status open
  1465. \begin_layout Plain Layout
  1466. limma
  1467. \end_layout
  1468. \end_inset
  1469. is able to relax this assumption by identifying and down-weighting samples
  1470. that diverge more strongly from the linear model across many features
  1471. \begin_inset CommandInset citation
  1472. LatexCommand cite
  1473. key "Ritchie2006,Liu2015"
  1474. literal "false"
  1475. \end_inset
  1476. .
  1477. In addition,
  1478. \begin_inset Flex Code
  1479. status open
  1480. \begin_layout Plain Layout
  1481. limma
  1482. \end_layout
  1483. \end_inset
  1484. is also able to fit simple mixed models incorporating one random effect
  1485. in addition to the fixed effects represented by an ordinary linear model
  1486. \begin_inset CommandInset citation
  1487. LatexCommand cite
  1488. key "Smyth2005a"
  1489. literal "false"
  1490. \end_inset
  1491. .
  1492. Once again,
  1493. \begin_inset Flex Code
  1494. status open
  1495. \begin_layout Plain Layout
  1496. limma
  1497. \end_layout
  1498. \end_inset
  1499. shares information between features to obtain a robust estimate for the
  1500. random effect correlation.
  1501. \end_layout
  1502. \begin_layout Subsection
  1503. \begin_inset Flex Code
  1504. status open
  1505. \begin_layout Plain Layout
  1506. edgeR
  1507. \end_layout
  1508. \end_inset
  1509. provides
  1510. \begin_inset Flex Code
  1511. status open
  1512. \begin_layout Plain Layout
  1513. limma
  1514. \end_layout
  1515. \end_inset
  1516. -like analysis features for read count data
  1517. \end_layout
  1518. \begin_layout Standard
  1519. Although
  1520. \begin_inset Flex Code
  1521. status open
  1522. \begin_layout Plain Layout
  1523. limma
  1524. \end_layout
  1525. \end_inset
  1526. can be applied to read counts from
  1527. \begin_inset Flex Glossary Term
  1528. status open
  1529. \begin_layout Plain Layout
  1530. RNA-seq
  1531. \end_layout
  1532. \end_inset
  1533. data, it is less suitable for counts from
  1534. \begin_inset Flex Glossary Term
  1535. status open
  1536. \begin_layout Plain Layout
  1537. ChIP-seq
  1538. \end_layout
  1539. \end_inset
  1540. and other sources, which tend to be much smaller and therefore violate
  1541. the assumption of a normal distribution more severely.
  1542. For all count-based data, the
  1543. \begin_inset Flex Code
  1544. status open
  1545. \begin_layout Plain Layout
  1546. edgeR
  1547. \end_layout
  1548. \end_inset
  1549. package works similarly to
  1550. \begin_inset Flex Code
  1551. status open
  1552. \begin_layout Plain Layout
  1553. limma
  1554. \end_layout
  1555. \end_inset
  1556. , but uses a
  1557. \begin_inset Flex Glossary Term
  1558. status open
  1559. \begin_layout Plain Layout
  1560. GLM
  1561. \end_layout
  1562. \end_inset
  1563. instead of a linear model.
  1564. Relative to a linear model, a
  1565. \begin_inset Flex Glossary Term
  1566. status open
  1567. \begin_layout Plain Layout
  1568. GLM
  1569. \end_layout
  1570. \end_inset
  1571. gains flexibility by relaxing several assumptions, the most important of
  1572. which is the assumption of normally distributed errors.
  1573. This allows the
  1574. \begin_inset Flex Glossary Term
  1575. status open
  1576. \begin_layout Plain Layout
  1577. GLM
  1578. \end_layout
  1579. \end_inset
  1580. in
  1581. \begin_inset Flex Code
  1582. status open
  1583. \begin_layout Plain Layout
  1584. edgeR
  1585. \end_layout
  1586. \end_inset
  1587. to model the counts directly using a
  1588. \begin_inset Flex Glossary Term
  1589. status open
  1590. \begin_layout Plain Layout
  1591. NB
  1592. \end_layout
  1593. \end_inset
  1594. distribution rather than modeling the normalized log counts using a normal
  1595. distribution as
  1596. \begin_inset Flex Code
  1597. status open
  1598. \begin_layout Plain Layout
  1599. limma
  1600. \end_layout
  1601. \end_inset
  1602. does
  1603. \begin_inset CommandInset citation
  1604. LatexCommand cite
  1605. key "Chen2014,McCarthy2012,Robinson2010a"
  1606. literal "false"
  1607. \end_inset
  1608. .
  1609. \end_layout
  1610. \begin_layout Standard
  1611. The
  1612. \begin_inset Flex Glossary Term
  1613. status open
  1614. \begin_layout Plain Layout
  1615. NB
  1616. \end_layout
  1617. \end_inset
  1618. distribution is a good fit for count data because it can be derived as
  1619. a gamma-distributed mixture of Poisson distributions.
  1620. The reads in an
  1621. \begin_inset Flex Glossary Term
  1622. status open
  1623. \begin_layout Plain Layout
  1624. RNA-seq
  1625. \end_layout
  1626. \end_inset
  1627. sample are assumed to be sampled from a much larger population, such that
  1628. the sampling process does not significantly affect the proportions.
  1629. Under this assumption, a gene's read count in an
  1630. \begin_inset Flex Glossary Term
  1631. status open
  1632. \begin_layout Plain Layout
  1633. RNA-seq
  1634. \end_layout
  1635. \end_inset
  1636. sample is distributed as
  1637. \begin_inset Formula $\mathrm{Binomial}(n,p)$
  1638. \end_inset
  1639. , where
  1640. \begin_inset Formula $n$
  1641. \end_inset
  1642. is the total number of reads sequenced from the sample and
  1643. \begin_inset Formula $p$
  1644. \end_inset
  1645. is the proportion of total fragments in the sample derived from that gene.
  1646. When
  1647. \begin_inset Formula $n$
  1648. \end_inset
  1649. is large and
  1650. \begin_inset Formula $p$
  1651. \end_inset
  1652. is small, a
  1653. \begin_inset Formula $\mathrm{Binomial}(n,p)$
  1654. \end_inset
  1655. distribution is well-approximated by
  1656. \begin_inset Formula $\mathrm{Poisson}(np)$
  1657. \end_inset
  1658. .
  1659. Hence, if multiple sequencing runs are performed on the same
  1660. \begin_inset Flex Glossary Term
  1661. status open
  1662. \begin_layout Plain Layout
  1663. RNA-seq
  1664. \end_layout
  1665. \end_inset
  1666. sample (with the same gene mixing proportions each time), each gene's read
  1667. count is expected to follow a Poisson distribution.
  1668. If the abundance of a gene,
  1669. \begin_inset Formula $p,$
  1670. \end_inset
  1671. varies across biological replicates according to a gamma distribution,
  1672. and
  1673. \begin_inset Formula $n$
  1674. \end_inset
  1675. is held constant, then the result is a gamma-distributed mixture of Poisson
  1676. distributions, which is equivalent to the
  1677. \begin_inset Flex Glossary Term
  1678. status open
  1679. \begin_layout Plain Layout
  1680. NB
  1681. \end_layout
  1682. \end_inset
  1683. distribution.
  1684. The assumption of a gamma distribution for the mixing weights is arbitrary,
  1685. motivated by the convenience of the numerically tractable
  1686. \begin_inset Flex Glossary Term
  1687. status open
  1688. \begin_layout Plain Layout
  1689. NB
  1690. \end_layout
  1691. \end_inset
  1692. distribution and the need to select
  1693. \emph on
  1694. some
  1695. \emph default
  1696. distribution, since the true shape of the distribution of biological variance
  1697. is unknown.
  1698. \end_layout
  1699. \begin_layout Standard
  1700. Thus,
  1701. \begin_inset Flex Code
  1702. status open
  1703. \begin_layout Plain Layout
  1704. edgeR
  1705. \end_layout
  1706. \end_inset
  1707. 's use of the
  1708. \begin_inset Flex Glossary Term
  1709. status open
  1710. \begin_layout Plain Layout
  1711. NB
  1712. \end_layout
  1713. \end_inset
  1714. is equivalent to an
  1715. \emph on
  1716. a priori
  1717. \emph default
  1718. assumption that the variation in gene abundances between replicates follows
  1719. a gamma distribution.
  1720. The gamma shape parameter in the context of the
  1721. \begin_inset Flex Glossary Term
  1722. status open
  1723. \begin_layout Plain Layout
  1724. NB
  1725. \end_layout
  1726. \end_inset
  1727. is called the dispersion, and the square root of this dispersion is referred
  1728. to as the
  1729. \begin_inset Flex Glossary Term
  1730. status open
  1731. \begin_layout Plain Layout
  1732. BCV
  1733. \end_layout
  1734. \end_inset
  1735. , since it represents the variability in abundance that was present in the
  1736. biological samples prior to the Poisson
  1737. \begin_inset Quotes eld
  1738. \end_inset
  1739. noise
  1740. \begin_inset Quotes erd
  1741. \end_inset
  1742. that was generated by the random sampling of reads in proportion to feature
  1743. abundances.
  1744. Like
  1745. \begin_inset Flex Code
  1746. status open
  1747. \begin_layout Plain Layout
  1748. limma
  1749. \end_layout
  1750. \end_inset
  1751. ,
  1752. \begin_inset Flex Code
  1753. status open
  1754. \begin_layout Plain Layout
  1755. edgeR
  1756. \end_layout
  1757. \end_inset
  1758. estimates the
  1759. \begin_inset Flex Glossary Term
  1760. status open
  1761. \begin_layout Plain Layout
  1762. BCV
  1763. \end_layout
  1764. \end_inset
  1765. for each feature using an empirical Bayes procedure that represents a compromis
  1766. e between per-feature dispersions and a single pooled dispersion estimate
  1767. shared across all features.
  1768. For differential abundance testing,
  1769. \begin_inset Flex Code
  1770. status open
  1771. \begin_layout Plain Layout
  1772. edgeR
  1773. \end_layout
  1774. \end_inset
  1775. offers a likelihood ratio test based on the
  1776. \begin_inset Flex Glossary Term
  1777. status open
  1778. \begin_layout Plain Layout
  1779. NB
  1780. \end_layout
  1781. \end_inset
  1782. \begin_inset Flex Glossary Term
  1783. status open
  1784. \begin_layout Plain Layout
  1785. GLM
  1786. \end_layout
  1787. \end_inset
  1788. .
  1789. However, this test assumes the dispersion parameter is known exactly rather
  1790. than estimated from the data, which can result in overstating the significance
  1791. of differential abundance results.
  1792. More recently, a quasi-likelihood test has been introduced that properly
  1793. factors the uncertainty in dispersion estimation into the estimates of
  1794. statistical significance, and this test is recommended over the likelihood
  1795. ratio test in most cases
  1796. \begin_inset CommandInset citation
  1797. LatexCommand cite
  1798. key "Lund2012"
  1799. literal "false"
  1800. \end_inset
  1801. .
  1802. \end_layout
  1803. \begin_layout Subsection
  1804. Calling consensus peaks from ChIP-seq data
  1805. \end_layout
  1806. \begin_layout Standard
  1807. Unlike
  1808. \begin_inset Flex Glossary Term
  1809. status open
  1810. \begin_layout Plain Layout
  1811. RNA-seq
  1812. \end_layout
  1813. \end_inset
  1814. data, in which gene annotations provide a well-defined set of discrete
  1815. genomic regions in which to count reads,
  1816. \begin_inset Flex Glossary Term
  1817. status open
  1818. \begin_layout Plain Layout
  1819. ChIP-seq
  1820. \end_layout
  1821. \end_inset
  1822. reads can potentially occur anywhere in the genome.
  1823. However, most genome regions will not contain significant
  1824. \begin_inset Flex Glossary Term
  1825. status open
  1826. \begin_layout Plain Layout
  1827. ChIP-seq
  1828. \end_layout
  1829. \end_inset
  1830. read coverage, and analyzing every position in the entire genome is statistical
  1831. ly and computationally infeasible, so it is necessary to identify regions
  1832. of interest inside which
  1833. \begin_inset Flex Glossary Term
  1834. status open
  1835. \begin_layout Plain Layout
  1836. ChIP-seq
  1837. \end_layout
  1838. \end_inset
  1839. reads will be counted and analyzed.
  1840. One option is to define a set of interesting regions
  1841. \emph on
  1842. a priori
  1843. \emph default
  1844. , for example by defining a promoter region for each annotated gene.
  1845. However, it is also possible to use the
  1846. \begin_inset Flex Glossary Term
  1847. status open
  1848. \begin_layout Plain Layout
  1849. ChIP-seq
  1850. \end_layout
  1851. \end_inset
  1852. data itself to identify regions with
  1853. \begin_inset Flex Glossary Term
  1854. status open
  1855. \begin_layout Plain Layout
  1856. ChIP-seq
  1857. \end_layout
  1858. \end_inset
  1859. read coverage significantly above the background level, known as peaks.
  1860. \end_layout
  1861. \begin_layout Standard
  1862. The challenge in peak calling is that the immunoprecipitation step is not
  1863. 100% selective, so some fraction of reads are
  1864. \emph on
  1865. not
  1866. \emph default
  1867. derived from DNA fragments that were bound by the immunoprecipitated protein.
  1868. These are referred to as background reads.
  1869. Biases in amplification and sequencing, as well as the aforementioned Poisson
  1870. randomness of the sequencing itself, can cause fluctuations in the background
  1871. level of reads that resemble peaks, and the true peaks must be distinguished
  1872. from these.
  1873. It is common to sequence the input DNA to the ChIP-seq reaction alongside
  1874. the immunoprecipitated product in order to aid in estimating the fluctuations
  1875. in background level across the genome.
  1876. \end_layout
  1877. \begin_layout Standard
  1878. There are generally two kinds of peaks that can be identified: narrow peaks
  1879. and broadly enriched regions.
  1880. Proteins that bind specific sites in the genome (such as many transcription
  1881. factors) typically show most of their
  1882. \begin_inset Flex Glossary Term
  1883. status open
  1884. \begin_layout Plain Layout
  1885. ChIP-seq
  1886. \end_layout
  1887. \end_inset
  1888. read coverage at these specific sites and very little coverage anywhere
  1889. else.
  1890. Because the footprint of the protein is consistent wherever it binds, each
  1891. peak has a consistent width, typically tens to hundreds of base pairs,
  1892. representing the length of DNA that it binds to.
  1893. Algorithms like
  1894. \begin_inset Flex Glossary Term
  1895. status open
  1896. \begin_layout Plain Layout
  1897. MACS
  1898. \end_layout
  1899. \end_inset
  1900. exploit this pattern to identify specific loci at which such
  1901. \begin_inset Quotes eld
  1902. \end_inset
  1903. narrow peaks
  1904. \begin_inset Quotes erd
  1905. \end_inset
  1906. occur by looking for the characteristic peak shape in the
  1907. \begin_inset Flex Glossary Term
  1908. status open
  1909. \begin_layout Plain Layout
  1910. ChIP-seq
  1911. \end_layout
  1912. \end_inset
  1913. coverage rising above the surrounding background coverage
  1914. \begin_inset CommandInset citation
  1915. LatexCommand cite
  1916. key "Zhang2008"
  1917. literal "false"
  1918. \end_inset
  1919. .
  1920. In contrast, some proteins, chief among them histones, do not bind only
  1921. at a small number of specific sites, but rather bind potentially almost
  1922. everywhere in the entire genome.
  1923. When looking at histone marks, adjacent histones tend to be similarly marked,
  1924. and a given mark may be present on an arbitrary number of consecutive histones
  1925. along the genome.
  1926. Hence, there is no consistent
  1927. \begin_inset Quotes eld
  1928. \end_inset
  1929. footprint size
  1930. \begin_inset Quotes erd
  1931. \end_inset
  1932. for
  1933. \begin_inset Flex Glossary Term
  1934. status open
  1935. \begin_layout Plain Layout
  1936. ChIP-seq
  1937. \end_layout
  1938. \end_inset
  1939. peaks based on histone marks, and peaks typically span many histones.
  1940. Hence, typical peaks span many hundreds or even thousands of base pairs.
  1941. Instead of identifying specific loci of strong enrichment, algorithms like
  1942. \begin_inset Flex Glossary Term
  1943. status open
  1944. \begin_layout Plain Layout
  1945. SICER
  1946. \end_layout
  1947. \end_inset
  1948. assume that peaks are represented in the
  1949. \begin_inset Flex Glossary Term
  1950. status open
  1951. \begin_layout Plain Layout
  1952. ChIP-seq
  1953. \end_layout
  1954. \end_inset
  1955. data by modest enrichment above background occurring across broad regions,
  1956. and they attempt to identify the extent of those regions
  1957. \begin_inset CommandInset citation
  1958. LatexCommand cite
  1959. key "Zang2009"
  1960. literal "false"
  1961. \end_inset
  1962. .
  1963. \end_layout
  1964. \begin_layout Standard
  1965. Regardless of the type of peak identified, it is important to identify peaks
  1966. that occur consistently across biological replicates.
  1967. The
  1968. \begin_inset Flex Glossary Term
  1969. status open
  1970. \begin_layout Plain Layout
  1971. ENCODE
  1972. \end_layout
  1973. \end_inset
  1974. project has developed a method called
  1975. \begin_inset Flex Glossary Term
  1976. status open
  1977. \begin_layout Plain Layout
  1978. IDR
  1979. \end_layout
  1980. \end_inset
  1981. for this purpose
  1982. \begin_inset CommandInset citation
  1983. LatexCommand cite
  1984. key "Li2006"
  1985. literal "false"
  1986. \end_inset
  1987. .
  1988. The
  1989. \begin_inset Flex Glossary Term
  1990. status open
  1991. \begin_layout Plain Layout
  1992. IDR
  1993. \end_layout
  1994. \end_inset
  1995. is defined as the probability that a peak identified in one biological
  1996. replicate will
  1997. \emph on
  1998. not
  1999. \emph default
  2000. also be identified in a second replicate.
  2001. Where the more familiar false discovery rate measures the degree of corresponde
  2002. nce between a data-derived ranked list and the (unknown) true list of significan
  2003. t features,
  2004. \begin_inset Flex Glossary Term
  2005. status open
  2006. \begin_layout Plain Layout
  2007. IDR
  2008. \end_layout
  2009. \end_inset
  2010. instead measures the degree of correspondence between two ranked lists
  2011. derived from different data.
  2012. \begin_inset Flex Glossary Term
  2013. status open
  2014. \begin_layout Plain Layout
  2015. IDR
  2016. \end_layout
  2017. \end_inset
  2018. assumes that the highest-ranked features are
  2019. \begin_inset Quotes eld
  2020. \end_inset
  2021. signal
  2022. \begin_inset Quotes erd
  2023. \end_inset
  2024. peaks that tend to be listed in the same order in both lists, while the
  2025. lowest-ranked features are essentially noise peaks, listed in random order
  2026. with no correspondence between the lists.
  2027. \begin_inset Flex Glossary Term (Capital)
  2028. status open
  2029. \begin_layout Plain Layout
  2030. IDR
  2031. \end_layout
  2032. \end_inset
  2033. attempts to locate the
  2034. \begin_inset Quotes eld
  2035. \end_inset
  2036. crossover point
  2037. \begin_inset Quotes erd
  2038. \end_inset
  2039. between the signal and the noise by determining how far down the list the
  2040. rank consistency breaks down into randomness (Figure
  2041. \begin_inset CommandInset ref
  2042. LatexCommand ref
  2043. reference "fig:Example-IDR"
  2044. plural "false"
  2045. caps "false"
  2046. noprefix "false"
  2047. \end_inset
  2048. ).
  2049. \end_layout
  2050. \begin_layout Standard
  2051. \begin_inset Float figure
  2052. wide false
  2053. sideways false
  2054. status open
  2055. \begin_layout Plain Layout
  2056. \align center
  2057. \begin_inset Graphics
  2058. filename graphics/CD4-csaw/IDR/D4659vsD5053_epic-PAGE1-CROP-RASTER.png
  2059. lyxscale 25
  2060. width 100col%
  2061. groupId colwidth-raster
  2062. \end_inset
  2063. \end_layout
  2064. \begin_layout Plain Layout
  2065. \begin_inset Caption Standard
  2066. \begin_layout Plain Layout
  2067. \begin_inset Argument 1
  2068. status collapsed
  2069. \begin_layout Plain Layout
  2070. Example IDR consistency plot.
  2071. \end_layout
  2072. \end_inset
  2073. \begin_inset CommandInset label
  2074. LatexCommand label
  2075. name "fig:Example-IDR"
  2076. \end_inset
  2077. \series bold
  2078. Example IDR consistency plot.
  2079. \series default
  2080. Peak calls in two replicates are ranked from highest score (top and right)
  2081. to lowest score (bottom and left).
  2082. IDR identifies reproducible peaks, which rank highly in both replicates
  2083. (light blue), separating them from
  2084. \begin_inset Quotes eld
  2085. \end_inset
  2086. noise
  2087. \begin_inset Quotes erd
  2088. \end_inset
  2089. peak calls whose ranking is not reproducible between replicates (dark blue).
  2090. \end_layout
  2091. \end_inset
  2092. \end_layout
  2093. \begin_layout Plain Layout
  2094. \end_layout
  2095. \end_inset
  2096. \end_layout
  2097. \begin_layout Standard
  2098. In addition to other considerations, if called peaks are to be used as regions
  2099. of interest for differential abundance analysis, then care must be taken
  2100. to call peaks in a way that is blind to differential abundance between
  2101. experimental conditions, or else the statistical significance calculations
  2102. for differential abundance will overstate their confidence in the results.
  2103. The
  2104. \begin_inset Flex Code
  2105. status open
  2106. \begin_layout Plain Layout
  2107. csaw
  2108. \end_layout
  2109. \end_inset
  2110. package provides guidelines for calling peaks in this way: peaks are called
  2111. based on a combination of all
  2112. \begin_inset Flex Glossary Term
  2113. status open
  2114. \begin_layout Plain Layout
  2115. ChIP-seq
  2116. \end_layout
  2117. \end_inset
  2118. reads from all experimental conditions, so that the identified peaks are
  2119. based on the average abundance across all conditions, which is independent
  2120. of any differential abundance between conditions
  2121. \begin_inset CommandInset citation
  2122. LatexCommand cite
  2123. key "Lun2015a"
  2124. literal "false"
  2125. \end_inset
  2126. .
  2127. \end_layout
  2128. \begin_layout Subsection
  2129. Normalization of high-throughput data is non-trivial and application-dependent
  2130. \end_layout
  2131. \begin_layout Standard
  2132. High-throughput data sets invariably require some kind of normalization
  2133. before further analysis can be conducted.
  2134. In general, the goal of normalization is to remove effects in the data
  2135. that are caused by technical factors that have nothing to do with the biology
  2136. being studied.
  2137. \end_layout
  2138. \begin_layout Standard
  2139. For Affymetrix expression arrays, the standard normalization algorithm used
  2140. in most analyses is
  2141. \begin_inset Flex Glossary Term
  2142. status open
  2143. \begin_layout Plain Layout
  2144. RMA
  2145. \end_layout
  2146. \end_inset
  2147. \begin_inset CommandInset citation
  2148. LatexCommand cite
  2149. key "Irizarry2003a"
  2150. literal "false"
  2151. \end_inset
  2152. .
  2153. \begin_inset Flex Glossary Term
  2154. status open
  2155. \begin_layout Plain Layout
  2156. RMA
  2157. \end_layout
  2158. \end_inset
  2159. is designed with the assumption that some fraction of probes on each array
  2160. will be artifactual and takes advantage of the fact that each gene is represent
  2161. ed by multiple probes by implementing normalization and summarization steps
  2162. that are robust against outlier probes.
  2163. However,
  2164. \begin_inset Flex Glossary Term
  2165. status open
  2166. \begin_layout Plain Layout
  2167. RMA
  2168. \end_layout
  2169. \end_inset
  2170. uses the probe intensities of all arrays in the data set in the normalization
  2171. of each individual array, meaning that the normalized expression values
  2172. in each array depend on every array in the data set, and will necessarily
  2173. change each time an array is added or removed from the data set.
  2174. If this is undesirable,
  2175. \begin_inset Flex Glossary Term
  2176. status open
  2177. \begin_layout Plain Layout
  2178. fRMA
  2179. \end_layout
  2180. \end_inset
  2181. implements a variant of
  2182. \begin_inset Flex Glossary Term
  2183. status open
  2184. \begin_layout Plain Layout
  2185. RMA
  2186. \end_layout
  2187. \end_inset
  2188. where the relevant distributional parameters are learned from a large reference
  2189. set of diverse public array data sets and then
  2190. \begin_inset Quotes eld
  2191. \end_inset
  2192. frozen
  2193. \begin_inset Quotes erd
  2194. \end_inset
  2195. , so that each array is effectively normalized against this frozen reference
  2196. set rather than the other arrays in the data set under study
  2197. \begin_inset CommandInset citation
  2198. LatexCommand cite
  2199. key "McCall2010"
  2200. literal "false"
  2201. \end_inset
  2202. .
  2203. Other available array normalization methods considered include dChip,
  2204. \begin_inset Flex Glossary Term
  2205. status open
  2206. \begin_layout Plain Layout
  2207. GRSN
  2208. \end_layout
  2209. \end_inset
  2210. , and
  2211. \begin_inset Flex Glossary Term
  2212. status open
  2213. \begin_layout Plain Layout
  2214. SCAN
  2215. \end_layout
  2216. \end_inset
  2217. \begin_inset CommandInset citation
  2218. LatexCommand cite
  2219. key "Li2001,Pelz2008,Piccolo2012"
  2220. literal "false"
  2221. \end_inset
  2222. .
  2223. \end_layout
  2224. \begin_layout Standard
  2225. In contrast,
  2226. \begin_inset Flex Glossary Term
  2227. status open
  2228. \begin_layout Plain Layout
  2229. HTS
  2230. \end_layout
  2231. \end_inset
  2232. data present very different normalization challenges.
  2233. The simplest case is
  2234. \begin_inset Flex Glossary Term
  2235. status open
  2236. \begin_layout Plain Layout
  2237. RNA-seq
  2238. \end_layout
  2239. \end_inset
  2240. in which read counts are obtained for a set of gene annotations, yielding
  2241. a matrix of counts with rows representing genes and columns representing
  2242. samples.
  2243. Because
  2244. \begin_inset Flex Glossary Term
  2245. status open
  2246. \begin_layout Plain Layout
  2247. RNA-seq
  2248. \end_layout
  2249. \end_inset
  2250. approximates a process of sampling from a population with replacement,
  2251. each gene's count is only interpretable as a fraction of the total reads
  2252. for that sample.
  2253. For that reason,
  2254. \begin_inset Flex Glossary Term
  2255. status open
  2256. \begin_layout Plain Layout
  2257. RNA-seq
  2258. \end_layout
  2259. \end_inset
  2260. abundances are often reported as
  2261. \begin_inset Flex Glossary Term
  2262. status open
  2263. \begin_layout Plain Layout
  2264. CPM
  2265. \end_layout
  2266. \end_inset
  2267. .
  2268. Furthermore, if the abundance of a single gene increases, then in order
  2269. for its fraction of the total reads to increase, all other genes' fractions
  2270. must decrease to accommodate it.
  2271. This effect is known as composition bias, and it is an artifact of the
  2272. read sampling process that has nothing to do with the biology of the samples
  2273. and must therefore be normalized out.
  2274. The most commonly used methods to normalize for composition bias in
  2275. \begin_inset Flex Glossary Term
  2276. status open
  2277. \begin_layout Plain Layout
  2278. RNA-seq
  2279. \end_layout
  2280. \end_inset
  2281. data seek to equalize the average gene abundance across samples, under
  2282. the assumption that the average gene is likely not changing
  2283. \begin_inset CommandInset citation
  2284. LatexCommand cite
  2285. key "Robinson2010,Anders2010"
  2286. literal "false"
  2287. \end_inset
  2288. .
  2289. The effect of such normalizations is to center the distribution of
  2290. \begin_inset Flex Glossary Term (pl)
  2291. status open
  2292. \begin_layout Plain Layout
  2293. logFC
  2294. \end_layout
  2295. \end_inset
  2296. at zero.
  2297. Note that if a true global difference in gene expression is present in
  2298. the data, this difference will be normalized out as well, since it is indisting
  2299. uishable from composition bias.
  2300. In other words,
  2301. \begin_inset Flex Glossary Term
  2302. status open
  2303. \begin_layout Plain Layout
  2304. RNA-seq
  2305. \end_layout
  2306. \end_inset
  2307. cannot measure absolute gene expression, only gene expression as a fraction
  2308. of total reads.
  2309. \end_layout
  2310. \begin_layout Standard
  2311. In
  2312. \begin_inset Flex Glossary Term
  2313. status open
  2314. \begin_layout Plain Layout
  2315. ChIP-seq
  2316. \end_layout
  2317. \end_inset
  2318. data, normalization is not as straightforward.
  2319. The
  2320. \begin_inset Flex Code
  2321. status open
  2322. \begin_layout Plain Layout
  2323. csaw
  2324. \end_layout
  2325. \end_inset
  2326. package implements several different normalization strategies and provides
  2327. guidance on when to use each one
  2328. \begin_inset CommandInset citation
  2329. LatexCommand cite
  2330. key "Lun2015a"
  2331. literal "false"
  2332. \end_inset
  2333. .
  2334. Briefly, a typical
  2335. \begin_inset Flex Glossary Term
  2336. status open
  2337. \begin_layout Plain Layout
  2338. ChIP-seq
  2339. \end_layout
  2340. \end_inset
  2341. sample has a bimodal distribution of read counts: a low-abundance mode
  2342. representing background regions and a high-abundance mode representing
  2343. signal regions.
  2344. This offers two mutually incompatible normalization strategies: equalizing
  2345. background coverage or equalizing signal coverage (Figure
  2346. \begin_inset CommandInset ref
  2347. LatexCommand ref
  2348. reference "fig:chipseq-norm-example"
  2349. plural "false"
  2350. caps "false"
  2351. noprefix "false"
  2352. \end_inset
  2353. ).
  2354. If the experiment is well controlled and
  2355. \begin_inset Flex Glossary Term
  2356. status open
  2357. \begin_layout Plain Layout
  2358. ChIP
  2359. \end_layout
  2360. \end_inset
  2361. efficiency is known to be consistent across all samples, then normalizing
  2362. the background coverage to be equal across all samples is a reasonable
  2363. strategy.
  2364. If this is not a safe assumption, then the preferred strategy is to normalize
  2365. the signal regions in a way similar to
  2366. \begin_inset Flex Glossary Term
  2367. status open
  2368. \begin_layout Plain Layout
  2369. RNA-seq
  2370. \end_layout
  2371. \end_inset
  2372. data by assuming that the average signal region is not changing abundance
  2373. between samples.
  2374. Beyond this, if a
  2375. \begin_inset Flex Glossary Term
  2376. status open
  2377. \begin_layout Plain Layout
  2378. ChIP-seq
  2379. \end_layout
  2380. \end_inset
  2381. experiment has a more complicated structure that doesn't show the typical
  2382. bimodal count distribution, it may be necessary to implement a normalization
  2383. as a smooth function of abundance.
  2384. However, this strategy makes a much stronger assumption about the data:
  2385. that the average
  2386. \begin_inset Flex Glossary Term
  2387. status open
  2388. \begin_layout Plain Layout
  2389. logFC
  2390. \end_layout
  2391. \end_inset
  2392. is zero across all abundance levels.
  2393. Hence, the simpler scaling normalization based on background or signal
  2394. regions are generally preferred whenever possible.
  2395. \end_layout
  2396. \begin_layout Standard
  2397. \begin_inset Float figure
  2398. wide false
  2399. sideways false
  2400. status open
  2401. \begin_layout Plain Layout
  2402. \align center
  2403. \begin_inset Graphics
  2404. filename graphics/CD4-csaw/ChIP-seq/H3K4me2-sample-MAplot-bins-CROP.png
  2405. lyxscale 25
  2406. width 100col%
  2407. groupId colwidth-raster
  2408. \end_inset
  2409. \end_layout
  2410. \begin_layout Plain Layout
  2411. \begin_inset Caption Standard
  2412. \begin_layout Plain Layout
  2413. \begin_inset Argument 1
  2414. status collapsed
  2415. \begin_layout Plain Layout
  2416. Example MA plot of ChIP-seq read counts in 10kb bins for two arbitrary samples.
  2417. \end_layout
  2418. \end_inset
  2419. \begin_inset CommandInset label
  2420. LatexCommand label
  2421. name "fig:chipseq-norm-example"
  2422. \end_inset
  2423. \series bold
  2424. Example MA plot of ChIP-seq read counts in 10kb bins for two arbitrary samples.
  2425. \series default
  2426. The distribution of bins is bimodal along the x axis (average abundance),
  2427. with the left mode representing
  2428. \begin_inset Quotes eld
  2429. \end_inset
  2430. background
  2431. \begin_inset Quotes erd
  2432. \end_inset
  2433. regions with no protein binding and the right mode representing bound regions.
  2434. The modes are also separated on the y axis (logFC), motivating two conflicting
  2435. normalization strategies: background normalization (red) and signal normalizati
  2436. on (blue and green, two similar signal normalizations).
  2437. \end_layout
  2438. \end_inset
  2439. \end_layout
  2440. \end_inset
  2441. \end_layout
  2442. \begin_layout Subsection
  2443. ComBat and SVA for correction of known and unknown batch effects
  2444. \end_layout
  2445. \begin_layout Standard
  2446. In addition to well-understood effects that can be easily normalized out,
  2447. a data set often contains confounding biological effects that must be accounted
  2448. for in the modeling step.
  2449. For instance, in an experiment with pre-treatment and post-treatment samples
  2450. of cells from several different donors, donor variability represents a
  2451. known batch effect.
  2452. The most straightforward correction for known batches is to estimate the
  2453. mean for each batch independently and subtract out the differences, so
  2454. that all batches have identical means for each feature.
  2455. However, as with variance estimation, estimating the differences in batch
  2456. means is not necessarily robust at the feature level, so the ComBat method
  2457. adds empirical Bayes squeezing of the batch mean differences toward a common
  2458. value, analogous to
  2459. \begin_inset Flex Code
  2460. status open
  2461. \begin_layout Plain Layout
  2462. limma
  2463. \end_layout
  2464. \end_inset
  2465. 's empirical Bayes squeezing of feature variance estimates
  2466. \begin_inset CommandInset citation
  2467. LatexCommand cite
  2468. key "Johnson2007"
  2469. literal "false"
  2470. \end_inset
  2471. .
  2472. Effectively, ComBat assumes that modest differences between batch means
  2473. are real batch effects, but extreme differences between batch means are
  2474. more likely to be the result of outlier observations that happen to line
  2475. up with the batches rather than a genuine batch effect.
  2476. The result is a batch correction that is more robust against outliers than
  2477. simple subtraction of mean differences.
  2478. \end_layout
  2479. \begin_layout Standard
  2480. In some data sets, unknown batch effects may be present due to inherent
  2481. variability in the data, either caused by technical or biological effects.
  2482. Examples of unknown batch effects include variations in enrichment efficiency
  2483. between
  2484. \begin_inset Flex Glossary Term
  2485. status open
  2486. \begin_layout Plain Layout
  2487. ChIP-seq
  2488. \end_layout
  2489. \end_inset
  2490. samples, variations in populations of different cell types, and the effects
  2491. of uncontrolled environmental factors on gene expression in humans or live
  2492. animals.
  2493. In an ordinary linear model context, unknown batch effects cannot be inferred
  2494. and must be treated as random noise.
  2495. However, in high-throughput experiments, once again information can be
  2496. shared across features to identify patterns of un-modeled variation that
  2497. are repeated in many features.
  2498. One attractive strategy would be to perform
  2499. \begin_inset Flex Glossary Term
  2500. status open
  2501. \begin_layout Plain Layout
  2502. SVD
  2503. \end_layout
  2504. \end_inset
  2505. on the matrix of linear model residuals (which contain all the un-modeled
  2506. variation in the data) and take the first few singular vectors as batch
  2507. effects.
  2508. While this can be effective, it makes the unreasonable assumption that
  2509. all batch effects are completely uncorrelated with any of the effects being
  2510. modeled.
  2511. \begin_inset Flex Glossary Term
  2512. status open
  2513. \begin_layout Plain Layout
  2514. SVA
  2515. \end_layout
  2516. \end_inset
  2517. starts with this approach, but takes some additional steps to identify
  2518. batch effects in the full data that are both highly correlated with the
  2519. singular vectors in the residuals and least correlated with the effects
  2520. of interest
  2521. \begin_inset CommandInset citation
  2522. LatexCommand cite
  2523. key "Leek2007"
  2524. literal "false"
  2525. \end_inset
  2526. .
  2527. Since the final batch effects are estimated from the full data, moderate
  2528. correlations between the batch effects and effects of interest are allowed,
  2529. which gives
  2530. \begin_inset Flex Glossary Term
  2531. status open
  2532. \begin_layout Plain Layout
  2533. SVA
  2534. \end_layout
  2535. \end_inset
  2536. much more freedom to estimate the true extent of the batch effects compared
  2537. to simple residual
  2538. \begin_inset Flex Glossary Term
  2539. status open
  2540. \begin_layout Plain Layout
  2541. SVD
  2542. \end_layout
  2543. \end_inset
  2544. .
  2545. Once the surrogate variables are estimated, they can be included as coefficient
  2546. s in the linear model in a similar fashion to known batch effects in order
  2547. to subtract out their effects on each feature's abundance.
  2548. \end_layout
  2549. \begin_layout Subsection
  2550. Interpreting p-value distributions and estimating false discovery rates
  2551. \end_layout
  2552. \begin_layout Standard
  2553. When testing thousands of genes for differential expression or performing
  2554. thousands of statistical tests for other kinds of genomic data, the result
  2555. is thousands of p-values.
  2556. By construction, p-values have a
  2557. \begin_inset Formula $\mathrm{Uniform}(0,1)$
  2558. \end_inset
  2559. distribution under the null hypothesis.
  2560. This means that if all null hypotheses are true in a large number
  2561. \begin_inset Formula $N$
  2562. \end_inset
  2563. of tests, then for any significance threshold
  2564. \begin_inset Formula $T$
  2565. \end_inset
  2566. , approximately
  2567. \begin_inset Formula $N*T$
  2568. \end_inset
  2569. p-values would be called
  2570. \begin_inset Quotes eld
  2571. \end_inset
  2572. significant
  2573. \begin_inset Quotes erd
  2574. \end_inset
  2575. at that threshold even though the null hypotheses are all true.
  2576. These are called false discoveries.
  2577. \end_layout
  2578. \begin_layout Standard
  2579. When only a fraction of null hypotheses are true, the p-value distribution
  2580. will be a mixture of a uniform component representing the null hypotheses
  2581. that are true and a non-uniform component representing the null hypotheses
  2582. that are not true (Figure
  2583. \begin_inset CommandInset ref
  2584. LatexCommand ref
  2585. reference "fig:Example-pval-hist"
  2586. plural "false"
  2587. caps "false"
  2588. noprefix "false"
  2589. \end_inset
  2590. ).
  2591. The fraction belonging to the uniform component is referred to as
  2592. \begin_inset Formula $\pi_{0}$
  2593. \end_inset
  2594. , which ranges from 1 (all null hypotheses true) to 0 (all null hypotheses
  2595. false).
  2596. Furthermore, the non-uniform component must be biased toward zero, since
  2597. any evidence against the null hypothesis pushes the p-value for a test
  2598. toward zero.
  2599. We can exploit this fact to estimate the
  2600. \begin_inset Flex Glossary Term
  2601. status open
  2602. \begin_layout Plain Layout
  2603. FDR
  2604. \end_layout
  2605. \end_inset
  2606. for any significance threshold by estimating the degree to which the density
  2607. of p-values left of that threshold exceeds what would be expected for a
  2608. uniform distribution.
  2609. In genomics, the most commonly used
  2610. \begin_inset Flex Glossary Term
  2611. status open
  2612. \begin_layout Plain Layout
  2613. FDR
  2614. \end_layout
  2615. \end_inset
  2616. estimation method, and the one used in this work, is that of
  2617. \begin_inset ERT
  2618. status open
  2619. \begin_layout Plain Layout
  2620. \backslash
  2621. glsdisp{BH}{Benjamini and Hochberg}
  2622. \end_layout
  2623. \end_inset
  2624. \begin_inset CommandInset citation
  2625. LatexCommand cite
  2626. key "Benjamini1995"
  2627. literal "false"
  2628. \end_inset
  2629. .
  2630. This is a conservative method that effectively assumes
  2631. \begin_inset Formula $\pi_{0}=1$
  2632. \end_inset
  2633. .
  2634. Hence it gives an estimated upper bound for the
  2635. \begin_inset Flex Glossary Term
  2636. status open
  2637. \begin_layout Plain Layout
  2638. FDR
  2639. \end_layout
  2640. \end_inset
  2641. at any significance threshold, rather than a point estimate.
  2642. \end_layout
  2643. \begin_layout Standard
  2644. \begin_inset Float figure
  2645. wide false
  2646. sideways false
  2647. status collapsed
  2648. \begin_layout Plain Layout
  2649. \align center
  2650. \begin_inset Graphics
  2651. filename graphics/Intro/med-pval-hist-colored-CROP.pdf
  2652. lyxscale 50
  2653. width 100col%
  2654. groupId colfullwidth
  2655. \end_inset
  2656. \end_layout
  2657. \begin_layout Plain Layout
  2658. \begin_inset Caption Standard
  2659. \begin_layout Plain Layout
  2660. \begin_inset Argument 1
  2661. status collapsed
  2662. \begin_layout Plain Layout
  2663. Example p-value histogram.
  2664. \end_layout
  2665. \end_inset
  2666. \begin_inset CommandInset label
  2667. LatexCommand label
  2668. name "fig:Example-pval-hist"
  2669. \end_inset
  2670. \series bold
  2671. Example p-value histogram.
  2672. \series default
  2673. The distribution of p-values from a large number of independent tests (such
  2674. as differential expression tests for each gene in the genome) is a mixture
  2675. of a uniform component representing the null hypotheses that are true (blue
  2676. shading) and a zero-biased component representing the null hypotheses that
  2677. are false (red shading).
  2678. The FDR for any column in the histogram is the fraction of that column
  2679. that is blue.
  2680. The line
  2681. \begin_inset Formula $y=\pi_{0}$
  2682. \end_inset
  2683. represents the theoretical uniform component of this p-value distribution,
  2684. while the line
  2685. \begin_inset Formula $y=1$
  2686. \end_inset
  2687. represents the uniform component when all null hypotheses are true.
  2688. Note that in real data, the true status of each hypothesis is unknown,
  2689. so only the overall shape of the distribution is known.
  2690. \end_layout
  2691. \end_inset
  2692. \end_layout
  2693. \end_inset
  2694. \end_layout
  2695. \begin_layout Standard
  2696. We can also estimate
  2697. \begin_inset Formula $\pi_{0}$
  2698. \end_inset
  2699. for the entire distribution of p-values, which can give an idea of the
  2700. overall signal size in the data without setting any significance threshold
  2701. or making any decisions about which specific null hypotheses to reject.
  2702. As
  2703. \begin_inset Flex Glossary Term
  2704. status open
  2705. \begin_layout Plain Layout
  2706. FDR
  2707. \end_layout
  2708. \end_inset
  2709. estimation, there are many methods proposed for estimating
  2710. \begin_inset Formula $\pi_{0}$
  2711. \end_inset
  2712. .
  2713. The one used in this work is the Phipson method of averaging local
  2714. \begin_inset Flex Glossary Term
  2715. status open
  2716. \begin_layout Plain Layout
  2717. FDR
  2718. \end_layout
  2719. \end_inset
  2720. values
  2721. \begin_inset CommandInset citation
  2722. LatexCommand cite
  2723. key "Phipson2013Thesis"
  2724. literal "false"
  2725. \end_inset
  2726. .
  2727. Once
  2728. \begin_inset Formula $\pi_{0}$
  2729. \end_inset
  2730. is estimated, the number of null hypotheses that are false can be estimated
  2731. as
  2732. \begin_inset Formula $(1-\pi_{0})*N$
  2733. \end_inset
  2734. .
  2735. \end_layout
  2736. \begin_layout Standard
  2737. Conversely, a p-value distribution that is neither uniform nor zero-biased
  2738. is evidence of a modeling failure.
  2739. Such a distribution would imply that there is less than zero evidence against
  2740. the null hypothesis, which is not possible (in a frequentist setting).
  2741. Attempting to estimate
  2742. \begin_inset Formula $\pi_{0}$
  2743. \end_inset
  2744. from such a distribution would yield an estimate greater than 1, a nonsensical
  2745. result.
  2746. The usual cause of a poorly-behaving p-value distribution is a model assumption
  2747. that is violated by the data, such as assuming equal variance between groups
  2748. (homoskedasticity) when the variance of each group is not equal (heteroskedasti
  2749. city) or failing to model a strong confounding batch effect.
  2750. In particular, such a p-value distribution is
  2751. \emph on
  2752. not
  2753. \emph default
  2754. consistent with a simple lack of signal in the data, as this should result
  2755. in a uniform distribution.
  2756. Hence, observing such a p-value distribution should prompt a search for
  2757. violated model assumptions.
  2758. \end_layout
  2759. \begin_layout Standard
  2760. \begin_inset Note Note
  2761. status open
  2762. \begin_layout Subsection
  2763. Factor analysis: PCA, PCoA, MOFA
  2764. \end_layout
  2765. \begin_layout Plain Layout
  2766. \begin_inset Flex TODO Note (inline)
  2767. status open
  2768. \begin_layout Plain Layout
  2769. Not sure if this merits a subsection here.
  2770. \end_layout
  2771. \end_inset
  2772. \end_layout
  2773. \begin_layout Itemize
  2774. Batch-corrected
  2775. \begin_inset Flex Glossary Term
  2776. status open
  2777. \begin_layout Plain Layout
  2778. PCA
  2779. \end_layout
  2780. \end_inset
  2781. is informative, but careful application is required to avoid bias
  2782. \end_layout
  2783. \end_inset
  2784. \end_layout
  2785. \begin_layout Section
  2786. Structure of the thesis
  2787. \end_layout
  2788. \begin_layout Standard
  2789. This thesis presents 3 instances of using high-throughput genomic and epigenomic
  2790. assays to investigate hypotheses or solve problems relating to the study
  2791. of transplant rejection.
  2792. In Chapter
  2793. \begin_inset CommandInset ref
  2794. LatexCommand ref
  2795. reference "chap:CD4-ChIP-seq"
  2796. plural "false"
  2797. caps "false"
  2798. noprefix "false"
  2799. \end_inset
  2800. ,
  2801. \begin_inset Flex Glossary Term
  2802. status open
  2803. \begin_layout Plain Layout
  2804. ChIP-seq
  2805. \end_layout
  2806. \end_inset
  2807. and
  2808. \begin_inset Flex Glossary Term
  2809. status open
  2810. \begin_layout Plain Layout
  2811. RNA-seq
  2812. \end_layout
  2813. \end_inset
  2814. are used to investigate the dynamics of promoter histone methylation as
  2815. it relates to gene expression in T-cell activation and memory.
  2816. Chapter
  2817. \begin_inset CommandInset ref
  2818. LatexCommand ref
  2819. reference "chap:Improving-array-based-diagnostic"
  2820. plural "false"
  2821. caps "false"
  2822. noprefix "false"
  2823. \end_inset
  2824. looks at several array-based assays with the potential to diagnose transplant
  2825. rejection and shows that analyses of this array data are greatly improved
  2826. by paying careful attention to normalization and preprocessing.
  2827. Finally Chapter
  2828. \begin_inset CommandInset ref
  2829. LatexCommand ref
  2830. reference "chap:Globin-blocking-cyno"
  2831. plural "false"
  2832. caps "false"
  2833. noprefix "false"
  2834. \end_inset
  2835. presents a custom method for improving
  2836. \begin_inset Flex Glossary Term
  2837. status open
  2838. \begin_layout Plain Layout
  2839. RNA-seq
  2840. \end_layout
  2841. \end_inset
  2842. of non-human primate blood samples by preventing reverse transcription
  2843. of unwanted globin transcripts.
  2844. \end_layout
  2845. \begin_layout Standard
  2846. \begin_inset Flex TODO Note (inline)
  2847. status open
  2848. \begin_layout Plain Layout
  2849. Add a sentence about Ch5 once written
  2850. \end_layout
  2851. \end_inset
  2852. \end_layout
  2853. \begin_layout Chapter
  2854. \begin_inset CommandInset label
  2855. LatexCommand label
  2856. name "chap:CD4-ChIP-seq"
  2857. \end_inset
  2858. Reproducible genome-wide epigenetic analysis of H3K4 and H3K27 methylation
  2859. in naïve and memory CD4
  2860. \begin_inset Formula $^{+}$
  2861. \end_inset
  2862. T-cell activation
  2863. \end_layout
  2864. \begin_layout Standard
  2865. \size large
  2866. Ryan C.
  2867. Thompson, Sarah A.
  2868. Lamere, Daniel R.
  2869. Salomon
  2870. \end_layout
  2871. \begin_layout Standard
  2872. \begin_inset ERT
  2873. status collapsed
  2874. \begin_layout Plain Layout
  2875. \backslash
  2876. glsresetall
  2877. \end_layout
  2878. \end_inset
  2879. \begin_inset Note Note
  2880. status open
  2881. \begin_layout Plain Layout
  2882. This causes all abbreviations to be reintroduced.
  2883. \end_layout
  2884. \end_inset
  2885. \end_layout
  2886. \begin_layout Section
  2887. Introduction
  2888. \end_layout
  2889. \begin_layout Standard
  2890. CD4
  2891. \begin_inset Formula $^{+}$
  2892. \end_inset
  2893. T-cells are central to all adaptive immune responses, as well as immune
  2894. memory
  2895. \begin_inset CommandInset citation
  2896. LatexCommand cite
  2897. key "Murphy2012"
  2898. literal "false"
  2899. \end_inset
  2900. .
  2901. After an infection is cleared, a subset of the naïve CD4
  2902. \begin_inset Formula $^{+}$
  2903. \end_inset
  2904. T-cells that responded to that infection differentiate into memory CD4
  2905. \begin_inset Formula $^{+}$
  2906. \end_inset
  2907. T-cells, which are responsible for responding to the same pathogen in the
  2908. future.
  2909. Memory CD4
  2910. \begin_inset Formula $^{+}$
  2911. \end_inset
  2912. T-cells are functionally distinct, able to respond to an infection more
  2913. quickly and without the co-stimulation required by naïve CD4
  2914. \begin_inset Formula $^{+}$
  2915. \end_inset
  2916. T-cells.
  2917. However, the molecular mechanisms underlying this functional distinction
  2918. are not well-understood.
  2919. Epigenetic regulation via histone modification is thought to play an important
  2920. role, but while many studies have looked at static snapshots of histone
  2921. methylation in T-cells, few studies have looked at the dynamics of histone
  2922. regulation after T-cell activation, nor the differences in histone methylation
  2923. between naïve and memory T-cells.
  2924. H3K4me2, H3K4me3 and H3K27me3 are three histone marks thought to be major
  2925. epigenetic regulators of gene expression.
  2926. The goal of the present study is to investigate the role of these histone
  2927. marks in CD4
  2928. \begin_inset Formula $^{+}$
  2929. \end_inset
  2930. T-cell activation kinetics and memory differentiation.
  2931. In static snapshots, H3K4me2 and H3K4me3 are often observed in the promoters
  2932. of highly transcribed genes, while H3K27me3 is more often observed in promoters
  2933. of inactive genes with little to no transcription occurring.
  2934. As a result, the two H3K4 marks have been characterized as
  2935. \begin_inset Quotes eld
  2936. \end_inset
  2937. activating
  2938. \begin_inset Quotes erd
  2939. \end_inset
  2940. marks, while H3K27me3 has been characterized as
  2941. \begin_inset Quotes eld
  2942. \end_inset
  2943. deactivating
  2944. \begin_inset Quotes erd
  2945. \end_inset
  2946. .
  2947. Despite these characterizations, the actual causal relationship between
  2948. these histone modifications and gene transcription is complex and likely
  2949. involves positive and negative feedback loops between the two.
  2950. \end_layout
  2951. \begin_layout Section
  2952. Approach
  2953. \end_layout
  2954. \begin_layout Standard
  2955. In order to investigate the relationship between gene expression and these
  2956. histone modifications in the context of naïve and memory CD4
  2957. \begin_inset Formula $^{+}$
  2958. \end_inset
  2959. T-cell activation, a previously published data set of
  2960. \begin_inset Flex Glossary Term
  2961. status open
  2962. \begin_layout Plain Layout
  2963. RNA-seq
  2964. \end_layout
  2965. \end_inset
  2966. data and
  2967. \begin_inset Flex Glossary Term
  2968. status open
  2969. \begin_layout Plain Layout
  2970. ChIP-seq
  2971. \end_layout
  2972. \end_inset
  2973. data was re-analyzed using up-to-date methods designed to address the specific
  2974. analysis challenges posed by this data set.
  2975. The data set contains naïve and memory CD4
  2976. \begin_inset Formula $^{+}$
  2977. \end_inset
  2978. T-cell samples in a time course before and after activation.
  2979. Like the original analysis, this analysis looks at the dynamics of these
  2980. histone marks and compares them to gene expression dynamics at the same
  2981. time points during activation, as well as compares them between naïve and
  2982. memory cells, in hope of discovering evidence of new mechanistic details
  2983. in the interplay between them.
  2984. The original analysis of this data treated each gene promoter as a monolithic
  2985. unit and mostly assumed that
  2986. \begin_inset Flex Glossary Term
  2987. status open
  2988. \begin_layout Plain Layout
  2989. ChIP-seq
  2990. \end_layout
  2991. \end_inset
  2992. reads or peaks occurring anywhere within a promoter were equivalent, regardless
  2993. of where they occurred relative to the gene structure.
  2994. For an initial analysis of the data, this was a necessary simplifying assumptio
  2995. n.
  2996. The current analysis aims to relax this assumption, first by directly analyzing
  2997. \begin_inset Flex Glossary Term
  2998. status open
  2999. \begin_layout Plain Layout
  3000. ChIP-seq
  3001. \end_layout
  3002. \end_inset
  3003. peaks for differential modification, and second by taking a more granular
  3004. look at the
  3005. \begin_inset Flex Glossary Term
  3006. status open
  3007. \begin_layout Plain Layout
  3008. ChIP-seq
  3009. \end_layout
  3010. \end_inset
  3011. read coverage within promoter regions to ask whether the location of histone
  3012. modifications relative to the gene's
  3013. \begin_inset Flex Glossary Term
  3014. status open
  3015. \begin_layout Plain Layout
  3016. TSS
  3017. \end_layout
  3018. \end_inset
  3019. is an important factor, as opposed to simple proximity.
  3020. \end_layout
  3021. \begin_layout Section
  3022. Methods
  3023. \end_layout
  3024. \begin_layout Standard
  3025. A reproducible workflow was written to analyze the raw
  3026. \begin_inset Flex Glossary Term
  3027. status open
  3028. \begin_layout Plain Layout
  3029. ChIP-seq
  3030. \end_layout
  3031. \end_inset
  3032. and
  3033. \begin_inset Flex Glossary Term
  3034. status open
  3035. \begin_layout Plain Layout
  3036. RNA-seq
  3037. \end_layout
  3038. \end_inset
  3039. data from previous studies (
  3040. \begin_inset Flex Glossary Term
  3041. status open
  3042. \begin_layout Plain Layout
  3043. GEO
  3044. \end_layout
  3045. \end_inset
  3046. accession number
  3047. \begin_inset CommandInset href
  3048. LatexCommand href
  3049. name "GSE73214"
  3050. target "https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE73214"
  3051. literal "false"
  3052. \end_inset
  3053. )
  3054. \begin_inset CommandInset citation
  3055. LatexCommand cite
  3056. key "gh-cd4-csaw,LaMere2016,LaMere2017"
  3057. literal "true"
  3058. \end_inset
  3059. .
  3060. Briefly, this data consists of
  3061. \begin_inset Flex Glossary Term
  3062. status open
  3063. \begin_layout Plain Layout
  3064. RNA-seq
  3065. \end_layout
  3066. \end_inset
  3067. and
  3068. \begin_inset Flex Glossary Term
  3069. status open
  3070. \begin_layout Plain Layout
  3071. ChIP-seq
  3072. \end_layout
  3073. \end_inset
  3074. from CD4
  3075. \begin_inset Formula $^{+}$
  3076. \end_inset
  3077. T-cells from 4 donors.
  3078. From each donor, naïve and memory CD4
  3079. \begin_inset Formula $^{+}$
  3080. \end_inset
  3081. T-cells were isolated separately.
  3082. Then cultures of both cells were activated with CD3/CD28 beads, and samples
  3083. were taken at 4 time points: Day 0 (pre-activation), Day 1 (early activation),
  3084. Day 5 (peak activation), and Day 14 (post-activation).
  3085. For each combination of cell type and time point, RNA was isolated and
  3086. sequenced, and
  3087. \begin_inset Flex Glossary Term
  3088. status open
  3089. \begin_layout Plain Layout
  3090. ChIP-seq
  3091. \end_layout
  3092. \end_inset
  3093. was performed for each of 3 histone marks: H3K4me2, H3K4me3, and H3K27me3.
  3094. The
  3095. \begin_inset Flex Glossary Term
  3096. status open
  3097. \begin_layout Plain Layout
  3098. ChIP-seq
  3099. \end_layout
  3100. \end_inset
  3101. input DNA was also sequenced for each sample.
  3102. The result was 32 samples for each assay.
  3103. \end_layout
  3104. \begin_layout Subsection
  3105. RNA-seq differential expression analysis
  3106. \end_layout
  3107. \begin_layout Standard
  3108. \begin_inset Note Note
  3109. status collapsed
  3110. \begin_layout Plain Layout
  3111. \begin_inset Float figure
  3112. wide false
  3113. sideways false
  3114. status open
  3115. \begin_layout Plain Layout
  3116. \align center
  3117. \begin_inset Float figure
  3118. wide false
  3119. sideways false
  3120. status collapsed
  3121. \begin_layout Plain Layout
  3122. \align center
  3123. \begin_inset Graphics
  3124. filename graphics/CD4-csaw/rnaseq-compare/ensmebl-vs-entrez-star-CROP.png
  3125. lyxscale 25
  3126. width 35col%
  3127. groupId rna-comp-subfig
  3128. \end_inset
  3129. \end_layout
  3130. \begin_layout Plain Layout
  3131. \begin_inset Caption Standard
  3132. \begin_layout Plain Layout
  3133. STAR quantification, Entrez vs Ensembl gene annotation
  3134. \end_layout
  3135. \end_inset
  3136. \end_layout
  3137. \end_inset
  3138. \begin_inset space \qquad{}
  3139. \end_inset
  3140. \begin_inset Float figure
  3141. wide false
  3142. sideways false
  3143. status collapsed
  3144. \begin_layout Plain Layout
  3145. \align center
  3146. \begin_inset Graphics
  3147. filename graphics/CD4-csaw/rnaseq-compare/ensmebl-vs-entrez-shoal-CROP.png
  3148. lyxscale 25
  3149. width 35col%
  3150. groupId rna-comp-subfig
  3151. \end_inset
  3152. \end_layout
  3153. \begin_layout Plain Layout
  3154. \begin_inset Caption Standard
  3155. \begin_layout Plain Layout
  3156. Salmon+Shoal quantification, Entrez vs Ensembl gene annotation
  3157. \end_layout
  3158. \end_inset
  3159. \end_layout
  3160. \end_inset
  3161. \end_layout
  3162. \begin_layout Plain Layout
  3163. \align center
  3164. \begin_inset Float figure
  3165. wide false
  3166. sideways false
  3167. status collapsed
  3168. \begin_layout Plain Layout
  3169. \align center
  3170. \begin_inset Graphics
  3171. filename graphics/CD4-csaw/rnaseq-compare/star-vs-hisat2-CROP.png
  3172. lyxscale 25
  3173. width 35col%
  3174. groupId rna-comp-subfig
  3175. \end_inset
  3176. \end_layout
  3177. \begin_layout Plain Layout
  3178. \begin_inset Caption Standard
  3179. \begin_layout Plain Layout
  3180. STAR vs HISAT2 quantification, Ensembl gene annotation
  3181. \end_layout
  3182. \end_inset
  3183. \end_layout
  3184. \end_inset
  3185. \begin_inset space \qquad{}
  3186. \end_inset
  3187. \begin_inset Float figure
  3188. wide false
  3189. sideways false
  3190. status collapsed
  3191. \begin_layout Plain Layout
  3192. \align center
  3193. \begin_inset Graphics
  3194. filename graphics/CD4-csaw/rnaseq-compare/star-vs-salmon-CROP.png
  3195. lyxscale 25
  3196. width 35col%
  3197. groupId rna-comp-subfig
  3198. \end_inset
  3199. \end_layout
  3200. \begin_layout Plain Layout
  3201. \begin_inset Caption Standard
  3202. \begin_layout Plain Layout
  3203. Salmon vs STAR quantification, Ensembl gene annotation
  3204. \end_layout
  3205. \end_inset
  3206. \end_layout
  3207. \end_inset
  3208. \end_layout
  3209. \begin_layout Plain Layout
  3210. \align center
  3211. \begin_inset Float figure
  3212. wide false
  3213. sideways false
  3214. status collapsed
  3215. \begin_layout Plain Layout
  3216. \align center
  3217. \begin_inset Graphics
  3218. filename graphics/CD4-csaw/rnaseq-compare/salmon-vs-kallisto-CROP.png
  3219. lyxscale 25
  3220. width 35col%
  3221. groupId rna-comp-subfig
  3222. \end_inset
  3223. \end_layout
  3224. \begin_layout Plain Layout
  3225. \begin_inset Caption Standard
  3226. \begin_layout Plain Layout
  3227. Salmon vs Kallisto quantification, Ensembl gene annotation
  3228. \end_layout
  3229. \end_inset
  3230. \end_layout
  3231. \end_inset
  3232. \begin_inset space \qquad{}
  3233. \end_inset
  3234. \begin_inset Float figure
  3235. wide false
  3236. sideways false
  3237. status collapsed
  3238. \begin_layout Plain Layout
  3239. \align center
  3240. \begin_inset Graphics
  3241. filename graphics/CD4-csaw/rnaseq-compare/salmon-vs-shoal-CROP.png
  3242. lyxscale 25
  3243. width 35col%
  3244. groupId rna-comp-subfig
  3245. \end_inset
  3246. \end_layout
  3247. \begin_layout Plain Layout
  3248. \begin_inset Caption Standard
  3249. \begin_layout Plain Layout
  3250. Salmon+Shoal vs Salmon alone, Ensembl gene annotation
  3251. \end_layout
  3252. \end_inset
  3253. \end_layout
  3254. \end_inset
  3255. \end_layout
  3256. \begin_layout Plain Layout
  3257. \begin_inset Caption Standard
  3258. \begin_layout Plain Layout
  3259. \begin_inset CommandInset label
  3260. LatexCommand label
  3261. name "fig:RNA-norm-comp"
  3262. \end_inset
  3263. RNA-seq comparisons
  3264. \end_layout
  3265. \end_inset
  3266. \end_layout
  3267. \end_inset
  3268. \end_layout
  3269. \end_inset
  3270. \end_layout
  3271. \begin_layout Standard
  3272. Sequence reads were retrieved from the
  3273. \begin_inset Flex Glossary Term
  3274. status open
  3275. \begin_layout Plain Layout
  3276. SRA
  3277. \end_layout
  3278. \end_inset
  3279. \begin_inset CommandInset citation
  3280. LatexCommand cite
  3281. key "Leinonen2011"
  3282. literal "false"
  3283. \end_inset
  3284. .
  3285. Five different alignment and quantification methods were tested for the
  3286. \begin_inset Flex Glossary Term
  3287. status open
  3288. \begin_layout Plain Layout
  3289. RNA-seq
  3290. \end_layout
  3291. \end_inset
  3292. data
  3293. \begin_inset CommandInset citation
  3294. LatexCommand cite
  3295. key "Dobin2012,Kim2019,Liao2014,Pimentel2016,Patro2017,gh-shoal,gh-hg38-ref"
  3296. literal "false"
  3297. \end_inset
  3298. .
  3299. Each quantification was tested with both Ensembl transcripts and GENCODE
  3300. known gene annotations
  3301. \begin_inset CommandInset citation
  3302. LatexCommand cite
  3303. key "Zerbino2018,Harrow2012"
  3304. literal "false"
  3305. \end_inset
  3306. .
  3307. Comparisons of downstream results from each combination of quantification
  3308. method and reference revealed that all quantifications gave broadly similar
  3309. results for most genes, with non being obviously superior.
  3310. Salmon quantification with regularization by shoal with the Ensembl annotation
  3311. was chosen as the method theoretically most likely to partially mitigate
  3312. some of the batch effect in the data
  3313. \begin_inset CommandInset citation
  3314. LatexCommand cite
  3315. key "Patro2017,gh-shoal"
  3316. literal "false"
  3317. \end_inset
  3318. .
  3319. \end_layout
  3320. \begin_layout Standard
  3321. Due to an error in sample preparation, the RNA from the samples for days
  3322. 0 and 5 were sequenced using a different kit than those for days 1 and
  3323. 14.
  3324. This induced a substantial batch effect in the data due to differences
  3325. in sequencing biases between the two kits, and this batch effect is unfortunate
  3326. ly confounded with the time point variable (Figure
  3327. \begin_inset CommandInset ref
  3328. LatexCommand ref
  3329. reference "fig:RNA-PCA-no-batchsub"
  3330. plural "false"
  3331. caps "false"
  3332. noprefix "false"
  3333. \end_inset
  3334. ).
  3335. To do the best possible analysis with this data, this batch effect was
  3336. subtracted out from the data using ComBat
  3337. \begin_inset CommandInset citation
  3338. LatexCommand cite
  3339. key "Johnson2007"
  3340. literal "false"
  3341. \end_inset
  3342. , ignoring the time point variable due to the confounding with the batch
  3343. variable.
  3344. The result is a marked improvement, but the unavoidable confounding with
  3345. time point means that certain real patterns of gene expression will be
  3346. indistinguishable from the batch effect and subtracted out as a result.
  3347. Specifically, any
  3348. \begin_inset Quotes eld
  3349. \end_inset
  3350. zig-zag
  3351. \begin_inset Quotes erd
  3352. \end_inset
  3353. pattern, such as a gene whose expression goes up on day 1, down on day
  3354. 5, and back up again on day 14, will be attenuated or eliminated entirely.
  3355. In the context of a T-cell activation time course, it is unlikely that
  3356. many genes of interest will follow such an expression pattern, so this
  3357. loss was deemed an acceptable cost for correcting the batch effect.
  3358. \end_layout
  3359. \begin_layout Standard
  3360. \begin_inset Float figure
  3361. wide false
  3362. sideways false
  3363. status collapsed
  3364. \begin_layout Plain Layout
  3365. \align center
  3366. \begin_inset Float figure
  3367. wide false
  3368. sideways false
  3369. status open
  3370. \begin_layout Plain Layout
  3371. \align center
  3372. \begin_inset Graphics
  3373. filename graphics/CD4-csaw/RNA-seq/PCA-no-batchsub-CROP.png
  3374. lyxscale 25
  3375. width 75col%
  3376. groupId rna-pca-subfig
  3377. \end_inset
  3378. \end_layout
  3379. \begin_layout Plain Layout
  3380. \begin_inset Caption Standard
  3381. \begin_layout Plain Layout
  3382. \begin_inset CommandInset label
  3383. LatexCommand label
  3384. name "fig:RNA-PCA-no-batchsub"
  3385. \end_inset
  3386. Before batch correction
  3387. \end_layout
  3388. \end_inset
  3389. \end_layout
  3390. \end_inset
  3391. \end_layout
  3392. \begin_layout Plain Layout
  3393. \align center
  3394. \begin_inset Float figure
  3395. wide false
  3396. sideways false
  3397. status open
  3398. \begin_layout Plain Layout
  3399. \align center
  3400. \begin_inset Graphics
  3401. filename graphics/CD4-csaw/RNA-seq/PCA-combat-batchsub-CROP.png
  3402. lyxscale 25
  3403. width 75col%
  3404. groupId rna-pca-subfig
  3405. \end_inset
  3406. \end_layout
  3407. \begin_layout Plain Layout
  3408. \begin_inset Caption Standard
  3409. \begin_layout Plain Layout
  3410. \begin_inset CommandInset label
  3411. LatexCommand label
  3412. name "fig:RNA-PCA-ComBat-batchsub"
  3413. \end_inset
  3414. After batch correction with ComBat
  3415. \end_layout
  3416. \end_inset
  3417. \end_layout
  3418. \end_inset
  3419. \end_layout
  3420. \begin_layout Plain Layout
  3421. \begin_inset Caption Standard
  3422. \begin_layout Plain Layout
  3423. \begin_inset Argument 1
  3424. status collapsed
  3425. \begin_layout Plain Layout
  3426. PCoA plots of RNA-seq data showing effect of batch correction.
  3427. \end_layout
  3428. \end_inset
  3429. \begin_inset CommandInset label
  3430. LatexCommand label
  3431. name "fig:RNA-PCA"
  3432. \end_inset
  3433. \series bold
  3434. PCoA plots of RNA-seq data showing effect of batch correction.
  3435. \series default
  3436. The uncorrected data (a) shows a clear separation between samples from the
  3437. two batches (red and blue) dominating the first principal coordinate.
  3438. After correction with ComBat (b), the two batches now have approximately
  3439. the same center, and the first two principal coordinates both show separation
  3440. between experimental conditions rather than batches.
  3441. (Note that time points are shown in hours rather than days in these plots.)
  3442. \end_layout
  3443. \end_inset
  3444. \end_layout
  3445. \end_inset
  3446. \end_layout
  3447. \begin_layout Standard
  3448. However, removing the systematic component of the batch effect still leaves
  3449. the noise component.
  3450. The gene quantifications from the first batch are substantially noisier
  3451. than those in the second batch.
  3452. This analysis corrected for this by using
  3453. \begin_inset Flex Code
  3454. status open
  3455. \begin_layout Plain Layout
  3456. limma
  3457. \end_layout
  3458. \end_inset
  3459. 's sample weighting method to assign lower weights to the noisy samples
  3460. of batch 1 (Figure
  3461. \begin_inset CommandInset ref
  3462. LatexCommand ref
  3463. reference "fig:RNA-seq-weights-vs-covars"
  3464. plural "false"
  3465. caps "false"
  3466. noprefix "false"
  3467. \end_inset
  3468. )
  3469. \begin_inset CommandInset citation
  3470. LatexCommand cite
  3471. key "Ritchie2006,Liu2015"
  3472. literal "false"
  3473. \end_inset
  3474. .
  3475. The resulting analysis gives an accurate assessment of statistical significance
  3476. for all comparisons, which unfortunately means a loss of statistical power
  3477. for comparisons involving samples in batch 1.
  3478. \end_layout
  3479. \begin_layout Standard
  3480. In any case, the
  3481. \begin_inset Flex Glossary Term
  3482. status open
  3483. \begin_layout Plain Layout
  3484. RNA-seq
  3485. \end_layout
  3486. \end_inset
  3487. counts were first normalized using
  3488. \begin_inset Flex Glossary Term
  3489. status open
  3490. \begin_layout Plain Layout
  3491. TMM
  3492. \end_layout
  3493. \end_inset
  3494. \begin_inset CommandInset citation
  3495. LatexCommand cite
  3496. key "Robinson2010"
  3497. literal "false"
  3498. \end_inset
  3499. , converted to normalized
  3500. \begin_inset Flex Glossary Term
  3501. status open
  3502. \begin_layout Plain Layout
  3503. logCPM
  3504. \end_layout
  3505. \end_inset
  3506. with quality weights using
  3507. \begin_inset Flex Code
  3508. status open
  3509. \begin_layout Plain Layout
  3510. voomWithQualityWeights
  3511. \end_layout
  3512. \end_inset
  3513. \begin_inset CommandInset citation
  3514. LatexCommand cite
  3515. key "Law2014,Liu2015"
  3516. literal "false"
  3517. \end_inset
  3518. , and batch-corrected at this point using ComBat.
  3519. A linear model was fit to the batch-corrected, quality-weighted data for
  3520. each gene using
  3521. \begin_inset Flex Code
  3522. status open
  3523. \begin_layout Plain Layout
  3524. limma
  3525. \end_layout
  3526. \end_inset
  3527. , and each gene was tested for differential expression using
  3528. \begin_inset Flex Code
  3529. status open
  3530. \begin_layout Plain Layout
  3531. limma
  3532. \end_layout
  3533. \end_inset
  3534. 's empirical Bayes moderated
  3535. \begin_inset Formula $t$
  3536. \end_inset
  3537. -test
  3538. \begin_inset CommandInset citation
  3539. LatexCommand cite
  3540. key "Smyth2005,Law2014,Phipson2016"
  3541. literal "false"
  3542. \end_inset
  3543. .
  3544. P-values were corrected for multiple testing using the
  3545. \begin_inset Flex Glossary Term
  3546. status open
  3547. \begin_layout Plain Layout
  3548. BH
  3549. \end_layout
  3550. \end_inset
  3551. procedure for
  3552. \begin_inset Flex Glossary Term
  3553. status open
  3554. \begin_layout Plain Layout
  3555. FDR
  3556. \end_layout
  3557. \end_inset
  3558. control
  3559. \begin_inset CommandInset citation
  3560. LatexCommand cite
  3561. key "Benjamini1995"
  3562. literal "false"
  3563. \end_inset
  3564. .
  3565. \end_layout
  3566. \begin_layout Standard
  3567. \begin_inset Float figure
  3568. wide false
  3569. sideways false
  3570. status open
  3571. \begin_layout Plain Layout
  3572. \align center
  3573. \begin_inset Graphics
  3574. filename graphics/CD4-csaw/RNA-seq/weights-vs-covars-nobcv-CROP.png
  3575. lyxscale 25
  3576. width 100col%
  3577. groupId colwidth-raster
  3578. \end_inset
  3579. \end_layout
  3580. \begin_layout Plain Layout
  3581. \begin_inset Caption Standard
  3582. \begin_layout Plain Layout
  3583. \begin_inset Argument 1
  3584. status collapsed
  3585. \begin_layout Plain Layout
  3586. RNA-seq sample weights, grouped by experimental and technical covariates.
  3587. \end_layout
  3588. \end_inset
  3589. \begin_inset CommandInset label
  3590. LatexCommand label
  3591. name "fig:RNA-seq-weights-vs-covars"
  3592. \end_inset
  3593. \series bold
  3594. RNA-seq sample weights, grouped by experimental and technical covariates.
  3595. \series default
  3596. Inverse variance weights were estimated for each sample using
  3597. \begin_inset Flex Code
  3598. status open
  3599. \begin_layout Plain Layout
  3600. limma
  3601. \end_layout
  3602. \end_inset
  3603. 's
  3604. \begin_inset Flex Code
  3605. status open
  3606. \begin_layout Plain Layout
  3607. arrayWeights
  3608. \end_layout
  3609. \end_inset
  3610. function (part of
  3611. \begin_inset Flex Code
  3612. status open
  3613. \begin_layout Plain Layout
  3614. voomWithQualityWeights
  3615. \end_layout
  3616. \end_inset
  3617. ).
  3618. The samples were grouped by each known covariate and the distribution of
  3619. weights was plotted for each group.
  3620. \end_layout
  3621. \end_inset
  3622. \end_layout
  3623. \end_inset
  3624. \end_layout
  3625. \begin_layout Subsection
  3626. ChIP-seq analyses
  3627. \end_layout
  3628. \begin_layout Standard
  3629. \begin_inset Flex TODO Note (inline)
  3630. status open
  3631. \begin_layout Plain Layout
  3632. Be consistent about use of
  3633. \begin_inset Quotes eld
  3634. \end_inset
  3635. differential binding
  3636. \begin_inset Quotes erd
  3637. \end_inset
  3638. vs
  3639. \begin_inset Quotes eld
  3640. \end_inset
  3641. differential modification
  3642. \begin_inset Quotes erd
  3643. \end_inset
  3644. throughout this chapter.
  3645. The latter is usually preferred.
  3646. \end_layout
  3647. \end_inset
  3648. \end_layout
  3649. \begin_layout Standard
  3650. Sequence reads were retrieved from
  3651. \begin_inset Flex Glossary Term
  3652. status open
  3653. \begin_layout Plain Layout
  3654. SRA
  3655. \end_layout
  3656. \end_inset
  3657. \begin_inset CommandInset citation
  3658. LatexCommand cite
  3659. key "Leinonen2011"
  3660. literal "false"
  3661. \end_inset
  3662. .
  3663. \begin_inset Flex Glossary Term (Capital)
  3664. status open
  3665. \begin_layout Plain Layout
  3666. ChIP-seq
  3667. \end_layout
  3668. \end_inset
  3669. (and input) reads were aligned to the
  3670. \begin_inset Flex Glossary Term
  3671. status open
  3672. \begin_layout Plain Layout
  3673. GRCh38
  3674. \end_layout
  3675. \end_inset
  3676. genome assembly using Bowtie 2
  3677. \begin_inset CommandInset citation
  3678. LatexCommand cite
  3679. key "Langmead2012,Schneider2017,gh-hg38-ref"
  3680. literal "false"
  3681. \end_inset
  3682. .
  3683. Artifact regions were annotated using a custom implementation of the
  3684. \begin_inset Flex Code
  3685. status open
  3686. \begin_layout Plain Layout
  3687. GreyListChIP
  3688. \end_layout
  3689. \end_inset
  3690. algorithm, and these
  3691. \begin_inset Quotes eld
  3692. \end_inset
  3693. greylists
  3694. \begin_inset Quotes erd
  3695. \end_inset
  3696. were merged with the published
  3697. \begin_inset Flex Glossary Term
  3698. status open
  3699. \begin_layout Plain Layout
  3700. ENCODE
  3701. \end_layout
  3702. \end_inset
  3703. blacklists
  3704. \begin_inset CommandInset citation
  3705. LatexCommand cite
  3706. key "greylistchip,Dunham2012,Amemiya2019,gh-cd4-csaw"
  3707. literal "false"
  3708. \end_inset
  3709. .
  3710. Any read or called peak overlapping one of these regions was regarded as
  3711. artifactual and excluded from downstream analyses.
  3712. Figure
  3713. \begin_inset CommandInset ref
  3714. LatexCommand ref
  3715. reference "fig:CCF-master"
  3716. plural "false"
  3717. caps "false"
  3718. noprefix "false"
  3719. \end_inset
  3720. shows the improvement after blacklisting in the strand cross-correlation
  3721. plots, a common quality control plot for
  3722. \begin_inset Flex Glossary Term
  3723. status open
  3724. \begin_layout Plain Layout
  3725. ChIP-seq
  3726. \end_layout
  3727. \end_inset
  3728. data
  3729. \begin_inset CommandInset citation
  3730. LatexCommand cite
  3731. key "Kharchenko2008,Lun2015a"
  3732. literal "false"
  3733. \end_inset
  3734. .
  3735. Peaks were called using
  3736. \begin_inset Flex Code
  3737. status open
  3738. \begin_layout Plain Layout
  3739. epic
  3740. \end_layout
  3741. \end_inset
  3742. , an implementation of the
  3743. \begin_inset Flex Glossary Term
  3744. status open
  3745. \begin_layout Plain Layout
  3746. SICER
  3747. \end_layout
  3748. \end_inset
  3749. algorithm
  3750. \begin_inset CommandInset citation
  3751. LatexCommand cite
  3752. key "Zang2009,gh-epic"
  3753. literal "false"
  3754. \end_inset
  3755. .
  3756. Peaks were also called separately using
  3757. \begin_inset Flex Glossary Term
  3758. status open
  3759. \begin_layout Plain Layout
  3760. MACS
  3761. \end_layout
  3762. \end_inset
  3763. , but
  3764. \begin_inset Flex Glossary Term
  3765. status open
  3766. \begin_layout Plain Layout
  3767. MACS
  3768. \end_layout
  3769. \end_inset
  3770. was determined to be a poor fit for the data, and these peak calls are
  3771. not used in any further analyses
  3772. \begin_inset CommandInset citation
  3773. LatexCommand cite
  3774. key "Zhang2008"
  3775. literal "false"
  3776. \end_inset
  3777. .
  3778. Consensus peaks were determined by applying the
  3779. \begin_inset Flex Glossary Term
  3780. status open
  3781. \begin_layout Plain Layout
  3782. IDR
  3783. \end_layout
  3784. \end_inset
  3785. framework
  3786. \begin_inset CommandInset citation
  3787. LatexCommand cite
  3788. key "Li2006,gh-idr"
  3789. literal "false"
  3790. \end_inset
  3791. to find peaks consistently called in the same locations across all 4 donors.
  3792. \end_layout
  3793. \begin_layout Standard
  3794. \begin_inset ERT
  3795. status open
  3796. \begin_layout Plain Layout
  3797. \backslash
  3798. afterpage{
  3799. \end_layout
  3800. \begin_layout Plain Layout
  3801. \backslash
  3802. begin{landscape}
  3803. \end_layout
  3804. \end_inset
  3805. \end_layout
  3806. \begin_layout Standard
  3807. \begin_inset Float figure
  3808. wide false
  3809. sideways false
  3810. status open
  3811. \begin_layout Plain Layout
  3812. \align center
  3813. \begin_inset Float figure
  3814. wide false
  3815. sideways false
  3816. status open
  3817. \begin_layout Plain Layout
  3818. \align center
  3819. \begin_inset Graphics
  3820. filename graphics/CD4-csaw/csaw/CCF-plots-noBL-PAGE2-CROP.pdf
  3821. lyxscale 75
  3822. width 47col%
  3823. groupId ccf-subfig
  3824. \end_inset
  3825. \end_layout
  3826. \begin_layout Plain Layout
  3827. \begin_inset Caption Standard
  3828. \begin_layout Plain Layout
  3829. \series bold
  3830. \begin_inset CommandInset label
  3831. LatexCommand label
  3832. name "fig:CCF-without-blacklist"
  3833. \end_inset
  3834. Cross-correlation plots without removing blacklisted reads.
  3835. \series default
  3836. Without blacklisting, many artifactual peaks are visible in the cross-correlatio
  3837. ns of the ChIP-seq samples, and the peak at the true fragment size (147
  3838. \begin_inset space ~
  3839. \end_inset
  3840. bp) is frequently overshadowed by the artifactual peak at the read length
  3841. (100
  3842. \begin_inset space ~
  3843. \end_inset
  3844. bp).
  3845. \end_layout
  3846. \end_inset
  3847. \end_layout
  3848. \end_inset
  3849. \begin_inset space \hfill{}
  3850. \end_inset
  3851. \begin_inset Float figure
  3852. wide false
  3853. sideways false
  3854. status collapsed
  3855. \begin_layout Plain Layout
  3856. \align center
  3857. \begin_inset Graphics
  3858. filename graphics/CD4-csaw/csaw/CCF-plots-PAGE2-CROP.pdf
  3859. lyxscale 75
  3860. width 47col%
  3861. groupId ccf-subfig
  3862. \end_inset
  3863. \end_layout
  3864. \begin_layout Plain Layout
  3865. \begin_inset Caption Standard
  3866. \begin_layout Plain Layout
  3867. \series bold
  3868. \begin_inset CommandInset label
  3869. LatexCommand label
  3870. name "fig:CCF-with-blacklist"
  3871. \end_inset
  3872. Cross-correlation plots with blacklisted reads removed.
  3873. \series default
  3874. After blacklisting, most ChIP-seq samples have clean-looking periodic cross-cor
  3875. relation plots, with the largest peak around 147
  3876. \begin_inset space ~
  3877. \end_inset
  3878. bp, the expected size for a fragment of DNA from a single nucleosome, and
  3879. little to no peak at the read length, 100
  3880. \begin_inset space ~
  3881. \end_inset
  3882. bp.
  3883. \end_layout
  3884. \end_inset
  3885. \end_layout
  3886. \end_inset
  3887. \end_layout
  3888. \begin_layout Plain Layout
  3889. \begin_inset Flex TODO Note (inline)
  3890. status open
  3891. \begin_layout Plain Layout
  3892. Figure font too small
  3893. \end_layout
  3894. \end_inset
  3895. \end_layout
  3896. \begin_layout Plain Layout
  3897. \begin_inset Caption Standard
  3898. \begin_layout Plain Layout
  3899. \begin_inset Argument 1
  3900. status collapsed
  3901. \begin_layout Plain Layout
  3902. Strand cross-correlation plots for ChIP-seq data, before and after blacklisting.
  3903. \end_layout
  3904. \end_inset
  3905. \begin_inset CommandInset label
  3906. LatexCommand label
  3907. name "fig:CCF-master"
  3908. \end_inset
  3909. \series bold
  3910. Strand cross-correlation plots for ChIP-seq data, before and after blacklisting.
  3911. \series default
  3912. The number of reads starting at each position in the genome was counted
  3913. separately for the plus and minus strands, and then the correlation coefficient
  3914. between the read start counts for both strands (cross-correlation) was
  3915. computed after shifting the plus strand counts forward by a specified interval
  3916. (the delay).
  3917. This was repeated for every delay value from 0 to 1000, and the cross-correlati
  3918. on values were plotted as a function of the delay.
  3919. In good quality samples, cross-correlation is maximized when the delay
  3920. equals the fragment size; in poor quality samples, cross-correlation is
  3921. often maximized when the delay equals the read length, an artifactual peak
  3922. whose cause is not fully understood.
  3923. \end_layout
  3924. \end_inset
  3925. \end_layout
  3926. \end_inset
  3927. \end_layout
  3928. \begin_layout Standard
  3929. \begin_inset ERT
  3930. status open
  3931. \begin_layout Plain Layout
  3932. \backslash
  3933. end{landscape}
  3934. \end_layout
  3935. \begin_layout Plain Layout
  3936. }
  3937. \end_layout
  3938. \end_inset
  3939. \end_layout
  3940. \begin_layout Standard
  3941. Promoters were defined by computing the distance from each annotated
  3942. \begin_inset Flex Glossary Term
  3943. status open
  3944. \begin_layout Plain Layout
  3945. TSS
  3946. \end_layout
  3947. \end_inset
  3948. to the nearest called peak and examining the distribution of distances,
  3949. observing that peaks for each histone mark were enriched within a certain
  3950. distance of the
  3951. \begin_inset Flex Glossary Term
  3952. status open
  3953. \begin_layout Plain Layout
  3954. TSS
  3955. \end_layout
  3956. \end_inset
  3957. .
  3958. (Note: this analysis was performed using the original peak calls and expression
  3959. values from
  3960. \begin_inset Flex Glossary Term
  3961. status open
  3962. \begin_layout Plain Layout
  3963. GEO
  3964. \end_layout
  3965. \end_inset
  3966. \begin_inset CommandInset citation
  3967. LatexCommand cite
  3968. key "LaMere2016"
  3969. literal "false"
  3970. \end_inset
  3971. .) For H3K4me2 and H3K4me3, this distance was about 1
  3972. \begin_inset space ~
  3973. \end_inset
  3974. kbp, while for H3K27me3 it was 2.5
  3975. \begin_inset space ~
  3976. \end_inset
  3977. kbp.
  3978. These distances were used as an
  3979. \begin_inset Quotes eld
  3980. \end_inset
  3981. effective promoter radius
  3982. \begin_inset Quotes erd
  3983. \end_inset
  3984. for each mark.
  3985. The promoter region for each gene was defined as the region of the genome
  3986. within this distance upstream or downstream of the gene's annotated
  3987. \begin_inset Flex Glossary Term
  3988. status open
  3989. \begin_layout Plain Layout
  3990. TSS
  3991. \end_layout
  3992. \end_inset
  3993. .
  3994. For genes with multiple annotated
  3995. \begin_inset Flex Glossary Term (pl)
  3996. status open
  3997. \begin_layout Plain Layout
  3998. TSS
  3999. \end_layout
  4000. \end_inset
  4001. , a promoter region was defined for each
  4002. \begin_inset Flex Glossary Term
  4003. status open
  4004. \begin_layout Plain Layout
  4005. TSS
  4006. \end_layout
  4007. \end_inset
  4008. individually, and any promoters that overlapped (due to multiple
  4009. \begin_inset Flex Glossary Term (pl)
  4010. status open
  4011. \begin_layout Plain Layout
  4012. TSS
  4013. \end_layout
  4014. \end_inset
  4015. being closer than 2 times the radius) were merged into one large promoter.
  4016. Thus, some genes had multiple promoters defined, which were each analyzed
  4017. separately for differential modification.
  4018. \end_layout
  4019. \begin_layout Standard
  4020. Reads in promoters, peaks, and sliding windows across the genome were counted
  4021. and normalized using
  4022. \begin_inset Flex Code
  4023. status open
  4024. \begin_layout Plain Layout
  4025. csaw
  4026. \end_layout
  4027. \end_inset
  4028. and analyzed for differential modification using
  4029. \begin_inset Flex Code
  4030. status open
  4031. \begin_layout Plain Layout
  4032. edgeR
  4033. \end_layout
  4034. \end_inset
  4035. \begin_inset CommandInset citation
  4036. LatexCommand cite
  4037. key "Lun2014,Lun2015a,Lund2012,Phipson2016"
  4038. literal "false"
  4039. \end_inset
  4040. .
  4041. Unobserved confounding factors in the
  4042. \begin_inset Flex Glossary Term
  4043. status open
  4044. \begin_layout Plain Layout
  4045. ChIP-seq
  4046. \end_layout
  4047. \end_inset
  4048. data were corrected using
  4049. \begin_inset Flex Glossary Term
  4050. status open
  4051. \begin_layout Plain Layout
  4052. SVA
  4053. \end_layout
  4054. \end_inset
  4055. \begin_inset CommandInset citation
  4056. LatexCommand cite
  4057. key "Leek2007,Leek2014"
  4058. literal "false"
  4059. \end_inset
  4060. .
  4061. Principal coordinate plots of the promoter count data for each histone
  4062. mark before and after subtracting surrogate variable effects are shown
  4063. in Figure
  4064. \begin_inset CommandInset ref
  4065. LatexCommand ref
  4066. reference "fig:PCoA-ChIP"
  4067. plural "false"
  4068. caps "false"
  4069. noprefix "false"
  4070. \end_inset
  4071. .
  4072. \end_layout
  4073. \begin_layout Standard
  4074. \begin_inset Float figure
  4075. wide false
  4076. sideways false
  4077. status collapsed
  4078. \begin_layout Plain Layout
  4079. \begin_inset Float figure
  4080. wide false
  4081. sideways false
  4082. status open
  4083. \begin_layout Plain Layout
  4084. \align center
  4085. \begin_inset Graphics
  4086. filename graphics/CD4-csaw/ChIP-seq/H3K4me2-PCA-raw-CROP.png
  4087. lyxscale 25
  4088. width 45col%
  4089. groupId pcoa-subfig
  4090. \end_inset
  4091. \end_layout
  4092. \begin_layout Plain Layout
  4093. \begin_inset Caption Standard
  4094. \begin_layout Plain Layout
  4095. \series bold
  4096. \begin_inset CommandInset label
  4097. LatexCommand label
  4098. name "fig:PCoA-H3K4me2-bad"
  4099. \end_inset
  4100. H3K4me2, no correction
  4101. \end_layout
  4102. \end_inset
  4103. \end_layout
  4104. \end_inset
  4105. \begin_inset space \hfill{}
  4106. \end_inset
  4107. \begin_inset Float figure
  4108. wide false
  4109. sideways false
  4110. status open
  4111. \begin_layout Plain Layout
  4112. \align center
  4113. \begin_inset Graphics
  4114. filename graphics/CD4-csaw/ChIP-seq/H3K4me2-PCA-SVsub-CROP.png
  4115. lyxscale 25
  4116. width 45col%
  4117. groupId pcoa-subfig
  4118. \end_inset
  4119. \end_layout
  4120. \begin_layout Plain Layout
  4121. \begin_inset Caption Standard
  4122. \begin_layout Plain Layout
  4123. \series bold
  4124. \begin_inset CommandInset label
  4125. LatexCommand label
  4126. name "fig:PCoA-H3K4me2-good"
  4127. \end_inset
  4128. H3K4me2, SVs subtracted
  4129. \end_layout
  4130. \end_inset
  4131. \end_layout
  4132. \end_inset
  4133. \end_layout
  4134. \begin_layout Plain Layout
  4135. \begin_inset Float figure
  4136. wide false
  4137. sideways false
  4138. status collapsed
  4139. \begin_layout Plain Layout
  4140. \align center
  4141. \begin_inset Graphics
  4142. filename graphics/CD4-csaw/ChIP-seq/H3K4me3-PCA-raw-CROP.png
  4143. lyxscale 25
  4144. width 45col%
  4145. groupId pcoa-subfig
  4146. \end_inset
  4147. \end_layout
  4148. \begin_layout Plain Layout
  4149. \begin_inset Caption Standard
  4150. \begin_layout Plain Layout
  4151. \series bold
  4152. \begin_inset CommandInset label
  4153. LatexCommand label
  4154. name "fig:PCoA-H3K4me3-bad"
  4155. \end_inset
  4156. H3K4me3, no correction
  4157. \end_layout
  4158. \end_inset
  4159. \end_layout
  4160. \end_inset
  4161. \begin_inset space \hfill{}
  4162. \end_inset
  4163. \begin_inset Float figure
  4164. wide false
  4165. sideways false
  4166. status collapsed
  4167. \begin_layout Plain Layout
  4168. \align center
  4169. \begin_inset Graphics
  4170. filename graphics/CD4-csaw/ChIP-seq/H3K4me3-PCA-SVsub-CROP.png
  4171. lyxscale 25
  4172. width 45col%
  4173. groupId pcoa-subfig
  4174. \end_inset
  4175. \end_layout
  4176. \begin_layout Plain Layout
  4177. \begin_inset Caption Standard
  4178. \begin_layout Plain Layout
  4179. \series bold
  4180. \begin_inset CommandInset label
  4181. LatexCommand label
  4182. name "fig:PCoA-H3K4me3-good"
  4183. \end_inset
  4184. H3K4me3, SVs subtracted
  4185. \end_layout
  4186. \end_inset
  4187. \end_layout
  4188. \end_inset
  4189. \end_layout
  4190. \begin_layout Plain Layout
  4191. \begin_inset Float figure
  4192. wide false
  4193. sideways false
  4194. status collapsed
  4195. \begin_layout Plain Layout
  4196. \align center
  4197. \begin_inset Graphics
  4198. filename graphics/CD4-csaw/ChIP-seq/H3K27me3-PCA-raw-CROP.png
  4199. lyxscale 25
  4200. width 45col%
  4201. groupId pcoa-subfig
  4202. \end_inset
  4203. \end_layout
  4204. \begin_layout Plain Layout
  4205. \begin_inset Caption Standard
  4206. \begin_layout Plain Layout
  4207. \series bold
  4208. \begin_inset CommandInset label
  4209. LatexCommand label
  4210. name "fig:PCoA-H3K27me3-bad"
  4211. \end_inset
  4212. H3K27me3, no correction
  4213. \end_layout
  4214. \end_inset
  4215. \end_layout
  4216. \end_inset
  4217. \begin_inset space \hfill{}
  4218. \end_inset
  4219. \begin_inset Float figure
  4220. wide false
  4221. sideways false
  4222. status collapsed
  4223. \begin_layout Plain Layout
  4224. \align center
  4225. \begin_inset Graphics
  4226. filename graphics/CD4-csaw/ChIP-seq/H3K27me3-PCA-SVsub-CROP.png
  4227. lyxscale 25
  4228. width 45col%
  4229. groupId pcoa-subfig
  4230. \end_inset
  4231. \end_layout
  4232. \begin_layout Plain Layout
  4233. \begin_inset Caption Standard
  4234. \begin_layout Plain Layout
  4235. \series bold
  4236. \begin_inset CommandInset label
  4237. LatexCommand label
  4238. name "fig:PCoA-H3K27me3-good"
  4239. \end_inset
  4240. H3K27me3, SVs subtracted
  4241. \end_layout
  4242. \end_inset
  4243. \end_layout
  4244. \end_inset
  4245. \end_layout
  4246. \begin_layout Plain Layout
  4247. \begin_inset Flex TODO Note (inline)
  4248. status collapsed
  4249. \begin_layout Plain Layout
  4250. Figure font too small
  4251. \end_layout
  4252. \end_inset
  4253. \end_layout
  4254. \begin_layout Plain Layout
  4255. \begin_inset Caption Standard
  4256. \begin_layout Plain Layout
  4257. \begin_inset Argument 1
  4258. status collapsed
  4259. \begin_layout Plain Layout
  4260. PCoA plots of ChIP-seq sliding window data, before and after subtracting
  4261. surrogate variables.
  4262. \end_layout
  4263. \end_inset
  4264. \begin_inset CommandInset label
  4265. LatexCommand label
  4266. name "fig:PCoA-ChIP"
  4267. \end_inset
  4268. \series bold
  4269. PCoA plots of ChIP-seq sliding window data, before and after subtracting
  4270. surrogate variables (SVs).
  4271. \series default
  4272. For each histone mark, a PCoA plot of the first 2 principal coordinates
  4273. was created before and after subtraction of SV effects.
  4274. Time points are shown by color and cell type by shape, and samples from
  4275. the same time point and cell type are enclosed in a shaded area to aid
  4276. in visial recognition (this shaded area has no meaning on the plot).
  4277. Samples of the same cell type from the same donor are connected with a
  4278. line in time point order, showing the
  4279. \begin_inset Quotes eld
  4280. \end_inset
  4281. trajectory
  4282. \begin_inset Quotes erd
  4283. \end_inset
  4284. of each donor's samples over time.
  4285. \end_layout
  4286. \end_inset
  4287. \end_layout
  4288. \end_inset
  4289. \end_layout
  4290. \begin_layout Standard
  4291. To investigate whether the location of a peak within the promoter region
  4292. was important,
  4293. \begin_inset Quotes eld
  4294. \end_inset
  4295. relative coverage profiles
  4296. \begin_inset Quotes erd
  4297. \end_inset
  4298. were generated.
  4299. First, 500-bp sliding windows were tiled around each annotated
  4300. \begin_inset Flex Glossary Term
  4301. status open
  4302. \begin_layout Plain Layout
  4303. TSS
  4304. \end_layout
  4305. \end_inset
  4306. : one window centered on the
  4307. \begin_inset Flex Glossary Term
  4308. status open
  4309. \begin_layout Plain Layout
  4310. TSS
  4311. \end_layout
  4312. \end_inset
  4313. itself, and 10 windows each upstream and downstream, thus covering a 10.5-kb
  4314. region centered on the
  4315. \begin_inset Flex Glossary Term
  4316. status open
  4317. \begin_layout Plain Layout
  4318. TSS
  4319. \end_layout
  4320. \end_inset
  4321. with 21 windows.
  4322. Reads in each window for each
  4323. \begin_inset Flex Glossary Term
  4324. status open
  4325. \begin_layout Plain Layout
  4326. TSS
  4327. \end_layout
  4328. \end_inset
  4329. were counted in each sample, and the counts were normalized and converted
  4330. to
  4331. \begin_inset Flex Glossary Term
  4332. status open
  4333. \begin_layout Plain Layout
  4334. logCPM
  4335. \end_layout
  4336. \end_inset
  4337. as in the differential modification analysis.
  4338. Then, the
  4339. \begin_inset Flex Glossary Term
  4340. status open
  4341. \begin_layout Plain Layout
  4342. logCPM
  4343. \end_layout
  4344. \end_inset
  4345. values within each promoter were normalized to an average of zero, such
  4346. that each window's normalized abundance now represents the relative read
  4347. depth of that window compared to all other windows in the same promoter.
  4348. The normalized abundance values for each window in a promoter are collectively
  4349. referred to as that promoter's
  4350. \begin_inset Quotes eld
  4351. \end_inset
  4352. relative coverage profile
  4353. \begin_inset Quotes erd
  4354. \end_inset
  4355. .
  4356. \end_layout
  4357. \begin_layout Subsection
  4358. MOFA analysis of cross-dataset variation patterns
  4359. \end_layout
  4360. \begin_layout Standard
  4361. \begin_inset Flex Glossary Term
  4362. status open
  4363. \begin_layout Plain Layout
  4364. MOFA
  4365. \end_layout
  4366. \end_inset
  4367. was run on all the
  4368. \begin_inset Flex Glossary Term
  4369. status open
  4370. \begin_layout Plain Layout
  4371. ChIP-seq
  4372. \end_layout
  4373. \end_inset
  4374. windows overlapping consensus peaks for each histone mark, as well as the
  4375. \begin_inset Flex Glossary Term
  4376. status open
  4377. \begin_layout Plain Layout
  4378. RNA-seq
  4379. \end_layout
  4380. \end_inset
  4381. data, in order to identify patterns of coordinated variation across all
  4382. data sets
  4383. \begin_inset CommandInset citation
  4384. LatexCommand cite
  4385. key "Argelaguet2018"
  4386. literal "false"
  4387. \end_inset
  4388. .
  4389. The results are summarized in Figure
  4390. \begin_inset CommandInset ref
  4391. LatexCommand ref
  4392. reference "fig:MOFA-master"
  4393. plural "false"
  4394. caps "false"
  4395. noprefix "false"
  4396. \end_inset
  4397. .
  4398. \begin_inset Flex Glossary Term (Capital, pl)
  4399. status open
  4400. \begin_layout Plain Layout
  4401. LF
  4402. \end_layout
  4403. \end_inset
  4404. 1, 4, and 5 were determined to explain the most variation consistently
  4405. across all data sets (Figure
  4406. \begin_inset CommandInset ref
  4407. LatexCommand ref
  4408. reference "fig:mofa-varexplained"
  4409. plural "false"
  4410. caps "false"
  4411. noprefix "false"
  4412. \end_inset
  4413. ), and scatter plots of these factors show that they also correlate best
  4414. with the experimental factors (Figure
  4415. \begin_inset CommandInset ref
  4416. LatexCommand ref
  4417. reference "fig:mofa-lf-scatter"
  4418. plural "false"
  4419. caps "false"
  4420. noprefix "false"
  4421. \end_inset
  4422. ).
  4423. \begin_inset Flex Glossary Term
  4424. status open
  4425. \begin_layout Plain Layout
  4426. LF
  4427. \end_layout
  4428. \end_inset
  4429. 2 captures the batch effect in the
  4430. \begin_inset Flex Glossary Term
  4431. status open
  4432. \begin_layout Plain Layout
  4433. RNA-seq
  4434. \end_layout
  4435. \end_inset
  4436. data.
  4437. Removing the effect of
  4438. \begin_inset Flex Glossary Term
  4439. status open
  4440. \begin_layout Plain Layout
  4441. LF
  4442. \end_layout
  4443. \end_inset
  4444. 2 using
  4445. \begin_inset Flex Glossary Term
  4446. status open
  4447. \begin_layout Plain Layout
  4448. MOFA
  4449. \end_layout
  4450. \end_inset
  4451. theoretically yields a batch correction that does not depend on knowing
  4452. the experimental factors.
  4453. When this was attempted, the resulting batch correction was comparable
  4454. to ComBat (see Figure
  4455. \begin_inset CommandInset ref
  4456. LatexCommand ref
  4457. reference "fig:RNA-PCA-ComBat-batchsub"
  4458. plural "false"
  4459. caps "false"
  4460. noprefix "false"
  4461. \end_inset
  4462. ), indicating that the ComBat-based batch correction has little room for
  4463. improvement given the problems with the data set.
  4464. \end_layout
  4465. \begin_layout Standard
  4466. \begin_inset ERT
  4467. status open
  4468. \begin_layout Plain Layout
  4469. \backslash
  4470. afterpage{
  4471. \end_layout
  4472. \begin_layout Plain Layout
  4473. \backslash
  4474. begin{landscape}
  4475. \end_layout
  4476. \end_inset
  4477. \end_layout
  4478. \begin_layout Standard
  4479. \begin_inset Float figure
  4480. wide false
  4481. sideways false
  4482. status open
  4483. \begin_layout Plain Layout
  4484. \begin_inset Float figure
  4485. wide false
  4486. sideways false
  4487. status collapsed
  4488. \begin_layout Plain Layout
  4489. \align center
  4490. \begin_inset Graphics
  4491. filename graphics/CD4-csaw/MOFA-varExplaiend-matrix-CROP.png
  4492. lyxscale 25
  4493. width 45col%
  4494. groupId mofa-subfig
  4495. \end_inset
  4496. \end_layout
  4497. \begin_layout Plain Layout
  4498. \begin_inset Caption Standard
  4499. \begin_layout Plain Layout
  4500. \series bold
  4501. \begin_inset CommandInset label
  4502. LatexCommand label
  4503. name "fig:mofa-varexplained"
  4504. \end_inset
  4505. Variance explained in each data set by each latent factor estimated by MOFA.
  4506. \series default
  4507. For each LF learned by MOFA, the variance explained by that factor in each
  4508. data set (
  4509. \begin_inset Quotes eld
  4510. \end_inset
  4511. view
  4512. \begin_inset Quotes erd
  4513. \end_inset
  4514. ) is shown by the shading of the cells in the lower section.
  4515. The upper section shows the total fraction of each data set's variance
  4516. that is explained by all LFs combined.
  4517. \end_layout
  4518. \end_inset
  4519. \end_layout
  4520. \end_inset
  4521. \begin_inset space \hfill{}
  4522. \end_inset
  4523. \begin_inset Float figure
  4524. wide false
  4525. sideways false
  4526. status collapsed
  4527. \begin_layout Plain Layout
  4528. \align center
  4529. \begin_inset Graphics
  4530. filename graphics/CD4-csaw/MOFA-LF-scatter-small.png
  4531. lyxscale 25
  4532. width 45col%
  4533. groupId mofa-subfig
  4534. \end_inset
  4535. \end_layout
  4536. \begin_layout Plain Layout
  4537. \begin_inset Caption Standard
  4538. \begin_layout Plain Layout
  4539. \series bold
  4540. \begin_inset CommandInset label
  4541. LatexCommand label
  4542. name "fig:mofa-lf-scatter"
  4543. \end_inset
  4544. Scatter plots of specific pairs of MOFA latent factors.
  4545. \series default
  4546. LFs 1, 4, and 5 explain substantial variation in all data sets, so they
  4547. were plotted against each other in order to reveal patterns of variation
  4548. that are shared across all data sets.
  4549. These plots can be interpreted similarly to PCA and PCoA plots.
  4550. \end_layout
  4551. \end_inset
  4552. \end_layout
  4553. \end_inset
  4554. \end_layout
  4555. \begin_layout Plain Layout
  4556. \begin_inset Flex TODO Note (inline)
  4557. status open
  4558. \begin_layout Plain Layout
  4559. Figure font a bit too small
  4560. \end_layout
  4561. \end_inset
  4562. \end_layout
  4563. \begin_layout Plain Layout
  4564. \begin_inset Caption Standard
  4565. \begin_layout Plain Layout
  4566. \begin_inset Argument 1
  4567. status collapsed
  4568. \begin_layout Plain Layout
  4569. MOFA latent factors identify shared patterns of variation.
  4570. \end_layout
  4571. \end_inset
  4572. \begin_inset CommandInset label
  4573. LatexCommand label
  4574. name "fig:MOFA-master"
  4575. \end_inset
  4576. \series bold
  4577. MOFA latent factors identify shared patterns of variation.
  4578. \series default
  4579. MOFA was used to estimate latent factors (LFs) that explain substantial
  4580. variation in the RNA-seq data and the ChIP-seq data (a).
  4581. Then specific LFs of interest were selected and plotted (b).
  4582. \end_layout
  4583. \end_inset
  4584. \end_layout
  4585. \end_inset
  4586. \end_layout
  4587. \begin_layout Standard
  4588. \begin_inset ERT
  4589. status open
  4590. \begin_layout Plain Layout
  4591. \backslash
  4592. end{landscape}
  4593. \end_layout
  4594. \begin_layout Plain Layout
  4595. }
  4596. \end_layout
  4597. \end_inset
  4598. \end_layout
  4599. \begin_layout Standard
  4600. \begin_inset Note Note
  4601. status collapsed
  4602. \begin_layout Plain Layout
  4603. \begin_inset Float figure
  4604. wide false
  4605. sideways false
  4606. status open
  4607. \begin_layout Plain Layout
  4608. \align center
  4609. \begin_inset Graphics
  4610. filename graphics/CD4-csaw/MOFA-batch-correct-CROP.png
  4611. lyxscale 25
  4612. width 100col%
  4613. groupId colwidth-raster
  4614. \end_inset
  4615. \end_layout
  4616. \begin_layout Plain Layout
  4617. \begin_inset Caption Standard
  4618. \begin_layout Plain Layout
  4619. \series bold
  4620. \begin_inset CommandInset label
  4621. LatexCommand label
  4622. name "fig:mofa-batchsub"
  4623. \end_inset
  4624. Result of RNA-seq batch-correction using MOFA latent factors
  4625. \end_layout
  4626. \end_inset
  4627. \end_layout
  4628. \end_inset
  4629. \end_layout
  4630. \end_inset
  4631. \end_layout
  4632. \begin_layout Section
  4633. Results
  4634. \end_layout
  4635. \begin_layout Standard
  4636. \begin_inset Flex TODO Note (inline)
  4637. status open
  4638. \begin_layout Plain Layout
  4639. Focus on what hypotheses were tested, then select figures that show how
  4640. those hypotheses were tested, even if the result is a negative.
  4641. Not every interesting result needs to be in here.
  4642. Chapter should tell a story.
  4643. \end_layout
  4644. \end_inset
  4645. \end_layout
  4646. \begin_layout Subsection
  4647. Interpretation of RNA-seq analysis is limited by a major confounding factor
  4648. \end_layout
  4649. \begin_layout Standard
  4650. Genes called as present in the
  4651. \begin_inset Flex Glossary Term
  4652. status open
  4653. \begin_layout Plain Layout
  4654. RNA-seq
  4655. \end_layout
  4656. \end_inset
  4657. data were tested for differential expression between all time points and
  4658. cell types.
  4659. The counts of differentially expressed genes are shown in Table
  4660. \begin_inset CommandInset ref
  4661. LatexCommand ref
  4662. reference "tab:Estimated-and-detected-rnaseq"
  4663. plural "false"
  4664. caps "false"
  4665. noprefix "false"
  4666. \end_inset
  4667. .
  4668. Notably, all the results for Day 0 and Day 5 have substantially fewer genes
  4669. called differentially expressed than any of the results for other time
  4670. points.
  4671. This is an unfortunate result of the difference in sample quality between
  4672. the two batches of
  4673. \begin_inset Flex Glossary Term
  4674. status open
  4675. \begin_layout Plain Layout
  4676. RNA-seq
  4677. \end_layout
  4678. \end_inset
  4679. data.
  4680. All the samples in Batch 1, which includes all the samples from Days 0
  4681. and 5, have substantially more variability than the samples in Batch 2,
  4682. which includes the other time points.
  4683. This is reflected in the substantially higher weights assigned to Batch
  4684. 2 (Figure
  4685. \begin_inset CommandInset ref
  4686. LatexCommand ref
  4687. reference "fig:RNA-seq-weights-vs-covars"
  4688. plural "false"
  4689. caps "false"
  4690. noprefix "false"
  4691. \end_inset
  4692. ).
  4693. \begin_inset Float table
  4694. wide false
  4695. sideways false
  4696. status collapsed
  4697. \begin_layout Plain Layout
  4698. \align center
  4699. \begin_inset Tabular
  4700. <lyxtabular version="3" rows="11" columns="3">
  4701. <features tabularvalignment="middle">
  4702. <column alignment="center" valignment="top">
  4703. <column alignment="center" valignment="top">
  4704. <column alignment="center" valignment="top">
  4705. <row>
  4706. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  4707. \begin_inset Text
  4708. \begin_layout Plain Layout
  4709. Test
  4710. \end_layout
  4711. \end_inset
  4712. </cell>
  4713. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  4714. \begin_inset Text
  4715. \begin_layout Plain Layout
  4716. Est.
  4717. non-null
  4718. \end_layout
  4719. \end_inset
  4720. </cell>
  4721. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  4722. \begin_inset Text
  4723. \begin_layout Plain Layout
  4724. \begin_inset Formula $\mathrm{FDR}\le10\%$
  4725. \end_inset
  4726. \end_layout
  4727. \end_inset
  4728. </cell>
  4729. </row>
  4730. <row>
  4731. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  4732. \begin_inset Text
  4733. \begin_layout Plain Layout
  4734. Naïve Day 0 vs Day 1
  4735. \end_layout
  4736. \end_inset
  4737. </cell>
  4738. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  4739. \begin_inset Text
  4740. \begin_layout Plain Layout
  4741. 5992
  4742. \end_layout
  4743. \end_inset
  4744. </cell>
  4745. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  4746. \begin_inset Text
  4747. \begin_layout Plain Layout
  4748. 1613
  4749. \end_layout
  4750. \end_inset
  4751. </cell>
  4752. </row>
  4753. <row>
  4754. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  4755. \begin_inset Text
  4756. \begin_layout Plain Layout
  4757. Naïve Day 0 vs Day 5
  4758. \end_layout
  4759. \end_inset
  4760. </cell>
  4761. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  4762. \begin_inset Text
  4763. \begin_layout Plain Layout
  4764. 3038
  4765. \end_layout
  4766. \end_inset
  4767. </cell>
  4768. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  4769. \begin_inset Text
  4770. \begin_layout Plain Layout
  4771. 32
  4772. \end_layout
  4773. \end_inset
  4774. </cell>
  4775. </row>
  4776. <row>
  4777. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  4778. \begin_inset Text
  4779. \begin_layout Plain Layout
  4780. Naïve Day 0 vs Day 14
  4781. \end_layout
  4782. \end_inset
  4783. </cell>
  4784. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  4785. \begin_inset Text
  4786. \begin_layout Plain Layout
  4787. 1870
  4788. \end_layout
  4789. \end_inset
  4790. </cell>
  4791. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  4792. \begin_inset Text
  4793. \begin_layout Plain Layout
  4794. 190
  4795. \end_layout
  4796. \end_inset
  4797. </cell>
  4798. </row>
  4799. <row>
  4800. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  4801. \begin_inset Text
  4802. \begin_layout Plain Layout
  4803. Memory Day 0 vs Day 1
  4804. \end_layout
  4805. \end_inset
  4806. </cell>
  4807. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  4808. \begin_inset Text
  4809. \begin_layout Plain Layout
  4810. 3195
  4811. \end_layout
  4812. \end_inset
  4813. </cell>
  4814. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  4815. \begin_inset Text
  4816. \begin_layout Plain Layout
  4817. 411
  4818. \end_layout
  4819. \end_inset
  4820. </cell>
  4821. </row>
  4822. <row>
  4823. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  4824. \begin_inset Text
  4825. \begin_layout Plain Layout
  4826. Memory Day 0 vs Day 5
  4827. \end_layout
  4828. \end_inset
  4829. </cell>
  4830. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  4831. \begin_inset Text
  4832. \begin_layout Plain Layout
  4833. 2688
  4834. \end_layout
  4835. \end_inset
  4836. </cell>
  4837. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  4838. \begin_inset Text
  4839. \begin_layout Plain Layout
  4840. 18
  4841. \end_layout
  4842. \end_inset
  4843. </cell>
  4844. </row>
  4845. <row>
  4846. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  4847. \begin_inset Text
  4848. \begin_layout Plain Layout
  4849. Memory Day 0 vs Day 14
  4850. \end_layout
  4851. \end_inset
  4852. </cell>
  4853. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  4854. \begin_inset Text
  4855. \begin_layout Plain Layout
  4856. 1911
  4857. \end_layout
  4858. \end_inset
  4859. </cell>
  4860. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  4861. \begin_inset Text
  4862. \begin_layout Plain Layout
  4863. 227
  4864. \end_layout
  4865. \end_inset
  4866. </cell>
  4867. </row>
  4868. <row>
  4869. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  4870. \begin_inset Text
  4871. \begin_layout Plain Layout
  4872. Day 0 Naïve vs Memory
  4873. \end_layout
  4874. \end_inset
  4875. </cell>
  4876. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  4877. \begin_inset Text
  4878. \begin_layout Plain Layout
  4879. 0
  4880. \end_layout
  4881. \end_inset
  4882. </cell>
  4883. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  4884. \begin_inset Text
  4885. \begin_layout Plain Layout
  4886. 2
  4887. \end_layout
  4888. \end_inset
  4889. </cell>
  4890. </row>
  4891. <row>
  4892. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  4893. \begin_inset Text
  4894. \begin_layout Plain Layout
  4895. Day 1 Naïve vs Memory
  4896. \end_layout
  4897. \end_inset
  4898. </cell>
  4899. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  4900. \begin_inset Text
  4901. \begin_layout Plain Layout
  4902. 9167
  4903. \end_layout
  4904. \end_inset
  4905. </cell>
  4906. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  4907. \begin_inset Text
  4908. \begin_layout Plain Layout
  4909. 5532
  4910. \end_layout
  4911. \end_inset
  4912. </cell>
  4913. </row>
  4914. <row>
  4915. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  4916. \begin_inset Text
  4917. \begin_layout Plain Layout
  4918. Day 5 Naïve vs Memory
  4919. \end_layout
  4920. \end_inset
  4921. </cell>
  4922. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  4923. \begin_inset Text
  4924. \begin_layout Plain Layout
  4925. 0
  4926. \end_layout
  4927. \end_inset
  4928. </cell>
  4929. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  4930. \begin_inset Text
  4931. \begin_layout Plain Layout
  4932. 0
  4933. \end_layout
  4934. \end_inset
  4935. </cell>
  4936. </row>
  4937. <row>
  4938. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  4939. \begin_inset Text
  4940. \begin_layout Plain Layout
  4941. Day 14 Naïve vs Memory
  4942. \end_layout
  4943. \end_inset
  4944. </cell>
  4945. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  4946. \begin_inset Text
  4947. \begin_layout Plain Layout
  4948. 6446
  4949. \end_layout
  4950. \end_inset
  4951. </cell>
  4952. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  4953. \begin_inset Text
  4954. \begin_layout Plain Layout
  4955. 2319
  4956. \end_layout
  4957. \end_inset
  4958. </cell>
  4959. </row>
  4960. </lyxtabular>
  4961. \end_inset
  4962. \end_layout
  4963. \begin_layout Plain Layout
  4964. \begin_inset Caption Standard
  4965. \begin_layout Plain Layout
  4966. \begin_inset Argument 1
  4967. status collapsed
  4968. \begin_layout Plain Layout
  4969. Estimated and detected differentially expressed genes.
  4970. \end_layout
  4971. \end_inset
  4972. \begin_inset CommandInset label
  4973. LatexCommand label
  4974. name "tab:Estimated-and-detected-rnaseq"
  4975. \end_inset
  4976. \series bold
  4977. Estimated and detected differentially expressed genes.
  4978. \series default
  4979. \begin_inset Quotes eld
  4980. \end_inset
  4981. Test
  4982. \begin_inset Quotes erd
  4983. \end_inset
  4984. : Which sample groups were compared;
  4985. \begin_inset Quotes eld
  4986. \end_inset
  4987. Est non-null
  4988. \begin_inset Quotes erd
  4989. \end_inset
  4990. : Estimated number of differentially expressed genes, using the method of
  4991. averaging local FDR values
  4992. \begin_inset CommandInset citation
  4993. LatexCommand cite
  4994. key "Phipson2013Thesis"
  4995. literal "false"
  4996. \end_inset
  4997. ;
  4998. \begin_inset Quotes eld
  4999. \end_inset
  5000. \begin_inset Formula $\mathrm{FDR}\le10\%$
  5001. \end_inset
  5002. \begin_inset Quotes erd
  5003. \end_inset
  5004. : Number of significantly differentially expressed genes at an FDR threshold
  5005. of 10%.
  5006. The total number of genes tested was 16707.
  5007. \end_layout
  5008. \end_inset
  5009. \end_layout
  5010. \end_inset
  5011. \begin_inset Note Note
  5012. status collapsed
  5013. \begin_layout Plain Layout
  5014. If float lost issues, reposition randomly until success.
  5015. \end_layout
  5016. \end_inset
  5017. The batch effect has both a systematic component and a random noise component.
  5018. While the systematic component was subtracted out using ComBat (Figure
  5019. \begin_inset CommandInset ref
  5020. LatexCommand ref
  5021. reference "fig:RNA-PCA"
  5022. plural "false"
  5023. caps "false"
  5024. noprefix "false"
  5025. \end_inset
  5026. ), no such correction is possible for the noise component: Batch 1 simply
  5027. has substantially more random noise in it, which reduces the statistical
  5028. power for any differential expression tests involving samples in that batch.
  5029. \end_layout
  5030. \begin_layout Standard
  5031. Despite the difficulty in detecting specific differentially expressed genes,
  5032. there is still evidence that differential expression is present for these
  5033. time points.
  5034. In Figure
  5035. \begin_inset CommandInset ref
  5036. LatexCommand ref
  5037. reference "fig:rna-pca-final"
  5038. plural "false"
  5039. caps "false"
  5040. noprefix "false"
  5041. \end_inset
  5042. , there is a clear separation between naïve and memory samples at Day 0,
  5043. despite the fact that only 2 genes were significantly differentially expressed
  5044. for this comparison.
  5045. Similarly, the small numbers of genes detected for the Day 0 vs Day 5 compariso
  5046. ns do not reflect the large separation between these time points in Figure
  5047. \begin_inset CommandInset ref
  5048. LatexCommand ref
  5049. reference "fig:rna-pca-final"
  5050. plural "false"
  5051. caps "false"
  5052. noprefix "false"
  5053. \end_inset
  5054. .
  5055. In addition, the
  5056. \begin_inset Flex Glossary Term
  5057. status open
  5058. \begin_layout Plain Layout
  5059. MOFA
  5060. \end_layout
  5061. \end_inset
  5062. \begin_inset Flex Glossary Term
  5063. status open
  5064. \begin_layout Plain Layout
  5065. LF
  5066. \end_layout
  5067. \end_inset
  5068. plots in Figure
  5069. \begin_inset CommandInset ref
  5070. LatexCommand ref
  5071. reference "fig:mofa-lf-scatter"
  5072. plural "false"
  5073. caps "false"
  5074. noprefix "false"
  5075. \end_inset
  5076. .
  5077. This suggests that there is indeed a differential expression signal present
  5078. in the data for these comparisons, but the large variability in the Batch
  5079. 1 samples obfuscates this signal at the individual gene level.
  5080. As a result, it is impossible to make any meaningful statements about the
  5081. \begin_inset Quotes eld
  5082. \end_inset
  5083. size
  5084. \begin_inset Quotes erd
  5085. \end_inset
  5086. of the gene signature for any time point, since the number of significant
  5087. genes as well as the estimated number of differentially expressed genes
  5088. depends so strongly on the variations in sample quality in addition to
  5089. the size of the differential expression signal in the data.
  5090. Gene-set enrichment analyses are similarly impractical.
  5091. However, analyses looking at genome-wide patterns of expression are still
  5092. practical.
  5093. \end_layout
  5094. \begin_layout Standard
  5095. \begin_inset Float figure
  5096. wide false
  5097. sideways false
  5098. status collapsed
  5099. \begin_layout Plain Layout
  5100. \align center
  5101. \begin_inset Graphics
  5102. filename graphics/CD4-csaw/RNA-seq/PCA-final-12-CROP.png
  5103. lyxscale 25
  5104. width 100col%
  5105. groupId colwidth-raster
  5106. \end_inset
  5107. \end_layout
  5108. \begin_layout Plain Layout
  5109. \begin_inset Caption Standard
  5110. \begin_layout Plain Layout
  5111. \begin_inset Argument 1
  5112. status collapsed
  5113. \begin_layout Plain Layout
  5114. PCoA plot of RNA-seq samples after ComBat batch correction.
  5115. \end_layout
  5116. \end_inset
  5117. \begin_inset CommandInset label
  5118. LatexCommand label
  5119. name "fig:rna-pca-final"
  5120. \end_inset
  5121. \series bold
  5122. PCoA plot of RNA-seq samples after ComBat batch correction.
  5123. \series default
  5124. Each point represents an individual sample.
  5125. Samples with the same combination of cell type and time point are encircled
  5126. with a shaded region to aid in visual identification of the sample groups.
  5127. Samples of the same cell type from the same donor are connected by lines
  5128. to indicate the
  5129. \begin_inset Quotes eld
  5130. \end_inset
  5131. trajectory
  5132. \begin_inset Quotes erd
  5133. \end_inset
  5134. of each donor's cells over time in PCoA space.
  5135. \end_layout
  5136. \end_inset
  5137. \end_layout
  5138. \end_inset
  5139. \end_layout
  5140. \begin_layout Subsection
  5141. H3K4 and H3K27 methylation occur in broad regions and are enriched near
  5142. promoters
  5143. \end_layout
  5144. \begin_layout Standard
  5145. \begin_inset Float table
  5146. wide false
  5147. sideways false
  5148. status open
  5149. \begin_layout Plain Layout
  5150. \align center
  5151. \begin_inset Flex TODO Note (inline)
  5152. status open
  5153. \begin_layout Plain Layout
  5154. Also get
  5155. \emph on
  5156. median
  5157. \emph default
  5158. peak width and maybe other quantiles (25%, 75%)
  5159. \end_layout
  5160. \end_inset
  5161. \end_layout
  5162. \begin_layout Plain Layout
  5163. \align center
  5164. \begin_inset Tabular
  5165. <lyxtabular version="3" rows="4" columns="5">
  5166. <features tabularvalignment="middle">
  5167. <column alignment="center" valignment="top">
  5168. <column alignment="center" valignment="top">
  5169. <column alignment="center" valignment="top">
  5170. <column alignment="center" valignment="top">
  5171. <column alignment="center" valignment="top">
  5172. <row>
  5173. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  5174. \begin_inset Text
  5175. \begin_layout Plain Layout
  5176. Histone Mark
  5177. \end_layout
  5178. \end_inset
  5179. </cell>
  5180. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  5181. \begin_inset Text
  5182. \begin_layout Plain Layout
  5183. # Peaks
  5184. \end_layout
  5185. \end_inset
  5186. </cell>
  5187. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  5188. \begin_inset Text
  5189. \begin_layout Plain Layout
  5190. Mean peak width
  5191. \end_layout
  5192. \end_inset
  5193. </cell>
  5194. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  5195. \begin_inset Text
  5196. \begin_layout Plain Layout
  5197. genome coverage
  5198. \end_layout
  5199. \end_inset
  5200. </cell>
  5201. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  5202. \begin_inset Text
  5203. \begin_layout Plain Layout
  5204. FRiP
  5205. \end_layout
  5206. \end_inset
  5207. </cell>
  5208. </row>
  5209. <row>
  5210. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  5211. \begin_inset Text
  5212. \begin_layout Plain Layout
  5213. H3K4me2
  5214. \end_layout
  5215. \end_inset
  5216. </cell>
  5217. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  5218. \begin_inset Text
  5219. \begin_layout Plain Layout
  5220. 14,965
  5221. \end_layout
  5222. \end_inset
  5223. </cell>
  5224. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  5225. \begin_inset Text
  5226. \begin_layout Plain Layout
  5227. 3,970
  5228. \end_layout
  5229. \end_inset
  5230. </cell>
  5231. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  5232. \begin_inset Text
  5233. \begin_layout Plain Layout
  5234. 1.92%
  5235. \end_layout
  5236. \end_inset
  5237. </cell>
  5238. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  5239. \begin_inset Text
  5240. \begin_layout Plain Layout
  5241. 14.2%
  5242. \end_layout
  5243. \end_inset
  5244. </cell>
  5245. </row>
  5246. <row>
  5247. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  5248. \begin_inset Text
  5249. \begin_layout Plain Layout
  5250. H3K4me3
  5251. \end_layout
  5252. \end_inset
  5253. </cell>
  5254. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  5255. \begin_inset Text
  5256. \begin_layout Plain Layout
  5257. 6,163
  5258. \end_layout
  5259. \end_inset
  5260. </cell>
  5261. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  5262. \begin_inset Text
  5263. \begin_layout Plain Layout
  5264. 2,946
  5265. \end_layout
  5266. \end_inset
  5267. </cell>
  5268. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  5269. \begin_inset Text
  5270. \begin_layout Plain Layout
  5271. 0.588%
  5272. \end_layout
  5273. \end_inset
  5274. </cell>
  5275. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  5276. \begin_inset Text
  5277. \begin_layout Plain Layout
  5278. 6.57%
  5279. \end_layout
  5280. \end_inset
  5281. </cell>
  5282. </row>
  5283. <row>
  5284. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  5285. \begin_inset Text
  5286. \begin_layout Plain Layout
  5287. H3K27me3
  5288. \end_layout
  5289. \end_inset
  5290. </cell>
  5291. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  5292. \begin_inset Text
  5293. \begin_layout Plain Layout
  5294. 18,139
  5295. \end_layout
  5296. \end_inset
  5297. </cell>
  5298. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  5299. \begin_inset Text
  5300. \begin_layout Plain Layout
  5301. 18,967
  5302. \end_layout
  5303. \end_inset
  5304. </cell>
  5305. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  5306. \begin_inset Text
  5307. \begin_layout Plain Layout
  5308. 11.1%
  5309. \end_layout
  5310. \end_inset
  5311. </cell>
  5312. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  5313. \begin_inset Text
  5314. \begin_layout Plain Layout
  5315. 22.5%
  5316. \end_layout
  5317. \end_inset
  5318. </cell>
  5319. </row>
  5320. </lyxtabular>
  5321. \end_inset
  5322. \end_layout
  5323. \begin_layout Plain Layout
  5324. \begin_inset Flex TODO Note (inline)
  5325. status open
  5326. \begin_layout Plain Layout
  5327. Get the IDR threshold
  5328. \end_layout
  5329. \end_inset
  5330. \end_layout
  5331. \begin_layout Plain Layout
  5332. \begin_inset Caption Standard
  5333. \begin_layout Plain Layout
  5334. \begin_inset Argument 1
  5335. status collapsed
  5336. \begin_layout Plain Layout
  5337. Summary of peak-calling statistics.
  5338. \end_layout
  5339. \end_inset
  5340. \begin_inset CommandInset label
  5341. LatexCommand label
  5342. name "tab:peak-calling-summary"
  5343. \end_inset
  5344. \series bold
  5345. Summary of peak-calling statistics.
  5346. \series default
  5347. For each histone mark, the number of peaks called using SICER at an IDR
  5348. threshold of ???, the mean width of those peaks, the fraction of the genome
  5349. covered by peaks, and the fraction of reads in peaks (FRiP).
  5350. \end_layout
  5351. \end_inset
  5352. \end_layout
  5353. \end_inset
  5354. \end_layout
  5355. \begin_layout Standard
  5356. Table
  5357. \begin_inset CommandInset ref
  5358. LatexCommand ref
  5359. reference "tab:peak-calling-summary"
  5360. plural "false"
  5361. caps "false"
  5362. noprefix "false"
  5363. \end_inset
  5364. gives a summary of the peak calling statistics for each histone mark.
  5365. Consistent with previous observations, all 3 histone marks occur in broad
  5366. regions spanning many consecutive nucleosomes, rather than in sharp peaks
  5367. as would be expected for a transcription factor or other molecule that
  5368. binds to specific sites.
  5369. This conclusion is further supported by Figure
  5370. \begin_inset CommandInset ref
  5371. LatexCommand ref
  5372. reference "fig:CCF-with-blacklist"
  5373. plural "false"
  5374. caps "false"
  5375. noprefix "false"
  5376. \end_inset
  5377. , in which a clear nucleosome-sized periodicity is visible in the cross-correlat
  5378. ion value for each sample, indicating that each time a given mark is present
  5379. on one histone, it is also likely to be found on adjacent histones as well.
  5380. H3K27me3 enrichment in particular is substantially more broad than either
  5381. H3K4 mark, with a mean peak width of almost 19,000 bp.
  5382. This is also reflected in the periodicity observed in Figure
  5383. \begin_inset CommandInset ref
  5384. LatexCommand ref
  5385. reference "fig:CCF-with-blacklist"
  5386. plural "false"
  5387. caps "false"
  5388. noprefix "false"
  5389. \end_inset
  5390. , which remains strong much farther out for H3K27me3 than the other marks,
  5391. showing H3K27me3 especially tends to be found on long runs of consecutive
  5392. histones.
  5393. \end_layout
  5394. \begin_layout Standard
  5395. \begin_inset Flex TODO Note (inline)
  5396. status open
  5397. \begin_layout Plain Layout
  5398. \end_layout
  5399. \end_inset
  5400. \end_layout
  5401. \begin_layout Standard
  5402. All 3 histone marks tend to occur more often near promoter regions, as shown
  5403. in Figure
  5404. \begin_inset CommandInset ref
  5405. LatexCommand ref
  5406. reference "fig:near-promoter-peak-enrich"
  5407. plural "false"
  5408. caps "false"
  5409. noprefix "false"
  5410. \end_inset
  5411. .
  5412. The majority of each density distribution is flat, representing the background
  5413. density of peaks genome-wide.
  5414. Each distribution has a peak near zero, representing an enrichment of peaks
  5415. close to
  5416. \begin_inset Flex Glossary Term
  5417. status open
  5418. \begin_layout Plain Layout
  5419. TSS
  5420. \end_layout
  5421. \end_inset
  5422. positions relative to the remainder of the genome.
  5423. Interestingly, the
  5424. \begin_inset Quotes eld
  5425. \end_inset
  5426. radius
  5427. \begin_inset Quotes erd
  5428. \end_inset
  5429. within which this enrichment occurs is not the same for every histone mark
  5430. (Table
  5431. \begin_inset CommandInset ref
  5432. LatexCommand ref
  5433. reference "tab:effective-promoter-radius"
  5434. plural "false"
  5435. caps "false"
  5436. noprefix "false"
  5437. \end_inset
  5438. ).
  5439. For H3K4me2 and H3K4me3, peaks are most enriched within 1
  5440. \begin_inset space ~
  5441. \end_inset
  5442. kbp of
  5443. \begin_inset Flex Glossary Term
  5444. status open
  5445. \begin_layout Plain Layout
  5446. TSS
  5447. \end_layout
  5448. \end_inset
  5449. positions, while for H3K27me3, enrichment is broader, extending to 2.5
  5450. \begin_inset space ~
  5451. \end_inset
  5452. kbp.
  5453. These
  5454. \begin_inset Quotes eld
  5455. \end_inset
  5456. effective promoter radii
  5457. \begin_inset Quotes erd
  5458. \end_inset
  5459. remain approximately the same across all combinations of experimental condition
  5460. (cell type, time point, and donor), so they appear to be a property of
  5461. the histone mark itself.
  5462. Hence, these radii were used to define the promoter regions for each histone
  5463. mark in all further analyses.
  5464. \end_layout
  5465. \begin_layout Standard
  5466. \begin_inset Float figure
  5467. wide false
  5468. sideways false
  5469. status open
  5470. \begin_layout Plain Layout
  5471. \align center
  5472. \begin_inset Graphics
  5473. filename graphics/CD4-csaw/Promoter-Peak-Distance-Profile-PAGE1-CROP.pdf
  5474. lyxscale 50
  5475. width 80col%
  5476. \end_inset
  5477. \end_layout
  5478. \begin_layout Plain Layout
  5479. \begin_inset Flex TODO Note (inline)
  5480. status open
  5481. \begin_layout Plain Layout
  5482. Future direction idea: Need a control: shuffle all peaks and repeat, N times.
  5483. \end_layout
  5484. \end_inset
  5485. \end_layout
  5486. \begin_layout Plain Layout
  5487. \begin_inset Caption Standard
  5488. \begin_layout Plain Layout
  5489. \begin_inset Argument 1
  5490. status collapsed
  5491. \begin_layout Plain Layout
  5492. Enrichment of peaks in promoter neighborhoods.
  5493. \end_layout
  5494. \end_inset
  5495. \begin_inset CommandInset label
  5496. LatexCommand label
  5497. name "fig:near-promoter-peak-enrich"
  5498. \end_inset
  5499. \series bold
  5500. Enrichment of peaks in promoter neighborhoods.
  5501. \series default
  5502. This plot shows the distribution of distances from each annotated transcription
  5503. start site in the genome to the nearest called peak.
  5504. Each line represents one combination of histone mark, cell type, and time
  5505. point.
  5506. Distributions are smoothed using kernel density estimation.
  5507. TSSs that occur
  5508. \emph on
  5509. within
  5510. \emph default
  5511. peaks were excluded from this plot to avoid a large spike at zero that
  5512. would overshadow the rest of the distribution.
  5513. (Note: this figure was generated using the original peak calls and expression
  5514. values from
  5515. \begin_inset Flex Glossary Term
  5516. status open
  5517. \begin_layout Plain Layout
  5518. GEO
  5519. \end_layout
  5520. \end_inset
  5521. \begin_inset CommandInset citation
  5522. LatexCommand cite
  5523. key "LaMere2016"
  5524. literal "false"
  5525. \end_inset
  5526. .)
  5527. \end_layout
  5528. \end_inset
  5529. \end_layout
  5530. \end_inset
  5531. \end_layout
  5532. \begin_layout Standard
  5533. \begin_inset Float table
  5534. wide false
  5535. sideways false
  5536. status collapsed
  5537. \begin_layout Plain Layout
  5538. \align center
  5539. \begin_inset Tabular
  5540. <lyxtabular version="3" rows="4" columns="2">
  5541. <features tabularvalignment="middle">
  5542. <column alignment="center" valignment="top">
  5543. <column alignment="center" valignment="top">
  5544. <row>
  5545. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  5546. \begin_inset Text
  5547. \begin_layout Plain Layout
  5548. Histone mark
  5549. \end_layout
  5550. \end_inset
  5551. </cell>
  5552. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  5553. \begin_inset Text
  5554. \begin_layout Plain Layout
  5555. Effective promoter radius
  5556. \end_layout
  5557. \end_inset
  5558. </cell>
  5559. </row>
  5560. <row>
  5561. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  5562. \begin_inset Text
  5563. \begin_layout Plain Layout
  5564. H3K4me2
  5565. \end_layout
  5566. \end_inset
  5567. </cell>
  5568. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  5569. \begin_inset Text
  5570. \begin_layout Plain Layout
  5571. 1 kbp
  5572. \end_layout
  5573. \end_inset
  5574. </cell>
  5575. </row>
  5576. <row>
  5577. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  5578. \begin_inset Text
  5579. \begin_layout Plain Layout
  5580. H3K4me3
  5581. \end_layout
  5582. \end_inset
  5583. </cell>
  5584. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  5585. \begin_inset Text
  5586. \begin_layout Plain Layout
  5587. 1 kbp
  5588. \end_layout
  5589. \end_inset
  5590. </cell>
  5591. </row>
  5592. <row>
  5593. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  5594. \begin_inset Text
  5595. \begin_layout Plain Layout
  5596. H3K27me3
  5597. \end_layout
  5598. \end_inset
  5599. </cell>
  5600. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  5601. \begin_inset Text
  5602. \begin_layout Plain Layout
  5603. 2.5 kbp
  5604. \end_layout
  5605. \end_inset
  5606. </cell>
  5607. </row>
  5608. </lyxtabular>
  5609. \end_inset
  5610. \end_layout
  5611. \begin_layout Plain Layout
  5612. \begin_inset Caption Standard
  5613. \begin_layout Plain Layout
  5614. \begin_inset Argument 1
  5615. status collapsed
  5616. \begin_layout Plain Layout
  5617. Effective promoter radius for each histone mark.
  5618. \end_layout
  5619. \end_inset
  5620. \begin_inset CommandInset label
  5621. LatexCommand label
  5622. name "tab:effective-promoter-radius"
  5623. \end_inset
  5624. \series bold
  5625. Effective promoter radius for each histone mark.
  5626. \series default
  5627. These values represent the approximate distance from transcription start
  5628. site positions within which an excess of peaks are found, as shown in Figure
  5629. \begin_inset CommandInset ref
  5630. LatexCommand ref
  5631. reference "fig:near-promoter-peak-enrich"
  5632. plural "false"
  5633. caps "false"
  5634. noprefix "false"
  5635. \end_inset
  5636. .
  5637. \end_layout
  5638. \end_inset
  5639. \end_layout
  5640. \end_inset
  5641. \end_layout
  5642. \begin_layout Standard
  5643. \begin_inset Flex TODO Note (inline)
  5644. status open
  5645. \begin_layout Plain Layout
  5646. Consider also showing figure for distance to nearest peak center, and reference
  5647. median peak size once that is known.
  5648. \end_layout
  5649. \end_inset
  5650. \end_layout
  5651. \begin_layout Subsection
  5652. Correlations between gene expression and promoter methylation follow expected
  5653. genome-wide trends
  5654. \end_layout
  5655. \begin_layout Standard
  5656. H3K4me2 and H3K4me2 have previously been reported as activating marks whose
  5657. presence in a gene's promoter is associated with higher gene expression,
  5658. while H3K27me3 has been reported as inactivating
  5659. \begin_inset CommandInset citation
  5660. LatexCommand cite
  5661. key "LaMere2016,LaMere2017"
  5662. literal "false"
  5663. \end_inset
  5664. .
  5665. The data are consistent with this characterization: genes whose promoters
  5666. (as defined by the radii for each histone mark listed in
  5667. \begin_inset CommandInset ref
  5668. LatexCommand ref
  5669. reference "tab:effective-promoter-radius"
  5670. plural "false"
  5671. caps "false"
  5672. noprefix "false"
  5673. \end_inset
  5674. ) overlap with a H3K4me2 or H3K4me3 peak tend to have higher expression
  5675. than those that don't, while H3K27me3 is likewise associated with lower
  5676. gene expression, as shown in
  5677. \begin_inset CommandInset ref
  5678. LatexCommand ref
  5679. reference "fig:fpkm-by-peak"
  5680. plural "false"
  5681. caps "false"
  5682. noprefix "false"
  5683. \end_inset
  5684. .
  5685. This pattern holds across all combinations of cell type and time point
  5686. (Welch's
  5687. \emph on
  5688. t
  5689. \emph default
  5690. -test, all
  5691. \begin_inset Formula $p\textrm{-values}\ll2.2\times10^{-16}$
  5692. \end_inset
  5693. ).
  5694. The difference in average
  5695. \begin_inset Formula $\log_{2}$
  5696. \end_inset
  5697. \begin_inset Flex Glossary Term
  5698. status open
  5699. \begin_layout Plain Layout
  5700. FPKM
  5701. \end_layout
  5702. \end_inset
  5703. values when a peak overlaps the promoter is about
  5704. \begin_inset Formula $+5.67$
  5705. \end_inset
  5706. for H3K4me2,
  5707. \begin_inset Formula $+5.76$
  5708. \end_inset
  5709. for H3K4me2, and
  5710. \begin_inset Formula $-4.00$
  5711. \end_inset
  5712. for H3K27me3.
  5713. \end_layout
  5714. \begin_layout Standard
  5715. \begin_inset ERT
  5716. status open
  5717. \begin_layout Plain Layout
  5718. \backslash
  5719. afterpage{
  5720. \end_layout
  5721. \begin_layout Plain Layout
  5722. \backslash
  5723. begin{landscape}
  5724. \end_layout
  5725. \end_inset
  5726. \end_layout
  5727. \begin_layout Standard
  5728. \begin_inset Float figure
  5729. wide false
  5730. sideways false
  5731. status collapsed
  5732. \begin_layout Plain Layout
  5733. \align center
  5734. \begin_inset Graphics
  5735. filename graphics/CD4-csaw/FPKM-by-Peak-Violin-Plots-CROP.pdf
  5736. lyxscale 50
  5737. height 80theight%
  5738. \end_inset
  5739. \end_layout
  5740. \begin_layout Plain Layout
  5741. \begin_inset Caption Standard
  5742. \begin_layout Plain Layout
  5743. \begin_inset Argument 1
  5744. status collapsed
  5745. \begin_layout Plain Layout
  5746. Expression distributions of genes with and without promoter peaks.
  5747. \end_layout
  5748. \end_inset
  5749. \begin_inset CommandInset label
  5750. LatexCommand label
  5751. name "fig:fpkm-by-peak"
  5752. \end_inset
  5753. \series bold
  5754. Expression distributions of genes with and without promoter peaks.
  5755. \series default
  5756. For each histone mark in each experimental condition, the average RNA-seq
  5757. abundance (
  5758. \begin_inset Formula $\log_{2}$
  5759. \end_inset
  5760. FPKM) of each gene across all 4 donors was calculated.
  5761. Genes were grouped based on whether or not a peak was called in their promoters
  5762. in that condition, and the distribution of abundance values was plotted
  5763. for the no-peak and peak groups.
  5764. (Note: this figure was generated using the original peak calls and expression
  5765. values from
  5766. \begin_inset Flex Glossary Term
  5767. status open
  5768. \begin_layout Plain Layout
  5769. GEO
  5770. \end_layout
  5771. \end_inset
  5772. \begin_inset CommandInset citation
  5773. LatexCommand cite
  5774. key "LaMere2016"
  5775. literal "false"
  5776. \end_inset
  5777. .)
  5778. \end_layout
  5779. \end_inset
  5780. \end_layout
  5781. \end_inset
  5782. \end_layout
  5783. \begin_layout Standard
  5784. \begin_inset ERT
  5785. status open
  5786. \begin_layout Plain Layout
  5787. \backslash
  5788. end{landscape}
  5789. \end_layout
  5790. \begin_layout Plain Layout
  5791. }
  5792. \end_layout
  5793. \end_inset
  5794. \end_layout
  5795. \begin_layout Subsection
  5796. Gene expression and promoter histone methylation patterns show convergence
  5797. between naïve and memory cells at day 14
  5798. \end_layout
  5799. \begin_layout Standard
  5800. We hypothesized that if naïve cells had differentiated into memory cells
  5801. by Day 14, then their patterns of expression and histone modification should
  5802. converge with those of memory cells at Day 14.
  5803. Figure
  5804. \begin_inset CommandInset ref
  5805. LatexCommand ref
  5806. reference "fig:PCoA-promoters"
  5807. plural "false"
  5808. caps "false"
  5809. noprefix "false"
  5810. \end_inset
  5811. shows the patterns of variation in all 3 histone marks in the promoter
  5812. regions of the genome using
  5813. \begin_inset Flex Glossary Term
  5814. status open
  5815. \begin_layout Plain Layout
  5816. PCoA
  5817. \end_layout
  5818. \end_inset
  5819. .
  5820. All 3 marks show a noticeable convergence between the naïve and memory
  5821. samples at day 14, visible as an overlapping of the day 14 groups on each
  5822. plot.
  5823. This is consistent with the counts of significantly differentially modified
  5824. promoters and estimates of the total numbers of differentially modified
  5825. promoters shown in Table
  5826. \begin_inset CommandInset ref
  5827. LatexCommand ref
  5828. reference "tab:Number-signif-promoters"
  5829. plural "false"
  5830. caps "false"
  5831. noprefix "false"
  5832. \end_inset
  5833. .
  5834. For all histone marks, evidence of differential modification between naïve
  5835. and memory samples was detected at every time point except day 14.
  5836. The day 14 convergence pattern is also present in the
  5837. \begin_inset Flex Glossary Term
  5838. status open
  5839. \begin_layout Plain Layout
  5840. RNA-seq
  5841. \end_layout
  5842. \end_inset
  5843. data (Figure
  5844. \begin_inset CommandInset ref
  5845. LatexCommand ref
  5846. reference "fig:RNA-PCA-group"
  5847. plural "false"
  5848. caps "false"
  5849. noprefix "false"
  5850. \end_inset
  5851. ), albeit in the 2nd and 3rd principal coordinates, indicating that it is
  5852. not the most dominant pattern driving gene expression.
  5853. Taken together, the data show that promoter histone methylation for these
  5854. 3 histone marks and RNA expression for naïve and memory cells are most
  5855. similar at day 14, the furthest time point after activation.
  5856. \begin_inset Flex Glossary Term
  5857. status open
  5858. \begin_layout Plain Layout
  5859. MOFA
  5860. \end_layout
  5861. \end_inset
  5862. was also able to capture this day 14 convergence pattern in
  5863. \begin_inset Flex Glossary Term
  5864. status open
  5865. \begin_layout Plain Layout
  5866. LF
  5867. \end_layout
  5868. \end_inset
  5869. 5 (Figure
  5870. \begin_inset CommandInset ref
  5871. LatexCommand ref
  5872. reference "fig:mofa-lf-scatter"
  5873. plural "false"
  5874. caps "false"
  5875. noprefix "false"
  5876. \end_inset
  5877. ), which accounts for shared variation across all 3 histone marks and the
  5878. \begin_inset Flex Glossary Term
  5879. status open
  5880. \begin_layout Plain Layout
  5881. RNA-seq
  5882. \end_layout
  5883. \end_inset
  5884. data, confirming that this convergence is a coordinated pattern across
  5885. all 4 data sets.
  5886. While this observation does not prove that the naïve cells have differentiated
  5887. into memory cells at Day 14, it is consistent with that hypothesis.
  5888. \end_layout
  5889. \begin_layout Standard
  5890. \begin_inset Float figure
  5891. placement p
  5892. wide false
  5893. sideways false
  5894. status collapsed
  5895. \begin_layout Plain Layout
  5896. \align center
  5897. \begin_inset Float figure
  5898. wide false
  5899. sideways false
  5900. status open
  5901. \begin_layout Plain Layout
  5902. \align center
  5903. \begin_inset Graphics
  5904. filename graphics/CD4-csaw/ChIP-seq/H3K4me2-promoter-PCA-group-CROP.png
  5905. lyxscale 25
  5906. width 45col%
  5907. groupId pcoa-prom-subfig
  5908. \end_inset
  5909. \end_layout
  5910. \begin_layout Plain Layout
  5911. \begin_inset Caption Standard
  5912. \begin_layout Plain Layout
  5913. \begin_inset CommandInset label
  5914. LatexCommand label
  5915. name "fig:PCoA-H3K4me2-prom"
  5916. \end_inset
  5917. PCoA plot of H3K4me2 promoters, after subtracting surrogate variables.
  5918. \end_layout
  5919. \end_inset
  5920. \end_layout
  5921. \end_inset
  5922. \begin_inset space \hfill{}
  5923. \end_inset
  5924. \begin_inset Float figure
  5925. wide false
  5926. sideways false
  5927. status open
  5928. \begin_layout Plain Layout
  5929. \align center
  5930. \begin_inset Graphics
  5931. filename graphics/CD4-csaw/ChIP-seq/H3K4me3-promoter-PCA-group-CROP.png
  5932. lyxscale 25
  5933. width 45col%
  5934. groupId pcoa-prom-subfig
  5935. \end_inset
  5936. \end_layout
  5937. \begin_layout Plain Layout
  5938. \begin_inset Caption Standard
  5939. \begin_layout Plain Layout
  5940. \begin_inset CommandInset label
  5941. LatexCommand label
  5942. name "fig:PCoA-H3K4me3-prom"
  5943. \end_inset
  5944. PCoA plot of H3K4me3 promoters, after subtracting surrogate variables.
  5945. \end_layout
  5946. \end_inset
  5947. \end_layout
  5948. \end_inset
  5949. \end_layout
  5950. \begin_layout Plain Layout
  5951. \align center
  5952. \begin_inset Float figure
  5953. wide false
  5954. sideways false
  5955. status open
  5956. \begin_layout Plain Layout
  5957. \align center
  5958. \begin_inset Graphics
  5959. filename graphics/CD4-csaw/ChIP-seq/H3K27me3-promoter-PCA-group-CROP.png
  5960. lyxscale 25
  5961. width 45col%
  5962. groupId pcoa-prom-subfig
  5963. \end_inset
  5964. \end_layout
  5965. \begin_layout Plain Layout
  5966. \begin_inset Caption Standard
  5967. \begin_layout Plain Layout
  5968. \begin_inset CommandInset label
  5969. LatexCommand label
  5970. name "fig:PCoA-H3K27me3-prom"
  5971. \end_inset
  5972. PCoA plot of H3K27me3 promoters, after subtracting surrogate variables.
  5973. \end_layout
  5974. \end_inset
  5975. \end_layout
  5976. \end_inset
  5977. \begin_inset space \hfill{}
  5978. \end_inset
  5979. \begin_inset Float figure
  5980. wide false
  5981. sideways false
  5982. status open
  5983. \begin_layout Plain Layout
  5984. \align center
  5985. \begin_inset Graphics
  5986. filename graphics/CD4-csaw/RNA-seq/PCA-final-23-CROP.png
  5987. lyxscale 25
  5988. width 45col%
  5989. groupId pcoa-prom-subfig
  5990. \end_inset
  5991. \end_layout
  5992. \begin_layout Plain Layout
  5993. \begin_inset Caption Standard
  5994. \begin_layout Plain Layout
  5995. \begin_inset CommandInset label
  5996. LatexCommand label
  5997. name "fig:RNA-PCA-group"
  5998. \end_inset
  5999. RNA-seq PCoA, after ComBat batch correction, showing principal coordinates
  6000. 2 and 3.
  6001. \end_layout
  6002. \end_inset
  6003. \end_layout
  6004. \end_inset
  6005. \end_layout
  6006. \begin_layout Plain Layout
  6007. \begin_inset Flex TODO Note (inline)
  6008. status open
  6009. \begin_layout Plain Layout
  6010. Figure font too small
  6011. \end_layout
  6012. \end_inset
  6013. \end_layout
  6014. \begin_layout Plain Layout
  6015. \begin_inset Caption Standard
  6016. \begin_layout Plain Layout
  6017. \begin_inset Argument 1
  6018. status collapsed
  6019. \begin_layout Plain Layout
  6020. PCoA plots for promoter ChIP-seq and expression RNA-seq data
  6021. \end_layout
  6022. \end_inset
  6023. \begin_inset CommandInset label
  6024. LatexCommand label
  6025. name "fig:PCoA-promoters"
  6026. \end_inset
  6027. \series bold
  6028. PCoA plots for promoter ChIP-seq and expression RNA-seq data.
  6029. \series default
  6030. Each point represents an individual sample.
  6031. Samples with the same combination of cell type and time point are encircled
  6032. with a shaded region to aid in visual identification of the sample groups.
  6033. Samples of the same cell type from the same donor are connected by lines
  6034. to indicate the
  6035. \begin_inset Quotes eld
  6036. \end_inset
  6037. trajectory
  6038. \begin_inset Quotes erd
  6039. \end_inset
  6040. of each donor's cells over time in PCoA space.
  6041. \end_layout
  6042. \end_inset
  6043. \end_layout
  6044. \end_inset
  6045. \end_layout
  6046. \begin_layout Standard
  6047. \begin_inset ERT
  6048. status open
  6049. \begin_layout Plain Layout
  6050. \backslash
  6051. afterpage{
  6052. \end_layout
  6053. \begin_layout Plain Layout
  6054. \backslash
  6055. begin{landscape}
  6056. \end_layout
  6057. \end_inset
  6058. \end_layout
  6059. \begin_layout Standard
  6060. \begin_inset Float table
  6061. wide false
  6062. sideways false
  6063. status collapsed
  6064. \begin_layout Plain Layout
  6065. \align center
  6066. \begin_inset Tabular
  6067. <lyxtabular version="3" rows="6" columns="7">
  6068. <features tabularvalignment="middle">
  6069. <column alignment="center" valignment="top">
  6070. <column alignment="center" valignment="top">
  6071. <column alignment="center" valignment="top">
  6072. <column alignment="center" valignment="top">
  6073. <column alignment="center" valignment="top">
  6074. <column alignment="center" valignment="top">
  6075. <column alignment="center" valignment="top">
  6076. <row>
  6077. <cell alignment="center" valignment="top" usebox="none">
  6078. \begin_inset Text
  6079. \begin_layout Plain Layout
  6080. \end_layout
  6081. \end_inset
  6082. </cell>
  6083. <cell multicolumn="1" alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  6084. \begin_inset Text
  6085. \begin_layout Plain Layout
  6086. Number of significant promoters
  6087. \end_layout
  6088. \end_inset
  6089. </cell>
  6090. <cell multicolumn="2" alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  6091. \begin_inset Text
  6092. \begin_layout Plain Layout
  6093. \end_layout
  6094. \end_inset
  6095. </cell>
  6096. <cell multicolumn="2" alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  6097. \begin_inset Text
  6098. \begin_layout Plain Layout
  6099. \end_layout
  6100. \end_inset
  6101. </cell>
  6102. <cell multicolumn="1" alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  6103. \begin_inset Text
  6104. \begin_layout Plain Layout
  6105. Est.
  6106. differentially modified promoters
  6107. \end_layout
  6108. \end_inset
  6109. </cell>
  6110. <cell multicolumn="2" alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  6111. \begin_inset Text
  6112. \begin_layout Plain Layout
  6113. \end_layout
  6114. \end_inset
  6115. </cell>
  6116. <cell multicolumn="2" alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  6117. \begin_inset Text
  6118. \begin_layout Plain Layout
  6119. \end_layout
  6120. \end_inset
  6121. </cell>
  6122. </row>
  6123. <row>
  6124. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  6125. \begin_inset Text
  6126. \begin_layout Plain Layout
  6127. Time Point
  6128. \end_layout
  6129. \end_inset
  6130. </cell>
  6131. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  6132. \begin_inset Text
  6133. \begin_layout Plain Layout
  6134. H3K4me2
  6135. \end_layout
  6136. \end_inset
  6137. </cell>
  6138. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  6139. \begin_inset Text
  6140. \begin_layout Plain Layout
  6141. H3K4me3
  6142. \end_layout
  6143. \end_inset
  6144. </cell>
  6145. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  6146. \begin_inset Text
  6147. \begin_layout Plain Layout
  6148. H3K27me3
  6149. \end_layout
  6150. \end_inset
  6151. </cell>
  6152. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  6153. \begin_inset Text
  6154. \begin_layout Plain Layout
  6155. H3K4me2
  6156. \end_layout
  6157. \end_inset
  6158. </cell>
  6159. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  6160. \begin_inset Text
  6161. \begin_layout Plain Layout
  6162. H3K4me3
  6163. \end_layout
  6164. \end_inset
  6165. </cell>
  6166. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  6167. \begin_inset Text
  6168. \begin_layout Plain Layout
  6169. H3K27me3
  6170. \end_layout
  6171. \end_inset
  6172. </cell>
  6173. </row>
  6174. <row>
  6175. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  6176. \begin_inset Text
  6177. \begin_layout Plain Layout
  6178. Day 0
  6179. \end_layout
  6180. \end_inset
  6181. </cell>
  6182. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  6183. \begin_inset Text
  6184. \begin_layout Plain Layout
  6185. 4553
  6186. \end_layout
  6187. \end_inset
  6188. </cell>
  6189. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  6190. \begin_inset Text
  6191. \begin_layout Plain Layout
  6192. 927
  6193. \end_layout
  6194. \end_inset
  6195. </cell>
  6196. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  6197. \begin_inset Text
  6198. \begin_layout Plain Layout
  6199. 6
  6200. \end_layout
  6201. \end_inset
  6202. </cell>
  6203. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  6204. \begin_inset Text
  6205. \begin_layout Plain Layout
  6206. 9967
  6207. \end_layout
  6208. \end_inset
  6209. </cell>
  6210. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  6211. \begin_inset Text
  6212. \begin_layout Plain Layout
  6213. 4149
  6214. \end_layout
  6215. \end_inset
  6216. </cell>
  6217. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  6218. \begin_inset Text
  6219. \begin_layout Plain Layout
  6220. 2404
  6221. \end_layout
  6222. \end_inset
  6223. </cell>
  6224. </row>
  6225. <row>
  6226. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  6227. \begin_inset Text
  6228. \begin_layout Plain Layout
  6229. Day 1
  6230. \end_layout
  6231. \end_inset
  6232. </cell>
  6233. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  6234. \begin_inset Text
  6235. \begin_layout Plain Layout
  6236. 567
  6237. \end_layout
  6238. \end_inset
  6239. </cell>
  6240. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  6241. \begin_inset Text
  6242. \begin_layout Plain Layout
  6243. 278
  6244. \end_layout
  6245. \end_inset
  6246. </cell>
  6247. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  6248. \begin_inset Text
  6249. \begin_layout Plain Layout
  6250. 1570
  6251. \end_layout
  6252. \end_inset
  6253. </cell>
  6254. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  6255. \begin_inset Text
  6256. \begin_layout Plain Layout
  6257. 4370
  6258. \end_layout
  6259. \end_inset
  6260. </cell>
  6261. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  6262. \begin_inset Text
  6263. \begin_layout Plain Layout
  6264. 2145
  6265. \end_layout
  6266. \end_inset
  6267. </cell>
  6268. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  6269. \begin_inset Text
  6270. \begin_layout Plain Layout
  6271. 6598
  6272. \end_layout
  6273. \end_inset
  6274. </cell>
  6275. </row>
  6276. <row>
  6277. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  6278. \begin_inset Text
  6279. \begin_layout Plain Layout
  6280. Day 5
  6281. \end_layout
  6282. \end_inset
  6283. </cell>
  6284. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  6285. \begin_inset Text
  6286. \begin_layout Plain Layout
  6287. 2313
  6288. \end_layout
  6289. \end_inset
  6290. </cell>
  6291. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  6292. \begin_inset Text
  6293. \begin_layout Plain Layout
  6294. 139
  6295. \end_layout
  6296. \end_inset
  6297. </cell>
  6298. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  6299. \begin_inset Text
  6300. \begin_layout Plain Layout
  6301. 490
  6302. \end_layout
  6303. \end_inset
  6304. </cell>
  6305. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  6306. \begin_inset Text
  6307. \begin_layout Plain Layout
  6308. 9450
  6309. \end_layout
  6310. \end_inset
  6311. </cell>
  6312. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  6313. \begin_inset Text
  6314. \begin_layout Plain Layout
  6315. 1148
  6316. \end_layout
  6317. \end_inset
  6318. </cell>
  6319. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  6320. \begin_inset Text
  6321. \begin_layout Plain Layout
  6322. 4141
  6323. \end_layout
  6324. \end_inset
  6325. </cell>
  6326. </row>
  6327. <row>
  6328. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  6329. \begin_inset Text
  6330. \begin_layout Plain Layout
  6331. Day 14
  6332. \end_layout
  6333. \end_inset
  6334. </cell>
  6335. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  6336. \begin_inset Text
  6337. \begin_layout Plain Layout
  6338. 0
  6339. \end_layout
  6340. \end_inset
  6341. </cell>
  6342. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  6343. \begin_inset Text
  6344. \begin_layout Plain Layout
  6345. 0
  6346. \end_layout
  6347. \end_inset
  6348. </cell>
  6349. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  6350. \begin_inset Text
  6351. \begin_layout Plain Layout
  6352. 0
  6353. \end_layout
  6354. \end_inset
  6355. </cell>
  6356. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  6357. \begin_inset Text
  6358. \begin_layout Plain Layout
  6359. 0
  6360. \end_layout
  6361. \end_inset
  6362. </cell>
  6363. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  6364. \begin_inset Text
  6365. \begin_layout Plain Layout
  6366. 0
  6367. \end_layout
  6368. \end_inset
  6369. </cell>
  6370. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  6371. \begin_inset Text
  6372. \begin_layout Plain Layout
  6373. 0
  6374. \end_layout
  6375. \end_inset
  6376. </cell>
  6377. </row>
  6378. </lyxtabular>
  6379. \end_inset
  6380. \end_layout
  6381. \begin_layout Plain Layout
  6382. \begin_inset Caption Standard
  6383. \begin_layout Plain Layout
  6384. \begin_inset Argument 1
  6385. status collapsed
  6386. \begin_layout Plain Layout
  6387. Number of differentially modified promoters between naïve and memory cells
  6388. at each time point after activation.
  6389. \end_layout
  6390. \end_inset
  6391. \begin_inset CommandInset label
  6392. LatexCommand label
  6393. name "tab:Number-signif-promoters"
  6394. \end_inset
  6395. \series bold
  6396. Number of differentially modified promoters between naïve and memory cells
  6397. at each time point after activation.
  6398. \series default
  6399. This table shows both the number of differentially modified promoters detected
  6400. at a 10% FDR threshold (left half), and the total number of differentially
  6401. modified promoters estimated using the method of averaging local FDR estimates
  6402. \begin_inset CommandInset citation
  6403. LatexCommand cite
  6404. key "Phipson2016"
  6405. literal "false"
  6406. \end_inset
  6407. (right half).
  6408. \end_layout
  6409. \end_inset
  6410. \end_layout
  6411. \end_inset
  6412. \end_layout
  6413. \begin_layout Standard
  6414. \begin_inset ERT
  6415. status open
  6416. \begin_layout Plain Layout
  6417. \backslash
  6418. end{landscape}
  6419. \end_layout
  6420. \begin_layout Plain Layout
  6421. }
  6422. \end_layout
  6423. \end_inset
  6424. \end_layout
  6425. \begin_layout Subsection
  6426. Association between resting H3K4me2 and H3K4me3 promoter coverage landscapes
  6427. and gene expression
  6428. \end_layout
  6429. \begin_layout Standard
  6430. \begin_inset Flex TODO Note (inline)
  6431. status open
  6432. \begin_layout Plain Layout
  6433. Need a better section title, for this and the next one.
  6434. \end_layout
  6435. \end_inset
  6436. \end_layout
  6437. \begin_layout Standard
  6438. \begin_inset Flex TODO Note (inline)
  6439. status open
  6440. \begin_layout Plain Layout
  6441. Make sure use of coverage/abundance/whatever is consistent.
  6442. \end_layout
  6443. \end_inset
  6444. \end_layout
  6445. \begin_layout Standard
  6446. \begin_inset Flex TODO Note (inline)
  6447. status open
  6448. \begin_layout Plain Layout
  6449. For the figures in this section and the next, the group labels are arbitrary,
  6450. so if time allows, it would be good to manually reorder them in a logical
  6451. way, e.g.
  6452. most upstream to most downstream.
  6453. If this is done, make sure to update the text with the correct group labels.
  6454. \end_layout
  6455. \end_inset
  6456. \end_layout
  6457. \begin_layout Standard
  6458. To test whether the position of a histone mark relative to a gene's
  6459. \begin_inset Flex Glossary Term
  6460. status open
  6461. \begin_layout Plain Layout
  6462. TSS
  6463. \end_layout
  6464. \end_inset
  6465. was important, we looked at the
  6466. \begin_inset Quotes eld
  6467. \end_inset
  6468. landscape
  6469. \begin_inset Quotes erd
  6470. \end_inset
  6471. of
  6472. \begin_inset Flex Glossary Term
  6473. status open
  6474. \begin_layout Plain Layout
  6475. ChIP-seq
  6476. \end_layout
  6477. \end_inset
  6478. read coverage in naïve Day 0 samples within 5 kbp of each gene's
  6479. \begin_inset Flex Glossary Term
  6480. status open
  6481. \begin_layout Plain Layout
  6482. TSS
  6483. \end_layout
  6484. \end_inset
  6485. by binning reads into 500-bp windows tiled across each promoter
  6486. \begin_inset Flex Glossary Term
  6487. status open
  6488. \begin_layout Plain Layout
  6489. logCPM
  6490. \end_layout
  6491. \end_inset
  6492. values were calculated for the bins in each promoter and then the average
  6493. \begin_inset Flex Glossary Term
  6494. status open
  6495. \begin_layout Plain Layout
  6496. logCPM
  6497. \end_layout
  6498. \end_inset
  6499. for each promoter's bins was normalized to zero, such that the values represent
  6500. coverage relative to other regions of the same promoter rather than being
  6501. proportional to absolute read count.
  6502. The promoters were then clustered based on the normalized bin abundances
  6503. using
  6504. \begin_inset Formula $k$
  6505. \end_inset
  6506. -means clustering with
  6507. \begin_inset Formula $K=6$
  6508. \end_inset
  6509. .
  6510. Different values of
  6511. \begin_inset Formula $K$
  6512. \end_inset
  6513. were also tested, but did not substantially change the interpretation of
  6514. the data.
  6515. \end_layout
  6516. \begin_layout Standard
  6517. For H3K4me2, plotting the average bin abundances for each cluster reveals
  6518. a simple pattern (Figure
  6519. \begin_inset CommandInset ref
  6520. LatexCommand ref
  6521. reference "fig:H3K4me2-neighborhood-clusters"
  6522. plural "false"
  6523. caps "false"
  6524. noprefix "false"
  6525. \end_inset
  6526. ): Cluster 5 represents a completely flat promoter coverage profile, likely
  6527. consisting of genes with no H3K4me2 methylation in the promoter.
  6528. All the other clusters represent a continuum of peak positions relative
  6529. to the
  6530. \begin_inset Flex Glossary Term
  6531. status open
  6532. \begin_layout Plain Layout
  6533. TSS
  6534. \end_layout
  6535. \end_inset
  6536. .
  6537. In order from most upstream to most downstream, they are Clusters 6, 4,
  6538. 3, 1, and 2.
  6539. There do not appear to be any clusters representing coverage patterns other
  6540. than lone peaks, such as coverage troughs or double peaks.
  6541. Next, all promoters were plotted in a
  6542. \begin_inset Flex Glossary Term
  6543. status open
  6544. \begin_layout Plain Layout
  6545. PCA
  6546. \end_layout
  6547. \end_inset
  6548. plot based on the same relative bin abundance data, and colored based on
  6549. cluster membership (Figure
  6550. \begin_inset CommandInset ref
  6551. LatexCommand ref
  6552. reference "fig:H3K4me2-neighborhood-pca"
  6553. plural "false"
  6554. caps "false"
  6555. noprefix "false"
  6556. \end_inset
  6557. ).
  6558. The
  6559. \begin_inset Flex Glossary Term
  6560. status open
  6561. \begin_layout Plain Layout
  6562. PCA
  6563. \end_layout
  6564. \end_inset
  6565. plot shows Cluster 5 (the
  6566. \begin_inset Quotes eld
  6567. \end_inset
  6568. no peak
  6569. \begin_inset Quotes erd
  6570. \end_inset
  6571. cluster) at the center, with the other clusters arranged in a counter-clockwise
  6572. arc around it in the order noted above, from most upstream peak to most
  6573. downstream.
  6574. Notably, the
  6575. \begin_inset Quotes eld
  6576. \end_inset
  6577. clusters
  6578. \begin_inset Quotes erd
  6579. \end_inset
  6580. form a single large
  6581. \begin_inset Quotes eld
  6582. \end_inset
  6583. cloud
  6584. \begin_inset Quotes erd
  6585. \end_inset
  6586. with no apparent separation between them, further supporting the conclusion
  6587. that these clusters represent an arbitrary partitioning of a continuous
  6588. distribution of promoter coverage landscapes.
  6589. While the clusters are a useful abstraction that aids in visualization,
  6590. they are ultimately not an accurate representation of the data.
  6591. The continuous nature of the distribution also explains why different values
  6592. of
  6593. \begin_inset Formula $K$
  6594. \end_inset
  6595. led to similar conclusions.
  6596. \end_layout
  6597. \begin_layout Standard
  6598. \begin_inset ERT
  6599. status open
  6600. \begin_layout Plain Layout
  6601. \backslash
  6602. afterpage{
  6603. \end_layout
  6604. \begin_layout Plain Layout
  6605. \backslash
  6606. begin{landscape}
  6607. \end_layout
  6608. \end_inset
  6609. \end_layout
  6610. \begin_layout Standard
  6611. \begin_inset Float figure
  6612. wide false
  6613. sideways false
  6614. status collapsed
  6615. \begin_layout Plain Layout
  6616. \align center
  6617. \begin_inset Float figure
  6618. wide false
  6619. sideways false
  6620. status open
  6621. \begin_layout Plain Layout
  6622. \align center
  6623. \begin_inset Graphics
  6624. filename graphics/CD4-csaw/ChIP-seq/H3K4me2-neighborhood-clusters-CROP.png
  6625. lyxscale 25
  6626. width 30col%
  6627. groupId covprof-subfig
  6628. \end_inset
  6629. \end_layout
  6630. \begin_layout Plain Layout
  6631. \begin_inset Caption Standard
  6632. \begin_layout Plain Layout
  6633. \series bold
  6634. \begin_inset CommandInset label
  6635. LatexCommand label
  6636. name "fig:H3K4me2-neighborhood-clusters"
  6637. \end_inset
  6638. Average relative coverage for each bin in each cluster.
  6639. \end_layout
  6640. \end_inset
  6641. \end_layout
  6642. \end_inset
  6643. \begin_inset space \hfill{}
  6644. \end_inset
  6645. \begin_inset Float figure
  6646. wide false
  6647. sideways false
  6648. status open
  6649. \begin_layout Plain Layout
  6650. \align center
  6651. \begin_inset Graphics
  6652. filename graphics/CD4-csaw/ChIP-seq/H3K4me2-neighborhood-PCA-CROP.png
  6653. lyxscale 25
  6654. width 30col%
  6655. groupId covprof-subfig
  6656. \end_inset
  6657. \end_layout
  6658. \begin_layout Plain Layout
  6659. \begin_inset Caption Standard
  6660. \begin_layout Plain Layout
  6661. \begin_inset CommandInset label
  6662. LatexCommand label
  6663. name "fig:H3K4me2-neighborhood-pca"
  6664. \end_inset
  6665. PCA of relative coverage depth, colored by K-means cluster membership.
  6666. \end_layout
  6667. \end_inset
  6668. \end_layout
  6669. \end_inset
  6670. \begin_inset space \hfill{}
  6671. \end_inset
  6672. \begin_inset Float figure
  6673. wide false
  6674. sideways false
  6675. status open
  6676. \begin_layout Plain Layout
  6677. \align center
  6678. \begin_inset Graphics
  6679. filename graphics/CD4-csaw/ChIP-seq/H3K4me2-neighborhood-expression-CROP.png
  6680. lyxscale 25
  6681. width 30col%
  6682. groupId covprof-subfig
  6683. \end_inset
  6684. \end_layout
  6685. \begin_layout Plain Layout
  6686. \begin_inset Caption Standard
  6687. \begin_layout Plain Layout
  6688. \begin_inset CommandInset label
  6689. LatexCommand label
  6690. name "fig:H3K4me2-neighborhood-expression"
  6691. \end_inset
  6692. Gene expression grouped by promoter coverage clusters.
  6693. \end_layout
  6694. \end_inset
  6695. \end_layout
  6696. \end_inset
  6697. \end_layout
  6698. \begin_layout Plain Layout
  6699. \begin_inset Flex TODO Note (inline)
  6700. status open
  6701. \begin_layout Plain Layout
  6702. Figure font too small
  6703. \end_layout
  6704. \end_inset
  6705. \end_layout
  6706. \begin_layout Plain Layout
  6707. \begin_inset Caption Standard
  6708. \begin_layout Plain Layout
  6709. \begin_inset Argument 1
  6710. status collapsed
  6711. \begin_layout Plain Layout
  6712. K-means clustering of promoter H3K4me2 relative coverage depth in naïve
  6713. day 0 samples.
  6714. \end_layout
  6715. \end_inset
  6716. \begin_inset CommandInset label
  6717. LatexCommand label
  6718. name "fig:H3K4me2-neighborhood"
  6719. \end_inset
  6720. \series bold
  6721. K-means clustering of promoter H3K4me2 relative coverage depth in naïve
  6722. day 0 samples.
  6723. \series default
  6724. H3K4me2 ChIP-seq reads were binned into 500-bp windows tiled across each
  6725. promoter from 5
  6726. \begin_inset space ~
  6727. \end_inset
  6728. kbp upstream to 5
  6729. \begin_inset space ~
  6730. \end_inset
  6731. kbp downstream, and the logCPM values were normalized within each promoter
  6732. to an average of 0, yielding relative coverage depths.
  6733. These were then grouped using K-means clustering with
  6734. \begin_inset Formula $K=6$
  6735. \end_inset
  6736. ,
  6737. \series bold
  6738. \series default
  6739. and the average bin values were plotted for each cluster (a).
  6740. The
  6741. \begin_inset Formula $x$
  6742. \end_inset
  6743. -axis is the genomic coordinate of each bin relative to the the transcription
  6744. start site, and the
  6745. \begin_inset Formula $y$
  6746. \end_inset
  6747. -axis is the mean relative coverage depth of that bin across all promoters
  6748. in the cluster.
  6749. Each line represents the average
  6750. \begin_inset Quotes eld
  6751. \end_inset
  6752. shape
  6753. \begin_inset Quotes erd
  6754. \end_inset
  6755. of the promoter coverage for promoters in that cluster.
  6756. PCA was performed on the same data, and the first two PCs were plotted,
  6757. coloring each point by its K-means cluster identity (b).
  6758. For each cluster, the distribution of gene expression values was plotted
  6759. (c).
  6760. \end_layout
  6761. \end_inset
  6762. \end_layout
  6763. \end_inset
  6764. \end_layout
  6765. \begin_layout Standard
  6766. \begin_inset ERT
  6767. status open
  6768. \begin_layout Plain Layout
  6769. \backslash
  6770. end{landscape}
  6771. \end_layout
  6772. \begin_layout Plain Layout
  6773. }
  6774. \end_layout
  6775. \end_inset
  6776. \end_layout
  6777. \begin_layout Standard
  6778. \begin_inset Flex TODO Note (inline)
  6779. status open
  6780. \begin_layout Plain Layout
  6781. Should have a table of p-values on difference of means between Cluster 5
  6782. and the others.
  6783. \end_layout
  6784. \end_inset
  6785. \end_layout
  6786. \begin_layout Standard
  6787. To investigate the association between relative peak position and gene expressio
  6788. n, we plotted the Naïve Day 0 expression for the genes in each cluster (Figure
  6789. \begin_inset CommandInset ref
  6790. LatexCommand ref
  6791. reference "fig:H3K4me2-neighborhood-expression"
  6792. plural "false"
  6793. caps "false"
  6794. noprefix "false"
  6795. \end_inset
  6796. ).
  6797. Most genes in Cluster 5, the
  6798. \begin_inset Quotes eld
  6799. \end_inset
  6800. no peak
  6801. \begin_inset Quotes erd
  6802. \end_inset
  6803. cluster, have low expression values.
  6804. Taking this as the
  6805. \begin_inset Quotes eld
  6806. \end_inset
  6807. baseline
  6808. \begin_inset Quotes erd
  6809. \end_inset
  6810. distribution when no H3K4me2 methylation is present, we can compare the
  6811. other clusters' distributions to determine which peak positions are associated
  6812. with elevated expression.
  6813. As might be expected, the 3 clusters representing peaks closest to the
  6814. \begin_inset Flex Glossary Term
  6815. status open
  6816. \begin_layout Plain Layout
  6817. TSS
  6818. \end_layout
  6819. \end_inset
  6820. , Clusters 1, 3, and 4, show the highest average expression distributions.
  6821. Specifically, these clusters all have their highest
  6822. \begin_inset Flex Glossary Term
  6823. status open
  6824. \begin_layout Plain Layout
  6825. ChIP-seq
  6826. \end_layout
  6827. \end_inset
  6828. abundance within 1kb of the
  6829. \begin_inset Flex Glossary Term
  6830. status open
  6831. \begin_layout Plain Layout
  6832. TSS
  6833. \end_layout
  6834. \end_inset
  6835. , consistent with the previously determined promoter radius.
  6836. In contrast, cluster 6, which represents peaks several kbp upstream of
  6837. the
  6838. \begin_inset Flex Glossary Term
  6839. status open
  6840. \begin_layout Plain Layout
  6841. TSS
  6842. \end_layout
  6843. \end_inset
  6844. , shows a slightly higher average expression than baseline, while Cluster
  6845. 2, which represents peaks several kbp downstream, doesn't appear to show
  6846. any appreciable difference.
  6847. Interestingly, the cluster with the highest average expression is Cluster
  6848. 1, which represents peaks about 1 kbp downstream of the
  6849. \begin_inset Flex Glossary Term
  6850. status open
  6851. \begin_layout Plain Layout
  6852. TSS
  6853. \end_layout
  6854. \end_inset
  6855. , rather than Cluster 3, which represents peaks centered directly at the
  6856. \begin_inset Flex Glossary Term
  6857. status open
  6858. \begin_layout Plain Layout
  6859. TSS
  6860. \end_layout
  6861. \end_inset
  6862. .
  6863. This suggests that conceptualizing the promoter as a region centered on
  6864. the
  6865. \begin_inset Flex Glossary Term
  6866. status open
  6867. \begin_layout Plain Layout
  6868. TSS
  6869. \end_layout
  6870. \end_inset
  6871. with a certain
  6872. \begin_inset Quotes eld
  6873. \end_inset
  6874. radius
  6875. \begin_inset Quotes erd
  6876. \end_inset
  6877. may be an oversimplification – a peak that is a specific distance from
  6878. the
  6879. \begin_inset Flex Glossary Term
  6880. status open
  6881. \begin_layout Plain Layout
  6882. TSS
  6883. \end_layout
  6884. \end_inset
  6885. may have a different degree of influence depending on whether it is upstream
  6886. or downstream of the
  6887. \begin_inset Flex Glossary Term
  6888. status open
  6889. \begin_layout Plain Layout
  6890. TSS
  6891. \end_layout
  6892. \end_inset
  6893. .
  6894. \end_layout
  6895. \begin_layout Standard
  6896. All observations described above for H3K4me2
  6897. \begin_inset Flex Glossary Term
  6898. status open
  6899. \begin_layout Plain Layout
  6900. ChIP-seq
  6901. \end_layout
  6902. \end_inset
  6903. also appear to hold for H3K4me3 as well (Figure
  6904. \begin_inset CommandInset ref
  6905. LatexCommand ref
  6906. reference "fig:H3K4me3-neighborhood"
  6907. plural "false"
  6908. caps "false"
  6909. noprefix "false"
  6910. \end_inset
  6911. ).
  6912. This is expected, since there is a high correlation between the positions
  6913. where both histone marks occur.
  6914. \end_layout
  6915. \begin_layout Standard
  6916. \begin_inset ERT
  6917. status open
  6918. \begin_layout Plain Layout
  6919. \backslash
  6920. afterpage{
  6921. \end_layout
  6922. \begin_layout Plain Layout
  6923. \backslash
  6924. begin{landscape}
  6925. \end_layout
  6926. \end_inset
  6927. \end_layout
  6928. \begin_layout Standard
  6929. \begin_inset Float figure
  6930. wide false
  6931. sideways false
  6932. status collapsed
  6933. \begin_layout Plain Layout
  6934. \align center
  6935. \begin_inset Float figure
  6936. wide false
  6937. sideways false
  6938. status open
  6939. \begin_layout Plain Layout
  6940. \align center
  6941. \begin_inset Graphics
  6942. filename graphics/CD4-csaw/ChIP-seq/H3K4me3-neighborhood-clusters-CROP.png
  6943. lyxscale 25
  6944. width 30col%
  6945. groupId covprof-subfig
  6946. \end_inset
  6947. \end_layout
  6948. \begin_layout Plain Layout
  6949. \begin_inset Caption Standard
  6950. \begin_layout Plain Layout
  6951. \begin_inset CommandInset label
  6952. LatexCommand label
  6953. name "fig:H3K4me3-neighborhood-clusters"
  6954. \end_inset
  6955. Average relative coverage for each bin in each cluster.
  6956. \end_layout
  6957. \end_inset
  6958. \end_layout
  6959. \end_inset
  6960. \begin_inset space \hfill{}
  6961. \end_inset
  6962. \begin_inset Float figure
  6963. wide false
  6964. sideways false
  6965. status open
  6966. \begin_layout Plain Layout
  6967. \align center
  6968. \begin_inset Graphics
  6969. filename graphics/CD4-csaw/ChIP-seq/H3K4me3-neighborhood-PCA-CROP.png
  6970. lyxscale 25
  6971. width 30col%
  6972. groupId covprof-subfig
  6973. \end_inset
  6974. \end_layout
  6975. \begin_layout Plain Layout
  6976. \begin_inset Caption Standard
  6977. \begin_layout Plain Layout
  6978. \begin_inset CommandInset label
  6979. LatexCommand label
  6980. name "fig:H3K4me3-neighborhood-pca"
  6981. \end_inset
  6982. PCA of relative coverage depth, colored by K-means cluster membership.
  6983. \end_layout
  6984. \end_inset
  6985. \end_layout
  6986. \end_inset
  6987. \begin_inset space \hfill{}
  6988. \end_inset
  6989. \begin_inset Float figure
  6990. wide false
  6991. sideways false
  6992. status open
  6993. \begin_layout Plain Layout
  6994. \align center
  6995. \begin_inset Graphics
  6996. filename graphics/CD4-csaw/ChIP-seq/H3K4me3-neighborhood-expression-CROP.png
  6997. lyxscale 25
  6998. width 30col%
  6999. groupId covprof-subfig
  7000. \end_inset
  7001. \end_layout
  7002. \begin_layout Plain Layout
  7003. \begin_inset Caption Standard
  7004. \begin_layout Plain Layout
  7005. \begin_inset CommandInset label
  7006. LatexCommand label
  7007. name "fig:H3K4me3-neighborhood-expression"
  7008. \end_inset
  7009. Gene expression grouped by promoter coverage clusters.
  7010. \end_layout
  7011. \end_inset
  7012. \end_layout
  7013. \end_inset
  7014. \end_layout
  7015. \begin_layout Plain Layout
  7016. \begin_inset Caption Standard
  7017. \begin_layout Plain Layout
  7018. \begin_inset Argument 1
  7019. status collapsed
  7020. \begin_layout Plain Layout
  7021. K-means clustering of promoter H3K4me3 relative coverage depth in naïve
  7022. day 0 samples.
  7023. \end_layout
  7024. \end_inset
  7025. \begin_inset CommandInset label
  7026. LatexCommand label
  7027. name "fig:H3K4me3-neighborhood"
  7028. \end_inset
  7029. \series bold
  7030. K-means clustering of promoter H3K4me3 relative coverage depth in naïve
  7031. day 0 samples.
  7032. \series default
  7033. H3K4me3 ChIP-seq reads were binned into 500-bp windows tiled across each
  7034. promoter from 5
  7035. \begin_inset space ~
  7036. \end_inset
  7037. kbp upstream to 5
  7038. \begin_inset space ~
  7039. \end_inset
  7040. kbp downstream, and the logCPM values were normalized within each promoter
  7041. to an average of 0, yielding relative coverage depths.
  7042. These were then grouped using K-means clustering with
  7043. \begin_inset Formula $K=6$
  7044. \end_inset
  7045. ,
  7046. \series bold
  7047. \series default
  7048. and the average bin values were plotted for each cluster (a).
  7049. The
  7050. \begin_inset Formula $x$
  7051. \end_inset
  7052. -axis is the genomic coordinate of each bin relative to the the transcription
  7053. start site, and the
  7054. \begin_inset Formula $y$
  7055. \end_inset
  7056. -axis is the mean relative coverage depth of that bin across all promoters
  7057. in the cluster.
  7058. Each line represents the average
  7059. \begin_inset Quotes eld
  7060. \end_inset
  7061. shape
  7062. \begin_inset Quotes erd
  7063. \end_inset
  7064. of the promoter coverage for promoters in that cluster.
  7065. PCA was performed on the same data, and the first two PCs were plotted,
  7066. coloring each point by its K-means cluster identity (b).
  7067. For each cluster, the distribution of gene expression values was plotted
  7068. (c).
  7069. \end_layout
  7070. \end_inset
  7071. \end_layout
  7072. \end_inset
  7073. \end_layout
  7074. \begin_layout Standard
  7075. \begin_inset ERT
  7076. status open
  7077. \begin_layout Plain Layout
  7078. \backslash
  7079. end{landscape}
  7080. \end_layout
  7081. \begin_layout Plain Layout
  7082. }
  7083. \end_layout
  7084. \end_inset
  7085. \end_layout
  7086. \begin_layout Subsection
  7087. Association between resting H3K27me3 promoter coverage landscapes and gene
  7088. expression
  7089. \end_layout
  7090. \begin_layout Standard
  7091. Unlike both H3K4 marks, whose main patterns of variation appear directly
  7092. related to the size and position of a single peak within the promoter,
  7093. the patterns of H3K27me3 methylation in promoters are more complex (Figure
  7094. \begin_inset CommandInset ref
  7095. LatexCommand ref
  7096. reference "fig:H3K27me3-neighborhood"
  7097. plural "false"
  7098. caps "false"
  7099. noprefix "false"
  7100. \end_inset
  7101. ).
  7102. Once again looking at the relative coverage in a 500-bp wide bins in a
  7103. 5kb radius around each
  7104. \begin_inset Flex Glossary Term
  7105. status open
  7106. \begin_layout Plain Layout
  7107. TSS
  7108. \end_layout
  7109. \end_inset
  7110. , promoters were clustered based on the normalized relative coverage values
  7111. in each bin using
  7112. \begin_inset Formula $k$
  7113. \end_inset
  7114. -means clustering with
  7115. \begin_inset Formula $K=6$
  7116. \end_inset
  7117. (Figure
  7118. \begin_inset CommandInset ref
  7119. LatexCommand ref
  7120. reference "fig:H3K27me3-neighborhood-clusters"
  7121. plural "false"
  7122. caps "false"
  7123. noprefix "false"
  7124. \end_inset
  7125. ).
  7126. This time, 3
  7127. \begin_inset Quotes eld
  7128. \end_inset
  7129. axes
  7130. \begin_inset Quotes erd
  7131. \end_inset
  7132. of variation can be observed, each represented by 2 clusters with opposing
  7133. patterns.
  7134. The first axis is greater upstream coverage (Cluster 1) vs.
  7135. greater downstream coverage (Cluster 3); the second axis is the coverage
  7136. at the
  7137. \begin_inset Flex Glossary Term
  7138. status open
  7139. \begin_layout Plain Layout
  7140. TSS
  7141. \end_layout
  7142. \end_inset
  7143. itself: peak (Cluster 4) or trough (Cluster 2); lastly, the third axis
  7144. represents a trough upstream of the
  7145. \begin_inset Flex Glossary Term
  7146. status open
  7147. \begin_layout Plain Layout
  7148. TSS
  7149. \end_layout
  7150. \end_inset
  7151. (Cluster 5) vs.
  7152. downstream of the
  7153. \begin_inset Flex Glossary Term
  7154. status open
  7155. \begin_layout Plain Layout
  7156. TSS
  7157. \end_layout
  7158. \end_inset
  7159. (Cluster 6).
  7160. Referring to these opposing pairs of clusters as axes of variation is justified
  7161. , because they correspond precisely to the first 3
  7162. \begin_inset Flex Glossary Term (pl)
  7163. status open
  7164. \begin_layout Plain Layout
  7165. PC
  7166. \end_layout
  7167. \end_inset
  7168. in the
  7169. \begin_inset Flex Glossary Term
  7170. status open
  7171. \begin_layout Plain Layout
  7172. PCA
  7173. \end_layout
  7174. \end_inset
  7175. plot of the relative coverage values (Figure
  7176. \begin_inset CommandInset ref
  7177. LatexCommand ref
  7178. reference "fig:H3K27me3-neighborhood-pca"
  7179. plural "false"
  7180. caps "false"
  7181. noprefix "false"
  7182. \end_inset
  7183. ).
  7184. The
  7185. \begin_inset Flex Glossary Term
  7186. status open
  7187. \begin_layout Plain Layout
  7188. PCA
  7189. \end_layout
  7190. \end_inset
  7191. plot reveals that as in the case of H3K4me2, all the
  7192. \begin_inset Quotes eld
  7193. \end_inset
  7194. clusters
  7195. \begin_inset Quotes erd
  7196. \end_inset
  7197. are really just sections of a single connected cloud rather than discrete
  7198. clusters.
  7199. The cloud is approximately ellipsoid-shaped, with each PC being an axis
  7200. of the ellipse, and each cluster consisting of a pyramidal section of the
  7201. ellipsoid.
  7202. \end_layout
  7203. \begin_layout Standard
  7204. \begin_inset ERT
  7205. status open
  7206. \begin_layout Plain Layout
  7207. \backslash
  7208. afterpage{
  7209. \end_layout
  7210. \begin_layout Plain Layout
  7211. \backslash
  7212. begin{landscape}
  7213. \end_layout
  7214. \end_inset
  7215. \end_layout
  7216. \begin_layout Standard
  7217. \begin_inset Float figure
  7218. wide false
  7219. sideways false
  7220. status open
  7221. \begin_layout Plain Layout
  7222. \align center
  7223. \begin_inset Float figure
  7224. wide false
  7225. sideways false
  7226. status open
  7227. \begin_layout Plain Layout
  7228. \align center
  7229. \begin_inset Graphics
  7230. filename graphics/CD4-csaw/ChIP-seq/H3K27me3-neighborhood-clusters-CROP.png
  7231. lyxscale 25
  7232. width 30col%
  7233. groupId covprof-subfig
  7234. \end_inset
  7235. \end_layout
  7236. \begin_layout Plain Layout
  7237. \begin_inset Caption Standard
  7238. \begin_layout Plain Layout
  7239. \begin_inset CommandInset label
  7240. LatexCommand label
  7241. name "fig:H3K27me3-neighborhood-clusters"
  7242. \end_inset
  7243. Average relative coverage for each bin in each cluster.
  7244. \end_layout
  7245. \end_inset
  7246. \end_layout
  7247. \end_inset
  7248. \begin_inset space \hfill{}
  7249. \end_inset
  7250. \begin_inset Float figure
  7251. wide false
  7252. sideways false
  7253. status open
  7254. \begin_layout Plain Layout
  7255. \align center
  7256. \begin_inset Graphics
  7257. filename graphics/CD4-csaw/ChIP-seq/H3K27me3-neighborhood-PCA-CROP.png
  7258. lyxscale 25
  7259. width 30col%
  7260. groupId covprof-subfig
  7261. \end_inset
  7262. \end_layout
  7263. \begin_layout Plain Layout
  7264. \begin_inset Caption Standard
  7265. \begin_layout Plain Layout
  7266. \begin_inset CommandInset label
  7267. LatexCommand label
  7268. name "fig:H3K27me3-neighborhood-pca"
  7269. \end_inset
  7270. PCA of relative coverage depth, colored by K-means cluster membership.
  7271. \end_layout
  7272. \end_inset
  7273. \end_layout
  7274. \end_inset
  7275. \begin_inset space \hfill{}
  7276. \end_inset
  7277. \begin_inset Float figure
  7278. wide false
  7279. sideways false
  7280. status open
  7281. \begin_layout Plain Layout
  7282. \align center
  7283. \begin_inset Graphics
  7284. filename graphics/CD4-csaw/ChIP-seq/H3K27me3-neighborhood-expression-CROP.png
  7285. lyxscale 25
  7286. width 30col%
  7287. groupId covprof-subfig
  7288. \end_inset
  7289. \end_layout
  7290. \begin_layout Plain Layout
  7291. \begin_inset Caption Standard
  7292. \begin_layout Plain Layout
  7293. \begin_inset CommandInset label
  7294. LatexCommand label
  7295. name "fig:H3K27me3-neighborhood-expression"
  7296. \end_inset
  7297. Gene expression grouped by promoter coverage clusters.
  7298. \end_layout
  7299. \end_inset
  7300. \end_layout
  7301. \end_inset
  7302. \end_layout
  7303. \begin_layout Plain Layout
  7304. \begin_inset Flex TODO Note (inline)
  7305. status open
  7306. \begin_layout Plain Layout
  7307. Repeated figure legends are kind of an issue here.
  7308. What to do?
  7309. \end_layout
  7310. \end_inset
  7311. \end_layout
  7312. \begin_layout Plain Layout
  7313. \begin_inset Caption Standard
  7314. \begin_layout Plain Layout
  7315. \begin_inset Argument 1
  7316. status collapsed
  7317. \begin_layout Plain Layout
  7318. K-means clustering of promoter H3K27me3 relative coverage depth in naïve
  7319. day 0 samples.
  7320. \end_layout
  7321. \end_inset
  7322. \begin_inset CommandInset label
  7323. LatexCommand label
  7324. name "fig:H3K27me3-neighborhood"
  7325. \end_inset
  7326. \series bold
  7327. K-means clustering of promoter H3K27me3 relative coverage depth in naïve
  7328. day 0 samples.
  7329. \series default
  7330. H3K27me3 ChIP-seq reads were binned into 500-bp windows tiled across each
  7331. promoter from 5
  7332. \begin_inset space ~
  7333. \end_inset
  7334. kbp upstream to 5
  7335. \begin_inset space ~
  7336. \end_inset
  7337. kbp downstream, and the logCPM values were normalized within each promoter
  7338. to an average of 0, yielding relative coverage depths.
  7339. These were then grouped using
  7340. \begin_inset Formula $k$
  7341. \end_inset
  7342. -means clustering with
  7343. \begin_inset Formula $K=6$
  7344. \end_inset
  7345. ,
  7346. \series bold
  7347. \series default
  7348. and the average bin values were plotted for each cluster (a).
  7349. The
  7350. \begin_inset Formula $x$
  7351. \end_inset
  7352. -axis is the genomic coordinate of each bin relative to the the transcription
  7353. start site, and the
  7354. \begin_inset Formula $y$
  7355. \end_inset
  7356. -axis is the mean relative coverage depth of that bin across all promoters
  7357. in the cluster.
  7358. Each line represents the average
  7359. \begin_inset Quotes eld
  7360. \end_inset
  7361. shape
  7362. \begin_inset Quotes erd
  7363. \end_inset
  7364. of the promoter coverage for promoters in that cluster.
  7365. PCA was performed on the same data, and the first two PCs were plotted,
  7366. coloring each point by its K-means cluster identity (b).
  7367. (Note: In (b), Cluster 6 is hidden behind all the other clusters.) For each
  7368. cluster, the distribution of gene expression values was plotted (c).
  7369. \end_layout
  7370. \end_inset
  7371. \end_layout
  7372. \end_inset
  7373. \end_layout
  7374. \begin_layout Standard
  7375. \begin_inset ERT
  7376. status open
  7377. \begin_layout Plain Layout
  7378. \backslash
  7379. end{landscape}
  7380. \end_layout
  7381. \begin_layout Plain Layout
  7382. }
  7383. \end_layout
  7384. \end_inset
  7385. \end_layout
  7386. \begin_layout Standard
  7387. In Figure
  7388. \begin_inset CommandInset ref
  7389. LatexCommand ref
  7390. reference "fig:H3K27me3-neighborhood-expression"
  7391. plural "false"
  7392. caps "false"
  7393. noprefix "false"
  7394. \end_inset
  7395. , we can see that Clusters 1 and 2 are the only clusters with higher gene
  7396. expression than the others.
  7397. For Cluster 2, this is expected, since this cluster represents genes with
  7398. depletion of H3K27me3 near the promoter.
  7399. Hence, elevated expression in cluster 2 is consistent with the conventional
  7400. view of H3K27me3 as a deactivating mark.
  7401. However, Cluster 1, the cluster with the most elevated gene expression,
  7402. represents genes with elevated coverage upstream of the
  7403. \begin_inset Flex Glossary Term
  7404. status open
  7405. \begin_layout Plain Layout
  7406. TSS
  7407. \end_layout
  7408. \end_inset
  7409. , or equivalently, decreased coverage downstream, inside the gene body.
  7410. The opposite pattern, in which H3K27me3 is more abundant within the gene
  7411. body and less abundance in the upstream promoter region, does not show
  7412. any elevation in gene expression.
  7413. As with H3K4me2, this shows that the location of H3K27 trimethylation relative
  7414. to the
  7415. \begin_inset Flex Glossary Term
  7416. status open
  7417. \begin_layout Plain Layout
  7418. TSS
  7419. \end_layout
  7420. \end_inset
  7421. is potentially an important factor beyond simple proximity.
  7422. \end_layout
  7423. \begin_layout Standard
  7424. \begin_inset Note Note
  7425. status open
  7426. \begin_layout Plain Layout
  7427. \begin_inset Flex TODO Note (inline)
  7428. status open
  7429. \begin_layout Plain Layout
  7430. Show the figures where the negative result ended this line of inquiry.
  7431. I need to debug some errors resulting from an R upgrade to do this.
  7432. \end_layout
  7433. \end_inset
  7434. \end_layout
  7435. \begin_layout Subsection
  7436. Defined pattern analysis
  7437. \end_layout
  7438. \begin_layout Plain Layout
  7439. \begin_inset Flex TODO Note (inline)
  7440. status open
  7441. \begin_layout Plain Layout
  7442. This was where I defined interesting expression patterns and then looked
  7443. at initial relative promoter coverage for each expression pattern.
  7444. Negative result.
  7445. I forgot about this until recently.
  7446. Worth including? Remember to also write methods.
  7447. \end_layout
  7448. \end_inset
  7449. \end_layout
  7450. \begin_layout Subsection
  7451. Promoter CpG islands?
  7452. \end_layout
  7453. \begin_layout Plain Layout
  7454. \begin_inset Flex TODO Note (inline)
  7455. status open
  7456. \begin_layout Plain Layout
  7457. I forgot until recently about the work I did on this.
  7458. Worth including? Remember to also write methods.
  7459. \end_layout
  7460. \end_inset
  7461. \end_layout
  7462. \end_inset
  7463. \end_layout
  7464. \begin_layout Section
  7465. Discussion
  7466. \end_layout
  7467. \begin_layout Standard
  7468. \begin_inset Flex TODO Note (inline)
  7469. status open
  7470. \begin_layout Plain Layout
  7471. Write better section headers
  7472. \end_layout
  7473. \end_inset
  7474. \end_layout
  7475. \begin_layout Subsection
  7476. Each histone mark's
  7477. \begin_inset Quotes eld
  7478. \end_inset
  7479. effective promoter extent
  7480. \begin_inset Quotes erd
  7481. \end_inset
  7482. must be determined empirically
  7483. \end_layout
  7484. \begin_layout Standard
  7485. Figure
  7486. \begin_inset CommandInset ref
  7487. LatexCommand ref
  7488. reference "fig:near-promoter-peak-enrich"
  7489. plural "false"
  7490. caps "false"
  7491. noprefix "false"
  7492. \end_inset
  7493. shows that H3K4me2, H3K4me3, and H3K27me3 are all enriched near promoters,
  7494. relative to the rest of the genome, consistent with their conventionally
  7495. understood role in regulating gene transcription.
  7496. Interestingly, the radius within this enrichment occurs is not the same
  7497. for each histone mark.
  7498. H3K4me2 and H3K4me3 are enriched within a 1
  7499. \begin_inset space ~
  7500. \end_inset
  7501. kbp radius, while H3K27me3 is enriched within 2.5
  7502. \begin_inset space ~
  7503. \end_inset
  7504. kbp.
  7505. Notably, the determined promoter radius was consistent across all experimental
  7506. conditions, varying only between different histone marks.
  7507. This suggests that the conventional
  7508. \begin_inset Quotes eld
  7509. \end_inset
  7510. one size fits all
  7511. \begin_inset Quotes erd
  7512. \end_inset
  7513. approach of defining a single promoter region for each gene (or each
  7514. \begin_inset Flex Glossary Term
  7515. status open
  7516. \begin_layout Plain Layout
  7517. TSS
  7518. \end_layout
  7519. \end_inset
  7520. ) and using that same promoter region for analyzing all types of genomic
  7521. data within an experiment may not be appropriate, and a better approach
  7522. may be to use a separate promoter radius for each kind of data, with each
  7523. radius being derived from the data itself.
  7524. Furthermore, the apparent asymmetry of upstream and downstream promoter
  7525. histone modification with respect to gene expression, seen in Figures
  7526. \begin_inset CommandInset ref
  7527. LatexCommand ref
  7528. reference "fig:H3K4me2-neighborhood"
  7529. plural "false"
  7530. caps "false"
  7531. noprefix "false"
  7532. \end_inset
  7533. ,
  7534. \begin_inset CommandInset ref
  7535. LatexCommand ref
  7536. reference "fig:H3K4me3-neighborhood"
  7537. plural "false"
  7538. caps "false"
  7539. noprefix "false"
  7540. \end_inset
  7541. , and
  7542. \begin_inset CommandInset ref
  7543. LatexCommand ref
  7544. reference "fig:H3K27me3-neighborhood"
  7545. plural "false"
  7546. caps "false"
  7547. noprefix "false"
  7548. \end_inset
  7549. , shows that even the concept of a promoter
  7550. \begin_inset Quotes eld
  7551. \end_inset
  7552. radius
  7553. \begin_inset Quotes erd
  7554. \end_inset
  7555. is likely an oversimplification.
  7556. At a minimum, nearby enrichment of peaks should be evaluated separately
  7557. for both upstream and downstream peaks, and an appropriate
  7558. \begin_inset Quotes eld
  7559. \end_inset
  7560. radius
  7561. \begin_inset Quotes erd
  7562. \end_inset
  7563. should be selected for each direction.
  7564. \end_layout
  7565. \begin_layout Standard
  7566. \begin_inset Flex TODO Note (inline)
  7567. status open
  7568. \begin_layout Plain Layout
  7569. Sarah: I would have to search the literature, but I believe this has been
  7570. observed before.
  7571. The position relative to the TSS likely has to do with recruitment of the
  7572. transcriptional machinery and the space required for that.
  7573. \end_layout
  7574. \end_inset
  7575. \end_layout
  7576. \begin_layout Standard
  7577. Figures
  7578. \begin_inset CommandInset ref
  7579. LatexCommand ref
  7580. reference "fig:H3K4me2-neighborhood"
  7581. plural "false"
  7582. caps "false"
  7583. noprefix "false"
  7584. \end_inset
  7585. and
  7586. \begin_inset CommandInset ref
  7587. LatexCommand ref
  7588. reference "fig:H3K4me3-neighborhood"
  7589. plural "false"
  7590. caps "false"
  7591. noprefix "false"
  7592. \end_inset
  7593. show that the determined promoter radius of 1
  7594. \begin_inset space ~
  7595. \end_inset
  7596. kbp is approximately consistent with the distance from the
  7597. \begin_inset Flex Glossary Term
  7598. status open
  7599. \begin_layout Plain Layout
  7600. TSS
  7601. \end_layout
  7602. \end_inset
  7603. at which enrichment of H3K4 methylation correlates with increased expression,
  7604. showing that this radius, which was determined by a simple analysis of
  7605. measuring the distance from each
  7606. \begin_inset Flex Glossary Term
  7607. status open
  7608. \begin_layout Plain Layout
  7609. TSS
  7610. \end_layout
  7611. \end_inset
  7612. to the nearest peak, also has functional significance.
  7613. For H3K27me3, the correlation between histone modification near the promoter
  7614. and gene expression is more complex, involving non-peak variations such
  7615. as troughs in coverage at the
  7616. \begin_inset Flex Glossary Term
  7617. status open
  7618. \begin_layout Plain Layout
  7619. TSS
  7620. \end_layout
  7621. \end_inset
  7622. and asymmetric coverage upstream and downstream, so it is difficult in
  7623. this case to evaluate whether the 2.5
  7624. \begin_inset space ~
  7625. \end_inset
  7626. kbp radius determined from TSS-to-peak distances is functionally significant.
  7627. However, the two patterns of coverage associated with elevated expression
  7628. levels both have interesting features within this radius.
  7629. \end_layout
  7630. \begin_layout Subsection
  7631. Day 14 convergence is consistent with naïve-to-memory differentiation
  7632. \end_layout
  7633. \begin_layout Standard
  7634. \begin_inset Flex TODO Note (inline)
  7635. status open
  7636. \begin_layout Plain Layout
  7637. Look up some more references for these histone marks being involved in memory
  7638. differentiation.
  7639. (Ask Sarah)
  7640. \end_layout
  7641. \end_inset
  7642. \end_layout
  7643. \begin_layout Standard
  7644. We observed that all 3 histone marks and the gene expression data all exhibit
  7645. evidence of convergence in abundance between naïve and memory cells by
  7646. day 14 after activation (Figure
  7647. \begin_inset CommandInset ref
  7648. LatexCommand ref
  7649. reference "fig:PCoA-promoters"
  7650. plural "false"
  7651. caps "false"
  7652. noprefix "false"
  7653. \end_inset
  7654. , Table
  7655. \begin_inset CommandInset ref
  7656. LatexCommand ref
  7657. reference "tab:Number-signif-promoters"
  7658. plural "false"
  7659. caps "false"
  7660. noprefix "false"
  7661. \end_inset
  7662. ).
  7663. The
  7664. \begin_inset Flex Glossary Term
  7665. status open
  7666. \begin_layout Plain Layout
  7667. MOFA
  7668. \end_layout
  7669. \end_inset
  7670. \begin_inset Flex Glossary Term
  7671. status open
  7672. \begin_layout Plain Layout
  7673. LF
  7674. \end_layout
  7675. \end_inset
  7676. scatter plots (Figure
  7677. \begin_inset CommandInset ref
  7678. LatexCommand ref
  7679. reference "fig:mofa-lf-scatter"
  7680. plural "false"
  7681. caps "false"
  7682. noprefix "false"
  7683. \end_inset
  7684. ) show that this pattern of convergence is captured in
  7685. \begin_inset Flex Glossary Term
  7686. status open
  7687. \begin_layout Plain Layout
  7688. LF
  7689. \end_layout
  7690. \end_inset
  7691. 5.
  7692. Like all the
  7693. \begin_inset Flex Glossary Term (pl)
  7694. status open
  7695. \begin_layout Plain Layout
  7696. LF
  7697. \end_layout
  7698. \end_inset
  7699. in this plot, this factor explains a substantial portion of the variance
  7700. in all 4 data sets, indicating a coordinated pattern of variation shared
  7701. across all histone marks and gene expression.
  7702. This is consistent with the expectation that any naïve CD4
  7703. \begin_inset Formula $^{+}$
  7704. \end_inset
  7705. T-cells remaining at day 14 should have differentiated into memory cells
  7706. by that time, and should therefore have a genomic and epigenomic state
  7707. similar to memory cells.
  7708. This convergence is evidence that these histone marks all play an important
  7709. role in the naïve-to-memory differentiation process.
  7710. A histone mark that was not involved in naïve-to-memory differentiation
  7711. would not be expected to converge in this way after activation.
  7712. \end_layout
  7713. \begin_layout Standard
  7714. In H3K4me2, H3K4me3, and
  7715. \begin_inset Flex Glossary Term
  7716. status open
  7717. \begin_layout Plain Layout
  7718. RNA-seq
  7719. \end_layout
  7720. \end_inset
  7721. , this convergence appears to be in progress already by Day 5, shown by
  7722. the smaller distance between naïve and memory cells at day 5 along the
  7723. \begin_inset Formula $y$
  7724. \end_inset
  7725. -axes in Figures
  7726. \begin_inset CommandInset ref
  7727. LatexCommand ref
  7728. reference "fig:PCoA-H3K4me2-prom"
  7729. plural "false"
  7730. caps "false"
  7731. noprefix "false"
  7732. \end_inset
  7733. ,
  7734. \begin_inset CommandInset ref
  7735. LatexCommand ref
  7736. reference "fig:PCoA-H3K4me3-prom"
  7737. plural "false"
  7738. caps "false"
  7739. noprefix "false"
  7740. \end_inset
  7741. , and
  7742. \begin_inset CommandInset ref
  7743. LatexCommand ref
  7744. reference "fig:RNA-PCA-group"
  7745. plural "false"
  7746. caps "false"
  7747. noprefix "false"
  7748. \end_inset
  7749. .
  7750. This agrees with the model proposed by Sarah Lamere based on an prior analysis
  7751. of the same data, shown in Figure
  7752. \begin_inset CommandInset ref
  7753. LatexCommand ref
  7754. reference "fig:Lamere2016-Fig8"
  7755. plural "false"
  7756. caps "false"
  7757. noprefix "false"
  7758. \end_inset
  7759. , which shows the pattern of H3K4 methylation and expression for naïve cells
  7760. and memory cells converging at day 5.
  7761. This model was developed without the benefit of the
  7762. \begin_inset Flex Glossary Term
  7763. status open
  7764. \begin_layout Plain Layout
  7765. PCoA
  7766. \end_layout
  7767. \end_inset
  7768. plots in Figure
  7769. \begin_inset CommandInset ref
  7770. LatexCommand ref
  7771. reference "fig:PCoA-promoters"
  7772. plural "false"
  7773. caps "false"
  7774. noprefix "false"
  7775. \end_inset
  7776. , which have been corrected for confounding factors by ComBat and
  7777. \begin_inset Flex Glossary Term
  7778. status open
  7779. \begin_layout Plain Layout
  7780. SVA
  7781. \end_layout
  7782. \end_inset
  7783. .
  7784. This shows that proper batch correction assists in extracting meaningful
  7785. patterns in the data while eliminating systematic sources of irrelevant
  7786. variation in the data, allowing simple automated procedures like
  7787. \begin_inset Flex Glossary Term
  7788. status open
  7789. \begin_layout Plain Layout
  7790. PCoA
  7791. \end_layout
  7792. \end_inset
  7793. to reveal interesting behaviors in the data that were previously only detectabl
  7794. e by a detailed manual analysis.
  7795. While the ideal comparison to demonstrate this convergence would be naïve
  7796. cells at day 14 to memory cells at day 0, this is not feasible in this
  7797. experimental system, since neither naïve nor memory cells are able to fully
  7798. return to their pre-activation state, as shown by the lack of overlap between
  7799. days 0 and 14 for either naïve or memory cells in Figure
  7800. \begin_inset CommandInset ref
  7801. LatexCommand ref
  7802. reference "fig:PCoA-promoters"
  7803. plural "false"
  7804. caps "false"
  7805. noprefix "false"
  7806. \end_inset
  7807. .
  7808. \end_layout
  7809. \begin_layout Standard
  7810. \begin_inset Float figure
  7811. wide false
  7812. sideways false
  7813. status collapsed
  7814. \begin_layout Plain Layout
  7815. \align center
  7816. \begin_inset Graphics
  7817. filename graphics/CD4-csaw/LaMere2016_fig8.pdf
  7818. lyxscale 50
  7819. width 100col%
  7820. groupId colfullwidth
  7821. \end_inset
  7822. \end_layout
  7823. \begin_layout Plain Layout
  7824. \begin_inset Caption Standard
  7825. \begin_layout Plain Layout
  7826. \begin_inset Argument 1
  7827. status collapsed
  7828. \begin_layout Plain Layout
  7829. Lamere 2016 Figure 8 “Model for the role of H3K4 methylation during CD4
  7830. \begin_inset Formula $^{+}$
  7831. \end_inset
  7832. T-cell activation.
  7833. \begin_inset Quotes erd
  7834. \end_inset
  7835. \end_layout
  7836. \end_inset
  7837. \begin_inset CommandInset label
  7838. LatexCommand label
  7839. name "fig:Lamere2016-Fig8"
  7840. \end_inset
  7841. \series bold
  7842. Lamere 2016 Figure 8
  7843. \begin_inset CommandInset citation
  7844. LatexCommand cite
  7845. key "LaMere2016"
  7846. literal "false"
  7847. \end_inset
  7848. ,
  7849. \begin_inset Quotes eld
  7850. \end_inset
  7851. Model for the role of H3K4 methylation during CD4
  7852. \begin_inset Formula $\mathbf{^{+}}$
  7853. \end_inset
  7854. T-cell activation.
  7855. \begin_inset Quotes erd
  7856. \end_inset
  7857. \series default
  7858. (Reproduced with permission.)
  7859. \end_layout
  7860. \end_inset
  7861. \end_layout
  7862. \end_inset
  7863. \end_layout
  7864. \begin_layout Subsection
  7865. The location of histone modifications within the promoter is important
  7866. \end_layout
  7867. \begin_layout Standard
  7868. When looking at patterns in the relative coverage of each histone mark near
  7869. the
  7870. \begin_inset Flex Glossary Term
  7871. status open
  7872. \begin_layout Plain Layout
  7873. TSS
  7874. \end_layout
  7875. \end_inset
  7876. of each gene, several interesting patterns were apparent.
  7877. For H3K4me2 and H3K4me3, the pattern was straightforward: the consistent
  7878. pattern across all promoters was a single peak a few kbp wide, with the
  7879. main axis of variation being the position of this peak relative to the
  7880. \begin_inset Flex Glossary Term
  7881. status open
  7882. \begin_layout Plain Layout
  7883. TSS
  7884. \end_layout
  7885. \end_inset
  7886. (Figures
  7887. \begin_inset CommandInset ref
  7888. LatexCommand ref
  7889. reference "fig:H3K4me2-neighborhood"
  7890. plural "false"
  7891. caps "false"
  7892. noprefix "false"
  7893. \end_inset
  7894. &
  7895. \begin_inset CommandInset ref
  7896. LatexCommand ref
  7897. reference "fig:H3K4me3-neighborhood"
  7898. plural "false"
  7899. caps "false"
  7900. noprefix "false"
  7901. \end_inset
  7902. ).
  7903. There were no obvious
  7904. \begin_inset Quotes eld
  7905. \end_inset
  7906. preferred
  7907. \begin_inset Quotes erd
  7908. \end_inset
  7909. positions, but rather a continuous distribution of relative positions ranging
  7910. all across the promoter region.
  7911. The association with gene expression was also straightforward: peaks closer
  7912. to the
  7913. \begin_inset Flex Glossary Term
  7914. status open
  7915. \begin_layout Plain Layout
  7916. TSS
  7917. \end_layout
  7918. \end_inset
  7919. were more strongly associated with elevated gene expression.
  7920. Coverage downstream of the
  7921. \begin_inset Flex Glossary Term
  7922. status open
  7923. \begin_layout Plain Layout
  7924. TSS
  7925. \end_layout
  7926. \end_inset
  7927. appears to be more strongly associated with elevated expression than coverage
  7928. at the same distance upstream, indicating that the
  7929. \begin_inset Quotes eld
  7930. \end_inset
  7931. effective promoter region
  7932. \begin_inset Quotes erd
  7933. \end_inset
  7934. for H3K4me2 and H3K4me3 may be centered downstream of the
  7935. \begin_inset Flex Glossary Term
  7936. status open
  7937. \begin_layout Plain Layout
  7938. TSS
  7939. \end_layout
  7940. \end_inset
  7941. .
  7942. \end_layout
  7943. \begin_layout Standard
  7944. The relative promoter coverage for H3K27me3 had a more complex pattern,
  7945. with two specific patterns of promoter coverage associated with elevated
  7946. expression: a sharp depletion of H3K27me3 around the
  7947. \begin_inset Flex Glossary Term
  7948. status open
  7949. \begin_layout Plain Layout
  7950. TSS
  7951. \end_layout
  7952. \end_inset
  7953. relative to the surrounding area, and a depletion of H3K27me3 downstream
  7954. of the
  7955. \begin_inset Flex Glossary Term
  7956. status open
  7957. \begin_layout Plain Layout
  7958. TSS
  7959. \end_layout
  7960. \end_inset
  7961. relative to upstream (Figure
  7962. \begin_inset CommandInset ref
  7963. LatexCommand ref
  7964. reference "fig:H3K27me3-neighborhood"
  7965. plural "false"
  7966. caps "false"
  7967. noprefix "false"
  7968. \end_inset
  7969. ).
  7970. A previous study found that H3K27me3 depletion within the gene body was
  7971. associated with elevated gene expression in 4 different cell types in mice
  7972. \begin_inset CommandInset citation
  7973. LatexCommand cite
  7974. key "Young2011"
  7975. literal "false"
  7976. \end_inset
  7977. .
  7978. This is consistent with the second pattern described here.
  7979. This study also reported that a spike in coverage at the
  7980. \begin_inset Flex Glossary Term
  7981. status open
  7982. \begin_layout Plain Layout
  7983. TSS
  7984. \end_layout
  7985. \end_inset
  7986. was associated with
  7987. \emph on
  7988. lower
  7989. \emph default
  7990. expression, which is indirectly consistent with the first pattern described
  7991. here, in the sense that it associates lower H3K27me3 levels near the
  7992. \begin_inset Flex Glossary Term
  7993. status open
  7994. \begin_layout Plain Layout
  7995. TSS
  7996. \end_layout
  7997. \end_inset
  7998. with higher expression.
  7999. \end_layout
  8000. \begin_layout Subsection
  8001. A reproducible workflow aids in analysis
  8002. \end_layout
  8003. \begin_layout Standard
  8004. The analyses described in this chapter were organized into a reproducible
  8005. workflow using the Snakemake workflow management system
  8006. \begin_inset CommandInset citation
  8007. LatexCommand cite
  8008. key "Koster2012"
  8009. literal "false"
  8010. \end_inset
  8011. .
  8012. As shown in Figure
  8013. \begin_inset CommandInset ref
  8014. LatexCommand ref
  8015. reference "fig:rulegraph"
  8016. plural "false"
  8017. caps "false"
  8018. noprefix "false"
  8019. \end_inset
  8020. , the workflow includes many steps with complex dependencies between them.
  8021. For example, the step that counts the number of
  8022. \begin_inset Flex Glossary Term
  8023. status open
  8024. \begin_layout Plain Layout
  8025. ChIP-seq
  8026. \end_layout
  8027. \end_inset
  8028. reads in 500
  8029. \begin_inset space ~
  8030. \end_inset
  8031. bp windows in each promoter (the starting point for Figures
  8032. \begin_inset CommandInset ref
  8033. LatexCommand ref
  8034. reference "fig:H3K4me2-neighborhood"
  8035. plural "false"
  8036. caps "false"
  8037. noprefix "false"
  8038. \end_inset
  8039. ,
  8040. \begin_inset CommandInset ref
  8041. LatexCommand ref
  8042. reference "fig:H3K4me3-neighborhood"
  8043. plural "false"
  8044. caps "false"
  8045. noprefix "false"
  8046. \end_inset
  8047. , and
  8048. \begin_inset CommandInset ref
  8049. LatexCommand ref
  8050. reference "fig:H3K27me3-neighborhood"
  8051. plural "false"
  8052. caps "false"
  8053. noprefix "false"
  8054. \end_inset
  8055. ), named
  8056. \begin_inset Flex Code
  8057. status open
  8058. \begin_layout Plain Layout
  8059. chipseq_count_tss_neighborhoods
  8060. \end_layout
  8061. \end_inset
  8062. , depends on the
  8063. \begin_inset Flex Glossary Term
  8064. status open
  8065. \begin_layout Plain Layout
  8066. RNA-seq
  8067. \end_layout
  8068. \end_inset
  8069. abundance estimates in order to select the most-used
  8070. \begin_inset Flex Glossary Term
  8071. status open
  8072. \begin_layout Plain Layout
  8073. TSS
  8074. \end_layout
  8075. \end_inset
  8076. for each gene, the aligned
  8077. \begin_inset Flex Glossary Term
  8078. status open
  8079. \begin_layout Plain Layout
  8080. ChIP-seq
  8081. \end_layout
  8082. \end_inset
  8083. reads, the index for those reads, and the blacklist of regions to be excluded
  8084. from
  8085. \begin_inset Flex Glossary Term
  8086. status open
  8087. \begin_layout Plain Layout
  8088. ChIP-seq
  8089. \end_layout
  8090. \end_inset
  8091. analysis.
  8092. Each step declares its inputs and outputs, and Snakemake uses these to
  8093. determine the dependencies between steps.
  8094. Each step is marked as depending on all the steps whose outputs match its
  8095. inputs, generating the workflow graph in Figure
  8096. \begin_inset CommandInset ref
  8097. LatexCommand ref
  8098. reference "fig:rulegraph"
  8099. plural "false"
  8100. caps "false"
  8101. noprefix "false"
  8102. \end_inset
  8103. , which Snakemake uses to determine order in which to execute each step
  8104. so that each step is executed only after all of the steps it depends on
  8105. have completed, thereby automating the entire workflow from start to finish.
  8106. \end_layout
  8107. \begin_layout Standard
  8108. \begin_inset ERT
  8109. status open
  8110. \begin_layout Plain Layout
  8111. \backslash
  8112. afterpage{
  8113. \end_layout
  8114. \begin_layout Plain Layout
  8115. \backslash
  8116. begin{landscape}
  8117. \end_layout
  8118. \end_inset
  8119. \end_layout
  8120. \begin_layout Standard
  8121. \begin_inset Float figure
  8122. wide false
  8123. sideways false
  8124. status collapsed
  8125. \begin_layout Plain Layout
  8126. \align center
  8127. \begin_inset Graphics
  8128. filename graphics/CD4-csaw/rulegraphs/rulegraph-all.pdf
  8129. lyxscale 50
  8130. width 100col%
  8131. height 95theight%
  8132. \end_inset
  8133. \end_layout
  8134. \begin_layout Plain Layout
  8135. \begin_inset Caption Standard
  8136. \begin_layout Plain Layout
  8137. \begin_inset Argument 1
  8138. status collapsed
  8139. \begin_layout Plain Layout
  8140. Dependency graph of steps in reproducible workflow.
  8141. \end_layout
  8142. \end_inset
  8143. \begin_inset CommandInset label
  8144. LatexCommand label
  8145. name "fig:rulegraph"
  8146. \end_inset
  8147. \series bold
  8148. Dependency graph of steps in reproducible workflow.
  8149. \series default
  8150. The analysis flows from left to right.
  8151. Arrows indicate which analysis steps depend on the output of other steps.
  8152. \end_layout
  8153. \end_inset
  8154. \end_layout
  8155. \end_inset
  8156. \end_layout
  8157. \begin_layout Standard
  8158. \begin_inset ERT
  8159. status open
  8160. \begin_layout Plain Layout
  8161. \backslash
  8162. end{landscape}
  8163. \end_layout
  8164. \begin_layout Plain Layout
  8165. }
  8166. \end_layout
  8167. \end_inset
  8168. \end_layout
  8169. \begin_layout Standard
  8170. In addition to simply making it easier to organize the steps in the analysis,
  8171. structuring the analysis as a workflow allowed for some analysis strategies
  8172. that would not have been practical otherwise.
  8173. For example, 5 different
  8174. \begin_inset Flex Glossary Term
  8175. status open
  8176. \begin_layout Plain Layout
  8177. RNA-seq
  8178. \end_layout
  8179. \end_inset
  8180. quantification methods were tested against two different reference transcriptom
  8181. e annotations for a total of 10 different quantifications of the same
  8182. \begin_inset Flex Glossary Term
  8183. status open
  8184. \begin_layout Plain Layout
  8185. RNA-seq
  8186. \end_layout
  8187. \end_inset
  8188. data.
  8189. These were then compared against each other in the exploratory data analysis
  8190. step, to determine that the results were not very sensitive to either the
  8191. choice of quantification method or the choice of annotation.
  8192. This was possible with a single script for the exploratory data analysis,
  8193. because Snakemake was able to automate running this script for every combinatio
  8194. n of method and reference.
  8195. In a similar manner, two different peak calling methods were tested against
  8196. each other, and in this case it was determined that
  8197. \begin_inset Flex Glossary Term
  8198. status open
  8199. \begin_layout Plain Layout
  8200. SICER
  8201. \end_layout
  8202. \end_inset
  8203. was unambiguously superior to
  8204. \begin_inset Flex Glossary Term
  8205. status open
  8206. \begin_layout Plain Layout
  8207. MACS
  8208. \end_layout
  8209. \end_inset
  8210. for all histone marks studied.
  8211. By enabling these types of comparisons, structuring the analysis as an
  8212. automated workflow allowed important analysis decisions to be made in a
  8213. data-driven way, by running every reasonable option through the downstream
  8214. steps, seeing the consequences of choosing each option, and deciding accordingl
  8215. y.
  8216. \end_layout
  8217. \begin_layout Standard
  8218. \begin_inset Note Note
  8219. status open
  8220. \begin_layout Subsection
  8221. Data quality issues limit conclusions
  8222. \end_layout
  8223. \begin_layout Plain Layout
  8224. \begin_inset Flex TODO Note (inline)
  8225. status open
  8226. \begin_layout Plain Layout
  8227. Is this needed?
  8228. \end_layout
  8229. \end_inset
  8230. \end_layout
  8231. \end_inset
  8232. \end_layout
  8233. \begin_layout Section
  8234. Future Directions
  8235. \end_layout
  8236. \begin_layout Standard
  8237. The analysis of
  8238. \begin_inset Flex Glossary Term
  8239. status open
  8240. \begin_layout Plain Layout
  8241. RNA-seq
  8242. \end_layout
  8243. \end_inset
  8244. and
  8245. \begin_inset Flex Glossary Term
  8246. status open
  8247. \begin_layout Plain Layout
  8248. ChIP-seq
  8249. \end_layout
  8250. \end_inset
  8251. in CD4
  8252. \begin_inset Formula $^{+}$
  8253. \end_inset
  8254. T-cells in Chapter 2 is in many ways a preliminary study that suggests
  8255. a multitude of new avenues of investigation.
  8256. Here we consider a selection of such avenues.
  8257. \end_layout
  8258. \begin_layout Subsection
  8259. Previous negative results
  8260. \end_layout
  8261. \begin_layout Standard
  8262. Two additional analyses were conducted beyond those reported in the results.
  8263. First, we searched for evidence that the presence or absence of a
  8264. \begin_inset Flex Glossary Term
  8265. status open
  8266. \begin_layout Plain Layout
  8267. CpGi
  8268. \end_layout
  8269. \end_inset
  8270. in the promoter was correlated with increases or decreases in gene expression
  8271. or any histone mark in any of the tested contrasts.
  8272. Second, we searched for evidence that the relative
  8273. \begin_inset Flex Glossary Term
  8274. status open
  8275. \begin_layout Plain Layout
  8276. ChIP-seq
  8277. \end_layout
  8278. \end_inset
  8279. coverage profiles prior to activations could predict the change in expression
  8280. of a gene after activation.
  8281. Neither analysis turned up any clear positive results.
  8282. \end_layout
  8283. \begin_layout Subsection
  8284. Improve on the idea of an effective promoter radius
  8285. \end_layout
  8286. \begin_layout Standard
  8287. This study introduced the concept of an
  8288. \begin_inset Quotes eld
  8289. \end_inset
  8290. effective promoter radius
  8291. \begin_inset Quotes erd
  8292. \end_inset
  8293. specific to each histone mark based on distance from the
  8294. \begin_inset Flex Glossary Term
  8295. status open
  8296. \begin_layout Plain Layout
  8297. TSS
  8298. \end_layout
  8299. \end_inset
  8300. within which an excess of peaks was called for that mark.
  8301. This concept was then used to guide further analyses throughout the study.
  8302. However, while the effective promoter radius was useful in those analyses,
  8303. it is both limited in theory and shown in practice to be a possible oversimplif
  8304. ication.
  8305. First, the effective promoter radii used in this study were chosen based
  8306. on manual inspection of the TSS-to-peak distance distributions in Figure
  8307. \begin_inset CommandInset ref
  8308. LatexCommand ref
  8309. reference "fig:near-promoter-peak-enrich"
  8310. plural "false"
  8311. caps "false"
  8312. noprefix "false"
  8313. \end_inset
  8314. , selecting round numbers of analyst convenience (Table
  8315. \begin_inset CommandInset ref
  8316. LatexCommand ref
  8317. reference "tab:effective-promoter-radius"
  8318. plural "false"
  8319. caps "false"
  8320. noprefix "false"
  8321. \end_inset
  8322. ).
  8323. It would be better to define an algorithm that selects a more precise radius
  8324. based on the features of the graph.
  8325. One possible way to do this would be to randomly rearrange the called peaks
  8326. throughout the genome many (while preserving the distribution of peak widths)
  8327. and re-generate the same plot as in Figure
  8328. \begin_inset CommandInset ref
  8329. LatexCommand ref
  8330. reference "fig:near-promoter-peak-enrich"
  8331. plural "false"
  8332. caps "false"
  8333. noprefix "false"
  8334. \end_inset
  8335. .
  8336. This would yield a better
  8337. \begin_inset Quotes eld
  8338. \end_inset
  8339. background
  8340. \begin_inset Quotes erd
  8341. \end_inset
  8342. distribution that demonstrates the degree of near-TSS enrichment that would
  8343. be expected by random chance.
  8344. The effective promoter radius could be defined as the point where the true
  8345. distribution diverges from the randomized background distribution.
  8346. \end_layout
  8347. \begin_layout Standard
  8348. Furthermore, the above definition of effective promoter radius has the significa
  8349. nt limitation of being based on the peak calling method.
  8350. It is thus very sensitive to the choice of peak caller and significance
  8351. threshold for calling peaks, as well as the degree of saturation in the
  8352. sequencing.
  8353. Calling peaks from
  8354. \begin_inset Flex Glossary Term
  8355. status open
  8356. \begin_layout Plain Layout
  8357. ChIP-seq
  8358. \end_layout
  8359. \end_inset
  8360. samples with insufficient coverage depth, with the wrong peak caller, or
  8361. with a different significance threshold could give a drastically different
  8362. number of called peaks, and hence a drastically different distribution
  8363. of peak-to-TSS distances.
  8364. To address this, it is desirable to develop a better method of determining
  8365. the effective promoter radius that relies only on the distribution of read
  8366. coverage around the
  8367. \begin_inset Flex Glossary Term
  8368. status open
  8369. \begin_layout Plain Layout
  8370. TSS
  8371. \end_layout
  8372. \end_inset
  8373. , independent of the peak calling.
  8374. Furthermore, as demonstrated by the upstream-downstream asymmetries observed
  8375. in Figures
  8376. \begin_inset CommandInset ref
  8377. LatexCommand ref
  8378. reference "fig:H3K4me2-neighborhood"
  8379. plural "false"
  8380. caps "false"
  8381. noprefix "false"
  8382. \end_inset
  8383. ,
  8384. \begin_inset CommandInset ref
  8385. LatexCommand ref
  8386. reference "fig:H3K4me3-neighborhood"
  8387. plural "false"
  8388. caps "false"
  8389. noprefix "false"
  8390. \end_inset
  8391. , and
  8392. \begin_inset CommandInset ref
  8393. LatexCommand ref
  8394. reference "fig:H3K27me3-neighborhood"
  8395. plural "false"
  8396. caps "false"
  8397. noprefix "false"
  8398. \end_inset
  8399. , this definition should determine a different radius for the upstream and
  8400. downstream directions.
  8401. At this point, it may be better to rename this concept
  8402. \begin_inset Quotes eld
  8403. \end_inset
  8404. effective promoter extent
  8405. \begin_inset Quotes erd
  8406. \end_inset
  8407. and avoid the word
  8408. \begin_inset Quotes eld
  8409. \end_inset
  8410. radius
  8411. \begin_inset Quotes erd
  8412. \end_inset
  8413. , since a radius implies a symmetry about the
  8414. \begin_inset Flex Glossary Term
  8415. status open
  8416. \begin_layout Plain Layout
  8417. TSS
  8418. \end_layout
  8419. \end_inset
  8420. that is not supported by the data.
  8421. \end_layout
  8422. \begin_layout Standard
  8423. Beyond improving the definition of effective promoter extent, functional
  8424. validation is necessary to show that this measure of near-TSS enrichment
  8425. has biological meaning.
  8426. Figures
  8427. \begin_inset CommandInset ref
  8428. LatexCommand ref
  8429. reference "fig:H3K4me2-neighborhood"
  8430. plural "false"
  8431. caps "false"
  8432. noprefix "false"
  8433. \end_inset
  8434. and
  8435. \begin_inset CommandInset ref
  8436. LatexCommand ref
  8437. reference "fig:H3K4me3-neighborhood"
  8438. plural "false"
  8439. caps "false"
  8440. noprefix "false"
  8441. \end_inset
  8442. already provide a very limited functional validation of the chosen promoter
  8443. extents for H3K4me2 and H3K4me3 by showing that spikes in coverage within
  8444. this region are most strongly correlated with elevated gene expression.
  8445. However, there are other ways to show functional relevance of the promoter
  8446. extent.
  8447. For example, correlations could be computed between read counts in peaks
  8448. nearby gene promoters and the expression level of those genes, and these
  8449. correlations could be plotted against the distance of the peak upstream
  8450. or downstream of the gene's
  8451. \begin_inset Flex Glossary Term
  8452. status open
  8453. \begin_layout Plain Layout
  8454. TSS
  8455. \end_layout
  8456. \end_inset
  8457. .
  8458. If the promoter extent truly defines a
  8459. \begin_inset Quotes eld
  8460. \end_inset
  8461. sphere of influence
  8462. \begin_inset Quotes erd
  8463. \end_inset
  8464. within which a histone mark is involved with the regulation of a gene,
  8465. then the correlations for peaks within this extent should be significantly
  8466. higher than those further upstream or downstream.
  8467. Peaks within these extents may also be more likely to show differential
  8468. modification than those outside genic regions of the genome.
  8469. \end_layout
  8470. \begin_layout Subsection
  8471. Design experiments to focus on post-activation convergence of naïve & memory
  8472. cells
  8473. \end_layout
  8474. \begin_layout Standard
  8475. In this study, a convergence between naïve and memory cells was observed
  8476. in both the pattern of gene expression and in epigenetic state of the 3
  8477. histone marks studied, consistent with the hypothesis that any naïve cells
  8478. remaining 14 days after activation have differentiated into memory cells,
  8479. and that both gene expression and these histone marks are involved in this
  8480. differentiation.
  8481. However, the current study was not designed with this specific hypothesis
  8482. in mind, and it therefore has some deficiencies with regard to testing
  8483. it.
  8484. The memory CD4
  8485. \begin_inset Formula $^{+}$
  8486. \end_inset
  8487. samples at day 14 do not resemble the memory samples at day 0, indicating
  8488. that in the specific model of activation used for this experiment, the
  8489. cells are not guaranteed to return to their original pre-activation state,
  8490. or perhaps this process takes substantially longer than 14 days.
  8491. This difference is expected, as the cell cultures in this experiment were
  8492. treated with IL2 from day 5 onward
  8493. \begin_inset CommandInset citation
  8494. LatexCommand cite
  8495. key "LaMere2016"
  8496. literal "false"
  8497. \end_inset
  8498. , so the signalling environments in which the cells are cultured are different
  8499. at day 0 and day 14.
  8500. This is a challenge for testing the convergence hypothesis because the
  8501. ideal comparison to prove that naïve cells are converging to a resting
  8502. memory state would be to compare the final naïve time point to the Day
  8503. 0 memory samples, but this comparison is only meaningful if memory cells
  8504. generally return to the same
  8505. \begin_inset Quotes eld
  8506. \end_inset
  8507. resting
  8508. \begin_inset Quotes erd
  8509. \end_inset
  8510. state that they started at.
  8511. \end_layout
  8512. \begin_layout Standard
  8513. Because pre-culture and post-culture cells will probably never behave identicall
  8514. y even if they both nominally have a
  8515. \begin_inset Quotes eld
  8516. \end_inset
  8517. resting
  8518. \begin_inset Quotes erd
  8519. \end_inset
  8520. phenotype, a different experiment should be designed in which post-activation
  8521. naive cells are compared to memory cells that were cultured for the same
  8522. amount of time but never activated, in addition to post-activation memory
  8523. cells.
  8524. If the convergence hypothesis is correct, both post-activation cultures
  8525. should converge on the culture of never-activated memory cells.
  8526. \end_layout
  8527. \begin_layout Standard
  8528. In addition, if naïve-to-memory convergence is a general pattern, it should
  8529. also be detectable in other epigenetic marks, including other histone marks
  8530. and DNA methylation.
  8531. An experiment should be designed studying a large number of epigenetic
  8532. marks known or suspected to be involved in regulation of gene expression,
  8533. assaying all of these at the same pre- and post-activation time points.
  8534. Multi-dataset factor analysis methods like
  8535. \begin_inset Flex Glossary Term
  8536. status open
  8537. \begin_layout Plain Layout
  8538. MOFA
  8539. \end_layout
  8540. \end_inset
  8541. can then be used to identify coordinated patterns of regulation shared
  8542. across many epigenetic marks.
  8543. Of course, CD4
  8544. \begin_inset Formula $^{+}$
  8545. \end_inset
  8546. T-cells are not the only adaptive immune cells that exhibit memory formation.
  8547. A similar study could be designed for CD8
  8548. \begin_inset Formula $^{+}$
  8549. \end_inset
  8550. T-cells, B-cells, and even specific subsets of CD4
  8551. \begin_inset Formula $^{+}$
  8552. \end_inset
  8553. T-cells, such as Th1, Th2, Treg, and Th17 cells, to determine whether these
  8554. also show convergence.
  8555. \end_layout
  8556. \begin_layout Subsection
  8557. Follow up on hints of interesting patterns in promoter relative coverage
  8558. profiles
  8559. \end_layout
  8560. \begin_layout Standard
  8561. The analysis of promoter coverage landscapes in resting naive CD4
  8562. \begin_inset Formula $^{+}$
  8563. \end_inset
  8564. T-cells and their correlations with gene expression raises many interesting
  8565. questions.
  8566. The chosen analysis strategy used a clustering approach, but this approach
  8567. was subsequently shown to be a poor fit for the data.
  8568. In light of this, a better means of dimension reduction for promoter landscape
  8569. data is required.
  8570. In the case of H3K4me2 and H3K4me3, one option is to define the first 3
  8571. principal componets as orthogonal promoter
  8572. \begin_inset Quotes eld
  8573. \end_inset
  8574. state variables
  8575. \begin_inset Quotes erd
  8576. \end_inset
  8577. : upstream vs downstream coverage, TSS-centered peak vs trough, and proximal
  8578. upstream trough vs proximal downstream trough.
  8579. Gene expression could then be modeled as a function of these three variables,
  8580. or possibly as a function of the first
  8581. \begin_inset Formula $N$
  8582. \end_inset
  8583. principal components for
  8584. \begin_inset Formula $N$
  8585. \end_inset
  8586. larger than 3.
  8587. For H3K4me2 and H3K4me3, a better representation might be obtained by transform
  8588. ing the first 2 principal coordinates into a polar coordinate system
  8589. \begin_inset Formula $(r,\theta)$
  8590. \end_inset
  8591. with the origin at the center of the
  8592. \begin_inset Quotes eld
  8593. \end_inset
  8594. no peak
  8595. \begin_inset Quotes erd
  8596. \end_inset
  8597. cluster, where the radius
  8598. \begin_inset Formula $r$
  8599. \end_inset
  8600. represents the peak height above the background and the angle
  8601. \begin_inset Formula $\theta$
  8602. \end_inset
  8603. represents the peak's position upstream or downstream of the
  8604. \begin_inset Flex Glossary Term
  8605. status open
  8606. \begin_layout Plain Layout
  8607. TSS
  8608. \end_layout
  8609. \end_inset
  8610. .
  8611. \end_layout
  8612. \begin_layout Standard
  8613. Another weakness in the current analysis is the normalization of the average
  8614. abundance of each promoter to an average of zero.
  8615. This allows the abundance value in each window to represent the relative
  8616. abundance of that window compared to all the other windows in the interrogated
  8617. area.
  8618. However, while using the remainder of the windows to set the
  8619. \begin_inset Quotes eld
  8620. \end_inset
  8621. background
  8622. \begin_inset Quotes erd
  8623. \end_inset
  8624. level against which each window is normalized is convenient, it is far
  8625. from optimal.
  8626. As shown in Table
  8627. \begin_inset CommandInset ref
  8628. LatexCommand ref
  8629. reference "tab:peak-calling-summary"
  8630. plural "false"
  8631. caps "false"
  8632. noprefix "false"
  8633. \end_inset
  8634. , many enriched regions are larger than the 5
  8635. \begin_inset space ~
  8636. \end_inset
  8637. kbp radius., which means there may not be any
  8638. \begin_inset Quotes eld
  8639. \end_inset
  8640. background
  8641. \begin_inset Quotes erd
  8642. \end_inset
  8643. regions within 5
  8644. \begin_inset space ~
  8645. \end_inset
  8646. kbp of the
  8647. \begin_inset Flex Glossary Term
  8648. status open
  8649. \begin_layout Plain Layout
  8650. TSS
  8651. \end_layout
  8652. \end_inset
  8653. to normalize against.
  8654. For example, this normalization strategy fails to distinguish between a
  8655. trough in coverage at the
  8656. \begin_inset Flex Glossary Term
  8657. status open
  8658. \begin_layout Plain Layout
  8659. TSS
  8660. \end_layout
  8661. \end_inset
  8662. and a pair of wide peaks upstream and downstream of the
  8663. \begin_inset Flex Glossary Term
  8664. status open
  8665. \begin_layout Plain Layout
  8666. TSS
  8667. \end_layout
  8668. \end_inset
  8669. .
  8670. Both cases would present as lower coverage in the windows immediately adjacent
  8671. to the
  8672. \begin_inset Flex Glossary Term
  8673. status open
  8674. \begin_layout Plain Layout
  8675. TSS
  8676. \end_layout
  8677. \end_inset
  8678. and higher coverage in windows further away, but the functional implications
  8679. of these two cases might be completely different.
  8680. To improve the normalization, the background estimation method used by
  8681. \begin_inset Flex Glossary Term
  8682. status open
  8683. \begin_layout Plain Layout
  8684. SICER
  8685. \end_layout
  8686. \end_inset
  8687. , which is specifically designed for finding broad regions of enrichment,
  8688. should be adapted to estimate the background sequencing depth in each window
  8689. from the
  8690. \begin_inset Flex Glossary Term
  8691. status open
  8692. \begin_layout Plain Layout
  8693. ChIP-seq
  8694. \end_layout
  8695. \end_inset
  8696. input samples, and each window's read count should be normalized against
  8697. the background and reported as a
  8698. \begin_inset Flex Glossary Term
  8699. status open
  8700. \begin_layout Plain Layout
  8701. logFC
  8702. \end_layout
  8703. \end_inset
  8704. relative to that background.
  8705. \end_layout
  8706. \begin_layout Standard
  8707. Lastly, the analysis of promoter coverage landscapes presented in this work
  8708. only looked at promoter coverage of resting naive CD4
  8709. \begin_inset Formula $^{+}$
  8710. \end_inset
  8711. T-cells, with the goal of determining whether this initial promoter state
  8712. was predictive of post-activation changes in gene expression.
  8713. Changes in the promoter coverage landscape over time have not yet been
  8714. considered.
  8715. This represents a significant analysis challenge, by adding yet another
  8716. dimension (genomic coordinate) in to the data.
  8717. \end_layout
  8718. \begin_layout Subsection
  8719. Investigate causes of high correlation between mutually exclusive histone
  8720. marks
  8721. \end_layout
  8722. \begin_layout Standard
  8723. The high correlation between coverage depth observed between H3K4me2 and
  8724. H3K4me3 is both expected and unexpected.
  8725. Since both marks are associated with elevated gene transcription, a positive
  8726. correlation between them is not surprising.
  8727. However, these two marks represent different post-translational modifications
  8728. of the
  8729. \emph on
  8730. same
  8731. \emph default
  8732. lysine residue on the histone H3 polypeptide, which means that they cannot
  8733. both be present on the same H3 subunit.
  8734. Thus, the high correlation between them has several potential explanations.
  8735. One possible reason is cell population heterogeneity: perhaps some genomic
  8736. loci are frequently marked with H3K4me2 in some cells, while in other cells
  8737. the same loci are marked with H3K4me3.
  8738. Another possibility is allele-specific modifications: the loci are marked
  8739. in each diploid cell with H3K4me2 on one allele and H3K4me3 on the other
  8740. allele.
  8741. Lastly, since each histone octamer contains 2 H3 subunits, it is possible
  8742. that having one H3K4me2 mark and one H3K4me3 mark on a given histone octamer
  8743. represents a distinct epigenetic state with a different function than either
  8744. double H3K4me2 or double H3K4me3.
  8745. \end_layout
  8746. \begin_layout Standard
  8747. The hypothesis of allele-specific histone modification can easily be tested
  8748. with existing data by locating all heterozygous loci occurring within both
  8749. H3K4me3 and H3K4me2 peaks and checking for opposite allelic imbalance between
  8750. H3K4me3 and H3K4me2 read at each locus.
  8751. If the allele fractions in the reads from the two histone marks for each
  8752. locus are plotted against each other, there should be a negative correlation.
  8753. If no such negative correlation is found, then allele-specific histone
  8754. modification is unlikely to be the reason for the high correlation between
  8755. these histone marks.
  8756. \end_layout
  8757. \begin_layout Standard
  8758. To test the hypothesis that H3K4me2 and H3K4me3 marks are occurring on the
  8759. same histones.
  8760. A double
  8761. \begin_inset Flex Glossary Term
  8762. status open
  8763. \begin_layout Plain Layout
  8764. ChIP
  8765. \end_layout
  8766. \end_inset
  8767. experiment can be performed
  8768. \begin_inset CommandInset citation
  8769. LatexCommand cite
  8770. key "Jin2007"
  8771. literal "false"
  8772. \end_inset
  8773. .
  8774. In this assay, the input DNA goes through two sequential immunoprecipitations
  8775. with different antibodies: first the anti-H3K4me2 antibody, then the anti-H3K4m
  8776. e3 antibody.
  8777. Only bearing both histone marks, and the DNA associated with them, should
  8778. be isolated.
  8779. This can be followed by
  8780. \begin_inset Flex Glossary Term
  8781. status open
  8782. \begin_layout Plain Layout
  8783. HTS
  8784. \end_layout
  8785. \end_inset
  8786. to form a
  8787. \begin_inset Quotes eld
  8788. \end_inset
  8789. double
  8790. \begin_inset Flex Glossary Term
  8791. status open
  8792. \begin_layout Plain Layout
  8793. ChIP-seq
  8794. \end_layout
  8795. \end_inset
  8796. \begin_inset Quotes erd
  8797. \end_inset
  8798. assay that can be used to identify DNA regions bound by the isolated histones
  8799. \begin_inset CommandInset citation
  8800. LatexCommand cite
  8801. key "Jin2009"
  8802. literal "false"
  8803. \end_inset
  8804. .
  8805. If peaks called from this double
  8806. \begin_inset Flex Glossary Term
  8807. status open
  8808. \begin_layout Plain Layout
  8809. ChIP-seq
  8810. \end_layout
  8811. \end_inset
  8812. assay are highly correlated with both H3K4me2 and H3K4me3 peaks, then this
  8813. is strong evidence that the correlation between the two marks is actually
  8814. caused by physical co-location on the same histone.
  8815. \end_layout
  8816. \begin_layout Chapter
  8817. \begin_inset CommandInset label
  8818. LatexCommand label
  8819. name "chap:Improving-array-based-diagnostic"
  8820. \end_inset
  8821. Improving array-based diagnostics for transplant rejection by optimizing
  8822. data preprocessing
  8823. \end_layout
  8824. \begin_layout Standard
  8825. \size large
  8826. Ryan C.
  8827. Thompson, Sunil M.
  8828. Kurian, Thomas Whisnant, Padmaja Natarajan, Daniel R.
  8829. Salomon
  8830. \end_layout
  8831. \begin_layout Standard
  8832. \begin_inset ERT
  8833. status collapsed
  8834. \begin_layout Plain Layout
  8835. \backslash
  8836. glsresetall
  8837. \end_layout
  8838. \end_inset
  8839. \begin_inset Note Note
  8840. status collapsed
  8841. \begin_layout Plain Layout
  8842. Reintroduce all abbreviations
  8843. \end_layout
  8844. \end_inset
  8845. \end_layout
  8846. \begin_layout Section
  8847. Introduction
  8848. \end_layout
  8849. \begin_layout Standard
  8850. \begin_inset Flex TODO Note (inline)
  8851. status open
  8852. \begin_layout Plain Layout
  8853. Fill this out
  8854. \end_layout
  8855. \end_inset
  8856. \end_layout
  8857. \begin_layout Subsection
  8858. Arrays for diagnostics
  8859. \end_layout
  8860. \begin_layout Standard
  8861. Arrays are an attractive platform for diagnostics
  8862. \end_layout
  8863. \begin_layout Subsection
  8864. Proper pre-processing is essential for array data
  8865. \end_layout
  8866. \begin_layout Standard
  8867. Microarrays, bead arrays, and similar assays produce raw data in the form
  8868. of fluorescence intensity measurements, with each intensity measurement
  8869. proportional to the abundance of some fluorescently labelled target DNA
  8870. or RNA sequence that base pairs to a specific probe sequence.
  8871. However, the fluorescence measurements for each probe are also affected
  8872. my many technical confounding factors, such as the concentration of target
  8873. material, strength of off-target binding, the sensitivity of the imaging
  8874. sensor, and visual artifacts in the image.
  8875. Some array designs also use multiple probe sequences for each target.
  8876. Hence, extensive pre-processing of array data is necessary to normalize
  8877. out the effects of these technical factors and summarize the information
  8878. from multiple probes to arrive at a single usable estimate of abundance
  8879. or other relevant quantity, such as a ratio of two abundances, for each
  8880. target
  8881. \begin_inset CommandInset citation
  8882. LatexCommand cite
  8883. key "Gentleman2005"
  8884. literal "false"
  8885. \end_inset
  8886. .
  8887. \end_layout
  8888. \begin_layout Standard
  8889. The choice of pre-processing algorithms used in the analysis of an array
  8890. data set can have a large effect on the results of that analysis.
  8891. However, despite their importance, these steps are often neglected or rushed
  8892. in order to get to the more scientifically interesting analysis steps involving
  8893. the actual biology of the system under study.
  8894. Hence, it is often possible to achieve substantial gains in statistical
  8895. power, model goodness-of-fit, or other relevant performance measures, by
  8896. checking the assumptions made by each preprocessing step and choosing specific
  8897. normalization methods tailored to the specific goals of the current analysis.
  8898. \end_layout
  8899. \begin_layout Section
  8900. Approach
  8901. \end_layout
  8902. \begin_layout Subsection
  8903. Clinical diagnostic applications for microarrays require single-channel
  8904. normalization
  8905. \end_layout
  8906. \begin_layout Standard
  8907. As the cost of performing microarray assays falls, there is increasing interest
  8908. in using genomic assays for diagnostic purposes, such as distinguishing
  8909. \begin_inset ERT
  8910. status collapsed
  8911. \begin_layout Plain Layout
  8912. \backslash
  8913. glsdisp*{TX}{healthy transplants (TX)}
  8914. \end_layout
  8915. \end_inset
  8916. from transplants undergoing
  8917. \begin_inset Flex Glossary Term
  8918. status open
  8919. \begin_layout Plain Layout
  8920. AR
  8921. \end_layout
  8922. \end_inset
  8923. or
  8924. \begin_inset Flex Glossary Term
  8925. status open
  8926. \begin_layout Plain Layout
  8927. ADNR
  8928. \end_layout
  8929. \end_inset
  8930. .
  8931. However, the the standard normalization algorithm used for microarray data,
  8932. \begin_inset Flex Glossary Term
  8933. status open
  8934. \begin_layout Plain Layout
  8935. RMA
  8936. \end_layout
  8937. \end_inset
  8938. \begin_inset CommandInset citation
  8939. LatexCommand cite
  8940. key "Irizarry2003a"
  8941. literal "false"
  8942. \end_inset
  8943. , is not applicable in a clinical setting.
  8944. Two of the steps in
  8945. \begin_inset Flex Glossary Term
  8946. status open
  8947. \begin_layout Plain Layout
  8948. RMA
  8949. \end_layout
  8950. \end_inset
  8951. , quantile normalization and probe summarization by median polish, depend
  8952. on every array in the data set being normalized.
  8953. This means that adding or removing any arrays from a data set changes the
  8954. normalized values for all arrays, and data sets that have been normalized
  8955. separately cannot be compared to each other.
  8956. Hence, when using
  8957. \begin_inset Flex Glossary Term
  8958. status open
  8959. \begin_layout Plain Layout
  8960. RMA
  8961. \end_layout
  8962. \end_inset
  8963. , any arrays to be analyzed together must also be normalized together, and
  8964. the set of arrays included in the data set must be held constant throughout
  8965. an analysis.
  8966. \end_layout
  8967. \begin_layout Standard
  8968. These limitations present serious impediments to the use of arrays as a
  8969. diagnostic tool.
  8970. When training a classifier, the samples to be classified must not be involved
  8971. in any step of the training process, lest their inclusion bias the training
  8972. process.
  8973. Once a classifier is deployed in a clinical setting, the samples to be
  8974. classified will not even
  8975. \emph on
  8976. exist
  8977. \emph default
  8978. at the time of training, so including them would be impossible even if
  8979. it were statistically justifiable.
  8980. Therefore, any machine learning application for microarrays demands that
  8981. the normalized expression values computed for an array must depend only
  8982. on information contained within that array.
  8983. This would ensure that each array's normalization is independent of every
  8984. other array, and that arrays normalized separately can still be compared
  8985. to each other without bias.
  8986. Such a normalization is commonly referred to as
  8987. \begin_inset Quotes eld
  8988. \end_inset
  8989. single-channel normalization
  8990. \begin_inset Quotes erd
  8991. \end_inset
  8992. .
  8993. \end_layout
  8994. \begin_layout Standard
  8995. \begin_inset Flex Glossary Term (Capital)
  8996. status open
  8997. \begin_layout Plain Layout
  8998. fRMA
  8999. \end_layout
  9000. \end_inset
  9001. addresses these concerns by replacing the quantile normalization and median
  9002. polish with alternatives that do not introduce inter-array dependence,
  9003. allowing each array to be normalized independently of all others
  9004. \begin_inset CommandInset citation
  9005. LatexCommand cite
  9006. key "McCall2010"
  9007. literal "false"
  9008. \end_inset
  9009. .
  9010. Quantile normalization is performed against a pre-generated set of quantiles
  9011. learned from a collection of 850 publicly available arrays sampled from
  9012. a wide variety of tissues in
  9013. \begin_inset ERT
  9014. status collapsed
  9015. \begin_layout Plain Layout
  9016. \backslash
  9017. glsdisp*{GEO}{the Gene Expression Omnibus (GEO)}
  9018. \end_layout
  9019. \end_inset
  9020. .
  9021. Each array's probe intensity distribution is normalized against these pre-gener
  9022. ated quantiles.
  9023. The median polish step is replaced with a robust weighted average of probe
  9024. intensities, using inverse variance weights learned from the same public
  9025. \begin_inset Flex Glossary Term
  9026. status open
  9027. \begin_layout Plain Layout
  9028. GEO
  9029. \end_layout
  9030. \end_inset
  9031. data.
  9032. The result is a normalization that satisfies the requirements mentioned
  9033. above: each array is normalized independently of all others, and any two
  9034. normalized arrays can be compared directly to each other.
  9035. \end_layout
  9036. \begin_layout Standard
  9037. One important limitation of
  9038. \begin_inset Flex Glossary Term
  9039. status open
  9040. \begin_layout Plain Layout
  9041. fRMA
  9042. \end_layout
  9043. \end_inset
  9044. is that it requires a separate reference data set from which to learn the
  9045. parameters (reference quantiles and probe weights) that will be used to
  9046. normalize each array.
  9047. These parameters are specific to a given array platform, and pre-generated
  9048. parameters are only provided for the most common platforms, such as Affymetrix
  9049. hgu133plus2.
  9050. For a less common platform, such as hthgu133pluspm, is is necessary to
  9051. learn custom parameters from in-house data before
  9052. \begin_inset Flex Glossary Term
  9053. status open
  9054. \begin_layout Plain Layout
  9055. fRMA
  9056. \end_layout
  9057. \end_inset
  9058. can be used to normalize samples on that platform
  9059. \begin_inset CommandInset citation
  9060. LatexCommand cite
  9061. key "McCall2011"
  9062. literal "false"
  9063. \end_inset
  9064. .
  9065. \end_layout
  9066. \begin_layout Standard
  9067. One other option is the aptly-named
  9068. \begin_inset ERT
  9069. status collapsed
  9070. \begin_layout Plain Layout
  9071. \backslash
  9072. glsdisp*{SCAN}{Single Channel Array Normalization (SCAN)}
  9073. \end_layout
  9074. \end_inset
  9075. , which adapts a normalization method originally designed for tiling arrays
  9076. \begin_inset CommandInset citation
  9077. LatexCommand cite
  9078. key "Piccolo2012"
  9079. literal "false"
  9080. \end_inset
  9081. .
  9082. \begin_inset Flex Glossary Term
  9083. status open
  9084. \begin_layout Plain Layout
  9085. SCAN
  9086. \end_layout
  9087. \end_inset
  9088. is truly single-channel in that it does not require a set of normalization
  9089. parameters estimated from an external set of reference samples like
  9090. \begin_inset Flex Glossary Term
  9091. status open
  9092. \begin_layout Plain Layout
  9093. fRMA
  9094. \end_layout
  9095. \end_inset
  9096. does.
  9097. \end_layout
  9098. \begin_layout Subsection
  9099. Heteroskedasticity must be accounted for in methylation array data
  9100. \end_layout
  9101. \begin_layout Standard
  9102. DNA methylation arrays are a relatively new kind of assay that uses microarrays
  9103. to measure the degree of methylation on cytosines in specific regions arrayed
  9104. across the genome.
  9105. First, bisulfite treatment converts all unmethylated cytosines to uracil
  9106. (which are read as thymine during amplification and sequencing) while leaving
  9107. methylated cytosines unaffected.
  9108. Then, each target region is interrogated with two probes: one binds to
  9109. the original genomic sequence and interrogates the level of methylated
  9110. DNA, and the other binds to the same sequence with all cytosines replaced
  9111. by thymidines and interrogates the level of unmethylated DNA.
  9112. \end_layout
  9113. \begin_layout Standard
  9114. After normalization, these two probe intensities are summarized in one of
  9115. two ways, each with advantages and disadvantages.
  9116. β
  9117. \series bold
  9118. \series default
  9119. values, interpreted as fraction of DNA copies methylated, range from 0 to
  9120. 1.
  9121. β
  9122. \series bold
  9123. \series default
  9124. values are conceptually easy to interpret, but the constrained range makes
  9125. them unsuitable for linear modeling, and their error distributions are
  9126. highly non-normal, which also frustrates linear modeling.
  9127. \begin_inset ERT
  9128. status collapsed
  9129. \begin_layout Plain Layout
  9130. \backslash
  9131. glsdisp*{M-value}{M-values}
  9132. \end_layout
  9133. \end_inset
  9134. , interpreted as the log ratios of methylated to unmethylated copies for
  9135. each probe region, are computed by mapping the beta values from
  9136. \begin_inset Formula $[0,1]$
  9137. \end_inset
  9138. onto
  9139. \begin_inset Formula $(-\infty,+\infty)$
  9140. \end_inset
  9141. using a sigmoid curve (Figure
  9142. \begin_inset CommandInset ref
  9143. LatexCommand ref
  9144. reference "fig:Sigmoid-beta-m-mapping"
  9145. plural "false"
  9146. caps "false"
  9147. noprefix "false"
  9148. \end_inset
  9149. ).
  9150. This transformation results in values with better statistical properties:
  9151. the unconstrained range is suitable for linear modeling, and the error
  9152. distributions are more normal.
  9153. Hence, most linear modeling and other statistical testing on methylation
  9154. arrays is performed using
  9155. \begin_inset Flex Glossary Term (pl)
  9156. status open
  9157. \begin_layout Plain Layout
  9158. M-value
  9159. \end_layout
  9160. \end_inset
  9161. .
  9162. \end_layout
  9163. \begin_layout Standard
  9164. \begin_inset Float figure
  9165. wide false
  9166. sideways false
  9167. status collapsed
  9168. \begin_layout Plain Layout
  9169. \align center
  9170. \begin_inset Graphics
  9171. filename graphics/methylvoom/sigmoid.pdf
  9172. lyxscale 50
  9173. width 60col%
  9174. groupId colwidth
  9175. \end_inset
  9176. \end_layout
  9177. \begin_layout Plain Layout
  9178. \begin_inset Caption Standard
  9179. \begin_layout Plain Layout
  9180. \begin_inset Argument 1
  9181. status collapsed
  9182. \begin_layout Plain Layout
  9183. Sigmoid shape of the mapping between β and M values.
  9184. \end_layout
  9185. \end_inset
  9186. \begin_inset CommandInset label
  9187. LatexCommand label
  9188. name "fig:Sigmoid-beta-m-mapping"
  9189. \end_inset
  9190. \series bold
  9191. Sigmoid shape of the mapping between β and M values.
  9192. \series default
  9193. This mapping is monotonic and non-linear, but it is approximately linear
  9194. in the neighborhood of
  9195. \begin_inset Formula $(\beta=0.5,M=0)$
  9196. \end_inset
  9197. .
  9198. \end_layout
  9199. \end_inset
  9200. \end_layout
  9201. \end_inset
  9202. \end_layout
  9203. \begin_layout Standard
  9204. However, the steep slope of the sigmoid transformation near 0 and 1 tends
  9205. to over-exaggerate small differences in β values near those extremes, which
  9206. in turn amplifies the error in those values, leading to a U-shaped trend
  9207. in the mean-variance curve: extreme values have higher variances than values
  9208. near the middle.
  9209. This mean-variance dependency must be accounted for when fitting the linear
  9210. model for differential methylation, or else the variance will be systematically
  9211. overestimated for probes with moderate
  9212. \begin_inset Flex Glossary Term (pl)
  9213. status open
  9214. \begin_layout Plain Layout
  9215. M-value
  9216. \end_layout
  9217. \end_inset
  9218. and underestimated for probes with extreme
  9219. \begin_inset Flex Glossary Term (pl)
  9220. status open
  9221. \begin_layout Plain Layout
  9222. M-value
  9223. \end_layout
  9224. \end_inset
  9225. .
  9226. This is particularly undesirable for methylation data because the intermediate
  9227. \begin_inset Flex Glossary Term (pl)
  9228. status open
  9229. \begin_layout Plain Layout
  9230. M-value
  9231. \end_layout
  9232. \end_inset
  9233. are the ones of most interest, since they are more likely to represent
  9234. areas of varying methylation, whereas extreme
  9235. \begin_inset Flex Glossary Term (pl)
  9236. status open
  9237. \begin_layout Plain Layout
  9238. M-value
  9239. \end_layout
  9240. \end_inset
  9241. typically represent complete methylation or complete lack of methylation.
  9242. \end_layout
  9243. \begin_layout Standard
  9244. \begin_inset Flex Glossary Term (Capital)
  9245. status open
  9246. \begin_layout Plain Layout
  9247. RNA-seq
  9248. \end_layout
  9249. \end_inset
  9250. read count data are also known to show heteroskedasticity, and the voom
  9251. method was introduced for modeling this heteroskedasticity by estimating
  9252. the mean-variance trend in the data and using this trend to assign precision
  9253. weights to each observation
  9254. \begin_inset CommandInset citation
  9255. LatexCommand cite
  9256. key "Law2014"
  9257. literal "false"
  9258. \end_inset
  9259. .
  9260. While methylation array data are not derived from counts and have a very
  9261. different mean-variance relationship from that of typical
  9262. \begin_inset Flex Glossary Term
  9263. status open
  9264. \begin_layout Plain Layout
  9265. RNA-seq
  9266. \end_layout
  9267. \end_inset
  9268. data, the voom method makes no specific assumptions on the shape of the
  9269. mean-variance relationship – it only assumes that the relationship can
  9270. be modeled as a smooth curve.
  9271. Hence, the method is sufficiently general to model the mean-variance relationsh
  9272. ip in methylation array data.
  9273. However, while the method does not require count data as input, the standard
  9274. implementation of voom assumes that the input is given in raw read counts,
  9275. and it must be adapted to run on methylation
  9276. \begin_inset Flex Glossary Term (pl)
  9277. status open
  9278. \begin_layout Plain Layout
  9279. M-value
  9280. \end_layout
  9281. \end_inset
  9282. .
  9283. \end_layout
  9284. \begin_layout Section
  9285. Methods
  9286. \end_layout
  9287. \begin_layout Subsection
  9288. Evaluation of classifier performance with different normalization methods
  9289. \end_layout
  9290. \begin_layout Standard
  9291. For testing different expression microarray normalizations, a data set of
  9292. 157 hgu133plus2 arrays was used, consisting of blood samples from kidney
  9293. transplant patients whose grafts had been graded as
  9294. \begin_inset Flex Glossary Term
  9295. status open
  9296. \begin_layout Plain Layout
  9297. TX
  9298. \end_layout
  9299. \end_inset
  9300. ,
  9301. \begin_inset Flex Glossary Term
  9302. status open
  9303. \begin_layout Plain Layout
  9304. AR
  9305. \end_layout
  9306. \end_inset
  9307. , or
  9308. \begin_inset Flex Glossary Term
  9309. status open
  9310. \begin_layout Plain Layout
  9311. ADNR
  9312. \end_layout
  9313. \end_inset
  9314. via biopsy and histology (46 TX, 69 AR, 42 ADNR)
  9315. \begin_inset CommandInset citation
  9316. LatexCommand cite
  9317. key "Kurian2014"
  9318. literal "true"
  9319. \end_inset
  9320. .
  9321. Additionally, an external validation set of 75 samples was gathered from
  9322. public
  9323. \begin_inset Flex Glossary Term
  9324. status open
  9325. \begin_layout Plain Layout
  9326. GEO
  9327. \end_layout
  9328. \end_inset
  9329. data (37 TX, 38 AR, no ADNR).
  9330. \end_layout
  9331. \begin_layout Standard
  9332. \begin_inset Flex TODO Note (inline)
  9333. status open
  9334. \begin_layout Plain Layout
  9335. Find appropriate GEO identifiers if possible.
  9336. Kurian 2014 says GSE15296, but this seems to be different data.
  9337. I also need to look up the GEO accession for the external validation set.
  9338. \end_layout
  9339. \end_inset
  9340. \end_layout
  9341. \begin_layout Standard
  9342. To evaluate the effect of each normalization on classifier performance,
  9343. the same classifier training and validation procedure was used after each
  9344. normalization method.
  9345. The
  9346. \begin_inset Flex Glossary Term
  9347. status open
  9348. \begin_layout Plain Layout
  9349. PAM
  9350. \end_layout
  9351. \end_inset
  9352. algorithm was used to train a nearest shrunken centroid classifier on the
  9353. training set and select the appropriate threshold for centroid shrinking
  9354. \begin_inset CommandInset citation
  9355. LatexCommand cite
  9356. key "Tibshirani2002"
  9357. literal "false"
  9358. \end_inset
  9359. .
  9360. Then the trained classifier was used to predict the class probabilities
  9361. of each validation sample.
  9362. From these class probabilities,
  9363. \begin_inset Flex Glossary Term
  9364. status open
  9365. \begin_layout Plain Layout
  9366. ROC
  9367. \end_layout
  9368. \end_inset
  9369. curves and
  9370. \begin_inset Flex Glossary Term
  9371. status open
  9372. \begin_layout Plain Layout
  9373. AUC
  9374. \end_layout
  9375. \end_inset
  9376. values were generated
  9377. \begin_inset CommandInset citation
  9378. LatexCommand cite
  9379. key "Turck2011"
  9380. literal "false"
  9381. \end_inset
  9382. .
  9383. Each normalization was tested on two different sets of training and validation
  9384. samples.
  9385. For internal validation, the 115
  9386. \begin_inset Flex Glossary Term
  9387. status open
  9388. \begin_layout Plain Layout
  9389. TX
  9390. \end_layout
  9391. \end_inset
  9392. and
  9393. \begin_inset Flex Glossary Term
  9394. status open
  9395. \begin_layout Plain Layout
  9396. AR
  9397. \end_layout
  9398. \end_inset
  9399. arrays in the internal set were split at random into two equal sized sets,
  9400. one for training and one for validation, each containing the same numbers
  9401. of
  9402. \begin_inset Flex Glossary Term
  9403. status open
  9404. \begin_layout Plain Layout
  9405. TX
  9406. \end_layout
  9407. \end_inset
  9408. and
  9409. \begin_inset Flex Glossary Term
  9410. status open
  9411. \begin_layout Plain Layout
  9412. AR
  9413. \end_layout
  9414. \end_inset
  9415. samples as the other set.
  9416. For external validation, the full set of 115
  9417. \begin_inset Flex Glossary Term
  9418. status open
  9419. \begin_layout Plain Layout
  9420. TX
  9421. \end_layout
  9422. \end_inset
  9423. and
  9424. \begin_inset Flex Glossary Term
  9425. status open
  9426. \begin_layout Plain Layout
  9427. AR
  9428. \end_layout
  9429. \end_inset
  9430. samples were used as a training set, and the 75 external
  9431. \begin_inset Flex Glossary Term
  9432. status open
  9433. \begin_layout Plain Layout
  9434. TX
  9435. \end_layout
  9436. \end_inset
  9437. and
  9438. \begin_inset Flex Glossary Term
  9439. status open
  9440. \begin_layout Plain Layout
  9441. AR
  9442. \end_layout
  9443. \end_inset
  9444. samples were used as the validation set.
  9445. Thus, 2
  9446. \begin_inset Flex Glossary Term
  9447. status open
  9448. \begin_layout Plain Layout
  9449. ROC
  9450. \end_layout
  9451. \end_inset
  9452. curves and
  9453. \begin_inset Flex Glossary Term
  9454. status open
  9455. \begin_layout Plain Layout
  9456. AUC
  9457. \end_layout
  9458. \end_inset
  9459. values were generated for each normalization method: one internal and one
  9460. external.
  9461. Because the external validation set contains no
  9462. \begin_inset Flex Glossary Term
  9463. status open
  9464. \begin_layout Plain Layout
  9465. ADNR
  9466. \end_layout
  9467. \end_inset
  9468. samples, only classification of
  9469. \begin_inset Flex Glossary Term
  9470. status open
  9471. \begin_layout Plain Layout
  9472. TX
  9473. \end_layout
  9474. \end_inset
  9475. and
  9476. \begin_inset Flex Glossary Term
  9477. status open
  9478. \begin_layout Plain Layout
  9479. AR
  9480. \end_layout
  9481. \end_inset
  9482. samples was considered.
  9483. The
  9484. \begin_inset Flex Glossary Term
  9485. status open
  9486. \begin_layout Plain Layout
  9487. ADNR
  9488. \end_layout
  9489. \end_inset
  9490. samples were included during normalization but excluded from all classifier
  9491. training and validation.
  9492. This ensures that the performance on internal and external validation sets
  9493. is directly comparable, since both are performing the same task: distinguishing
  9494. \begin_inset Flex Glossary Term
  9495. status open
  9496. \begin_layout Plain Layout
  9497. TX
  9498. \end_layout
  9499. \end_inset
  9500. from
  9501. \begin_inset Flex Glossary Term
  9502. status open
  9503. \begin_layout Plain Layout
  9504. AR
  9505. \end_layout
  9506. \end_inset
  9507. .
  9508. \end_layout
  9509. \begin_layout Standard
  9510. \begin_inset Flex TODO Note (inline)
  9511. status open
  9512. \begin_layout Plain Layout
  9513. Summarize the get.best.threshold algorithm for PAM threshold selection, or
  9514. just put the code online?
  9515. \end_layout
  9516. \end_inset
  9517. \end_layout
  9518. \begin_layout Standard
  9519. Six different normalization strategies were evaluated.
  9520. First, 2 well-known non-single-channel normalization methods were considered:
  9521. \begin_inset Flex Glossary Term
  9522. status open
  9523. \begin_layout Plain Layout
  9524. RMA
  9525. \end_layout
  9526. \end_inset
  9527. and dChip
  9528. \begin_inset CommandInset citation
  9529. LatexCommand cite
  9530. key "Li2001,Irizarry2003a"
  9531. literal "false"
  9532. \end_inset
  9533. .
  9534. Since
  9535. \begin_inset Flex Glossary Term
  9536. status open
  9537. \begin_layout Plain Layout
  9538. RMA
  9539. \end_layout
  9540. \end_inset
  9541. produces expression values on a
  9542. \begin_inset Formula $\log_{2}$
  9543. \end_inset
  9544. scale and dChip does not, the values from dChip were
  9545. \begin_inset Formula $\log_{2}$
  9546. \end_inset
  9547. transformed after normalization.
  9548. Next,
  9549. \begin_inset Flex Glossary Term
  9550. status open
  9551. \begin_layout Plain Layout
  9552. RMA
  9553. \end_layout
  9554. \end_inset
  9555. and dChip followed by
  9556. \begin_inset Flex Glossary Term
  9557. status open
  9558. \begin_layout Plain Layout
  9559. GRSN
  9560. \end_layout
  9561. \end_inset
  9562. were tested
  9563. \begin_inset CommandInset citation
  9564. LatexCommand cite
  9565. key "Pelz2008"
  9566. literal "false"
  9567. \end_inset
  9568. .
  9569. Post-processing with
  9570. \begin_inset Flex Glossary Term
  9571. status open
  9572. \begin_layout Plain Layout
  9573. GRSN
  9574. \end_layout
  9575. \end_inset
  9576. does not turn
  9577. \begin_inset Flex Glossary Term
  9578. status open
  9579. \begin_layout Plain Layout
  9580. RMA
  9581. \end_layout
  9582. \end_inset
  9583. or dChip into single-channel methods, but it may help mitigate batch effects
  9584. and is therefore useful as a benchmark.
  9585. Lastly, the two single-channel normalization methods,
  9586. \begin_inset Flex Glossary Term
  9587. status open
  9588. \begin_layout Plain Layout
  9589. fRMA
  9590. \end_layout
  9591. \end_inset
  9592. and
  9593. \begin_inset Flex Glossary Term
  9594. status open
  9595. \begin_layout Plain Layout
  9596. SCAN
  9597. \end_layout
  9598. \end_inset
  9599. , were tested
  9600. \begin_inset CommandInset citation
  9601. LatexCommand cite
  9602. key "McCall2010,Piccolo2012"
  9603. literal "false"
  9604. \end_inset
  9605. .
  9606. When evaluating internal validation performance, only the 157 internal
  9607. samples were normalized; when evaluating external validation performance,
  9608. all 157 internal samples and 75 external samples were normalized together.
  9609. \end_layout
  9610. \begin_layout Standard
  9611. For demonstrating the problem with separate normalization of training and
  9612. validation data, one additional normalization was performed: the internal
  9613. and external sets were each normalized separately using
  9614. \begin_inset Flex Glossary Term
  9615. status open
  9616. \begin_layout Plain Layout
  9617. RMA
  9618. \end_layout
  9619. \end_inset
  9620. , and the normalized data for each set were combined into a single set with
  9621. no further attempts at normalizing between the two sets.
  9622. This represents approximately how
  9623. \begin_inset Flex Glossary Term
  9624. status open
  9625. \begin_layout Plain Layout
  9626. RMA
  9627. \end_layout
  9628. \end_inset
  9629. would have to be used in a clinical setting, where the samples to be classified
  9630. are not available at the time the classifier is trained.
  9631. \end_layout
  9632. \begin_layout Subsection
  9633. Generating custom fRMA vectors for hthgu133pluspm array platform
  9634. \end_layout
  9635. \begin_layout Standard
  9636. In order to enable
  9637. \begin_inset Flex Glossary Term
  9638. status open
  9639. \begin_layout Plain Layout
  9640. fRMA
  9641. \end_layout
  9642. \end_inset
  9643. normalization for the hthgu133pluspm array platform, custom
  9644. \begin_inset Flex Glossary Term
  9645. status open
  9646. \begin_layout Plain Layout
  9647. fRMA
  9648. \end_layout
  9649. \end_inset
  9650. normalization vectors were trained using the
  9651. \begin_inset Flex Code
  9652. status open
  9653. \begin_layout Plain Layout
  9654. frmaTools
  9655. \end_layout
  9656. \end_inset
  9657. package
  9658. \begin_inset CommandInset citation
  9659. LatexCommand cite
  9660. key "McCall2011"
  9661. literal "false"
  9662. \end_inset
  9663. .
  9664. Separate vectors were created for two types of samples: kidney graft biopsy
  9665. samples and blood samples from graft recipients.
  9666. For training, 341 kidney biopsy samples from 2 data sets and 965 blood
  9667. samples from 5 data sets were used as the reference set.
  9668. Arrays were groups into batches based on unique combinations of sample
  9669. type (blood or biopsy), diagnosis (TX, AR, etc.), data set, and scan date.
  9670. Thus, each batch represents arrays of the same kind that were run together
  9671. on the same day.
  9672. For estimating the probe inverse variance weights, frmaTools requires equal-siz
  9673. ed batches, which means a batch size must be chosen, and then batches smaller
  9674. than that size must be ignored, while batches larger than the chosen size
  9675. must be downsampled.
  9676. This downsampling is performed randomly, so the sampling process is repeated
  9677. 5 times and the resulting normalizations are compared to each other.
  9678. \end_layout
  9679. \begin_layout Standard
  9680. To evaluate the consistency of the generated normalization vectors, the
  9681. 5
  9682. \begin_inset Flex Glossary Term
  9683. status open
  9684. \begin_layout Plain Layout
  9685. fRMA
  9686. \end_layout
  9687. \end_inset
  9688. vector sets generated from 5 random batch samplings were each used to normalize
  9689. the same 20 randomly selected samples from each tissue.
  9690. Then the normalized expression values for each probe on each array were
  9691. compared across all normalizations.
  9692. Each
  9693. \begin_inset Flex Glossary Term
  9694. status open
  9695. \begin_layout Plain Layout
  9696. fRMA
  9697. \end_layout
  9698. \end_inset
  9699. normalization was also compared against the normalized expression values
  9700. obtained by normalizing the same 20 samples with ordinary
  9701. \begin_inset Flex Glossary Term
  9702. status open
  9703. \begin_layout Plain Layout
  9704. RMA
  9705. \end_layout
  9706. \end_inset
  9707. .
  9708. \end_layout
  9709. \begin_layout Subsection
  9710. Modeling methylation array M-value heteroskedasticity with a modified voom
  9711. implementation
  9712. \end_layout
  9713. \begin_layout Standard
  9714. \begin_inset Flex TODO Note (inline)
  9715. status open
  9716. \begin_layout Plain Layout
  9717. Put code on Github and reference it.
  9718. \end_layout
  9719. \end_inset
  9720. \end_layout
  9721. \begin_layout Standard
  9722. To investigate the whether DNA methylation could be used to distinguish
  9723. between healthy and dysfunctional transplants, a data set of 78 Illumina
  9724. 450k methylation arrays from human kidney graft biopsies was analyzed for
  9725. differential methylation between 4 transplant statuses:
  9726. \begin_inset Flex Glossary Term
  9727. status open
  9728. \begin_layout Plain Layout
  9729. TX
  9730. \end_layout
  9731. \end_inset
  9732. , transplants undergoing
  9733. \begin_inset Flex Glossary Term
  9734. status open
  9735. \begin_layout Plain Layout
  9736. AR
  9737. \end_layout
  9738. \end_inset
  9739. ,
  9740. \begin_inset Flex Glossary Term
  9741. status open
  9742. \begin_layout Plain Layout
  9743. ADNR
  9744. \end_layout
  9745. \end_inset
  9746. , and
  9747. \begin_inset Flex Glossary Term
  9748. status open
  9749. \begin_layout Plain Layout
  9750. CAN
  9751. \end_layout
  9752. \end_inset
  9753. .
  9754. The data consisted of 33 TX, 9 AR, 8 ADNR, and 28 CAN samples.
  9755. The uneven group sizes are a result of taking the biopsy samples before
  9756. the eventual fate of the transplant was known.
  9757. Each sample was additionally annotated with a donor
  9758. \begin_inset Flex Glossary Term
  9759. status open
  9760. \begin_layout Plain Layout
  9761. ID
  9762. \end_layout
  9763. \end_inset
  9764. (anonymized), sex, age, ethnicity, creatinine level, and diabetes diagnosis
  9765. (all samples in this data set came from patients with either
  9766. \begin_inset Flex Glossary Term
  9767. status open
  9768. \begin_layout Plain Layout
  9769. T1D
  9770. \end_layout
  9771. \end_inset
  9772. or
  9773. \begin_inset Flex Glossary Term
  9774. status open
  9775. \begin_layout Plain Layout
  9776. T2D
  9777. \end_layout
  9778. \end_inset
  9779. ).
  9780. \end_layout
  9781. \begin_layout Standard
  9782. The intensity data were first normalized using
  9783. \begin_inset Flex Glossary Term
  9784. status open
  9785. \begin_layout Plain Layout
  9786. SWAN
  9787. \end_layout
  9788. \end_inset
  9789. \begin_inset CommandInset citation
  9790. LatexCommand cite
  9791. key "Maksimovic2012"
  9792. literal "false"
  9793. \end_inset
  9794. , then converted to intensity ratios (beta values)
  9795. \begin_inset CommandInset citation
  9796. LatexCommand cite
  9797. key "Aryee2014"
  9798. literal "false"
  9799. \end_inset
  9800. .
  9801. Any probes binding to loci that overlapped annotated SNPs were dropped,
  9802. and the annotated sex of each sample was verified against the sex inferred
  9803. from the ratio of median probe intensities for the X and Y chromosomes.
  9804. Then, the ratios were transformed to
  9805. \begin_inset Flex Glossary Term (pl)
  9806. status open
  9807. \begin_layout Plain Layout
  9808. M-value
  9809. \end_layout
  9810. \end_inset
  9811. .
  9812. \end_layout
  9813. \begin_layout Standard
  9814. \begin_inset Float table
  9815. wide false
  9816. sideways false
  9817. status collapsed
  9818. \begin_layout Plain Layout
  9819. \align center
  9820. \begin_inset Tabular
  9821. <lyxtabular version="3" rows="4" columns="6">
  9822. <features tabularvalignment="middle">
  9823. <column alignment="center" valignment="top">
  9824. <column alignment="center" valignment="top">
  9825. <column alignment="center" valignment="top">
  9826. <column alignment="center" valignment="top">
  9827. <column alignment="center" valignment="top">
  9828. <column alignment="center" valignment="top">
  9829. <row>
  9830. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  9831. \begin_inset Text
  9832. \begin_layout Plain Layout
  9833. Analysis
  9834. \end_layout
  9835. \end_inset
  9836. </cell>
  9837. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  9838. \begin_inset Text
  9839. \begin_layout Plain Layout
  9840. random effect
  9841. \end_layout
  9842. \end_inset
  9843. </cell>
  9844. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  9845. \begin_inset Text
  9846. \begin_layout Plain Layout
  9847. eBayes
  9848. \end_layout
  9849. \end_inset
  9850. </cell>
  9851. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  9852. \begin_inset Text
  9853. \begin_layout Plain Layout
  9854. SVA
  9855. \end_layout
  9856. \end_inset
  9857. </cell>
  9858. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  9859. \begin_inset Text
  9860. \begin_layout Plain Layout
  9861. weights
  9862. \end_layout
  9863. \end_inset
  9864. </cell>
  9865. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  9866. \begin_inset Text
  9867. \begin_layout Plain Layout
  9868. voom
  9869. \end_layout
  9870. \end_inset
  9871. </cell>
  9872. </row>
  9873. <row>
  9874. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  9875. \begin_inset Text
  9876. \begin_layout Plain Layout
  9877. A
  9878. \end_layout
  9879. \end_inset
  9880. </cell>
  9881. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  9882. \begin_inset Text
  9883. \begin_layout Plain Layout
  9884. Yes
  9885. \end_layout
  9886. \end_inset
  9887. </cell>
  9888. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  9889. \begin_inset Text
  9890. \begin_layout Plain Layout
  9891. Yes
  9892. \end_layout
  9893. \end_inset
  9894. </cell>
  9895. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  9896. \begin_inset Text
  9897. \begin_layout Plain Layout
  9898. No
  9899. \end_layout
  9900. \end_inset
  9901. </cell>
  9902. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  9903. \begin_inset Text
  9904. \begin_layout Plain Layout
  9905. No
  9906. \end_layout
  9907. \end_inset
  9908. </cell>
  9909. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  9910. \begin_inset Text
  9911. \begin_layout Plain Layout
  9912. No
  9913. \end_layout
  9914. \end_inset
  9915. </cell>
  9916. </row>
  9917. <row>
  9918. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  9919. \begin_inset Text
  9920. \begin_layout Plain Layout
  9921. B
  9922. \end_layout
  9923. \end_inset
  9924. </cell>
  9925. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  9926. \begin_inset Text
  9927. \begin_layout Plain Layout
  9928. Yes
  9929. \end_layout
  9930. \end_inset
  9931. </cell>
  9932. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  9933. \begin_inset Text
  9934. \begin_layout Plain Layout
  9935. Yes
  9936. \end_layout
  9937. \end_inset
  9938. </cell>
  9939. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  9940. \begin_inset Text
  9941. \begin_layout Plain Layout
  9942. Yes
  9943. \end_layout
  9944. \end_inset
  9945. </cell>
  9946. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  9947. \begin_inset Text
  9948. \begin_layout Plain Layout
  9949. Yes
  9950. \end_layout
  9951. \end_inset
  9952. </cell>
  9953. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  9954. \begin_inset Text
  9955. \begin_layout Plain Layout
  9956. No
  9957. \end_layout
  9958. \end_inset
  9959. </cell>
  9960. </row>
  9961. <row>
  9962. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  9963. \begin_inset Text
  9964. \begin_layout Plain Layout
  9965. C
  9966. \end_layout
  9967. \end_inset
  9968. </cell>
  9969. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  9970. \begin_inset Text
  9971. \begin_layout Plain Layout
  9972. Yes
  9973. \end_layout
  9974. \end_inset
  9975. </cell>
  9976. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  9977. \begin_inset Text
  9978. \begin_layout Plain Layout
  9979. Yes
  9980. \end_layout
  9981. \end_inset
  9982. </cell>
  9983. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  9984. \begin_inset Text
  9985. \begin_layout Plain Layout
  9986. Yes
  9987. \end_layout
  9988. \end_inset
  9989. </cell>
  9990. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  9991. \begin_inset Text
  9992. \begin_layout Plain Layout
  9993. Yes
  9994. \end_layout
  9995. \end_inset
  9996. </cell>
  9997. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  9998. \begin_inset Text
  9999. \begin_layout Plain Layout
  10000. Yes
  10001. \end_layout
  10002. \end_inset
  10003. </cell>
  10004. </row>
  10005. </lyxtabular>
  10006. \end_inset
  10007. \end_layout
  10008. \begin_layout Plain Layout
  10009. \begin_inset Caption Standard
  10010. \begin_layout Plain Layout
  10011. \begin_inset Argument 1
  10012. status collapsed
  10013. \begin_layout Plain Layout
  10014. Summary of analysis variants for methylation array data.
  10015. \end_layout
  10016. \end_inset
  10017. \begin_inset CommandInset label
  10018. LatexCommand label
  10019. name "tab:Summary-of-meth-analysis"
  10020. \end_inset
  10021. \series bold
  10022. Summary of analysis variants for methylation array data.
  10023. \series default
  10024. Each analysis included a different set of steps to adjust or account for
  10025. various systematic features of the data.
  10026. Random effect: The model included a random effect accounting for correlation
  10027. between samples from the same patient
  10028. \begin_inset CommandInset citation
  10029. LatexCommand cite
  10030. key "Smyth2005a"
  10031. literal "false"
  10032. \end_inset
  10033. ; eBayes: Empirical bayes squeezing of per-probe variances toward the mean-varia
  10034. nce trend
  10035. \begin_inset CommandInset citation
  10036. LatexCommand cite
  10037. key "Ritchie2015"
  10038. literal "false"
  10039. \end_inset
  10040. ; SVA: Surrogate variable analysis to account for unobserved confounders
  10041. \begin_inset CommandInset citation
  10042. LatexCommand cite
  10043. key "Leek2007"
  10044. literal "false"
  10045. \end_inset
  10046. ; Weights: Estimate sample weights to account for differences in sample
  10047. quality
  10048. \begin_inset CommandInset citation
  10049. LatexCommand cite
  10050. key "Liu2015,Ritchie2006"
  10051. literal "false"
  10052. \end_inset
  10053. ; voom: Use mean-variance trend to assign individual sample weights
  10054. \begin_inset CommandInset citation
  10055. LatexCommand cite
  10056. key "Law2014"
  10057. literal "false"
  10058. \end_inset
  10059. .
  10060. See the text for a more detailed explanation of each step.
  10061. \end_layout
  10062. \end_inset
  10063. \end_layout
  10064. \end_inset
  10065. \end_layout
  10066. \begin_layout Standard
  10067. From the
  10068. \begin_inset Flex Glossary Term (pl)
  10069. status open
  10070. \begin_layout Plain Layout
  10071. M-value
  10072. \end_layout
  10073. \end_inset
  10074. , a series of parallel analyses was performed, each adding additional steps
  10075. into the model fit to accommodate a feature of the data (see Table
  10076. \begin_inset CommandInset ref
  10077. LatexCommand ref
  10078. reference "tab:Summary-of-meth-analysis"
  10079. plural "false"
  10080. caps "false"
  10081. noprefix "false"
  10082. \end_inset
  10083. ).
  10084. For analysis A, a
  10085. \begin_inset Quotes eld
  10086. \end_inset
  10087. basic
  10088. \begin_inset Quotes erd
  10089. \end_inset
  10090. linear modeling analysis was performed, compensating for known confounders
  10091. by including terms for the factor of interest (transplant status) as well
  10092. as the known biological confounders: sex, age, ethnicity, and diabetes.
  10093. Since some samples came from the same patients at different times, the
  10094. intra-patient correlation was modeled as a random effect, estimating a
  10095. shared correlation value across all probes
  10096. \begin_inset CommandInset citation
  10097. LatexCommand cite
  10098. key "Smyth2005a"
  10099. literal "false"
  10100. \end_inset
  10101. .
  10102. Then the linear model was fit, and the variance was modeled using empirical
  10103. Bayes squeezing toward the mean-variance trend
  10104. \begin_inset CommandInset citation
  10105. LatexCommand cite
  10106. key "Ritchie2015"
  10107. literal "false"
  10108. \end_inset
  10109. .
  10110. Finally, t-tests or F-tests were performed as appropriate for each test:
  10111. t-tests for single contrasts, and F-tests for multiple contrasts.
  10112. P-values were corrected for multiple testing using the
  10113. \begin_inset Flex Glossary Term
  10114. status open
  10115. \begin_layout Plain Layout
  10116. BH
  10117. \end_layout
  10118. \end_inset
  10119. procedure for
  10120. \begin_inset Flex Glossary Term
  10121. status open
  10122. \begin_layout Plain Layout
  10123. FDR
  10124. \end_layout
  10125. \end_inset
  10126. control
  10127. \begin_inset CommandInset citation
  10128. LatexCommand cite
  10129. key "Benjamini1995"
  10130. literal "false"
  10131. \end_inset
  10132. .
  10133. \end_layout
  10134. \begin_layout Standard
  10135. For the analysis B,
  10136. \begin_inset Flex Glossary Term
  10137. status open
  10138. \begin_layout Plain Layout
  10139. SVA
  10140. \end_layout
  10141. \end_inset
  10142. was used to infer additional unobserved sources of heterogeneity in the
  10143. data
  10144. \begin_inset CommandInset citation
  10145. LatexCommand cite
  10146. key "Leek2007"
  10147. literal "false"
  10148. \end_inset
  10149. .
  10150. These surrogate variables were added to the design matrix before fitting
  10151. the linear model.
  10152. In addition, sample quality weights were estimated from the data and used
  10153. during linear modeling to down-weight the contribution of highly variable
  10154. arrays while increasing the weight to arrays with lower variability
  10155. \begin_inset CommandInset citation
  10156. LatexCommand cite
  10157. key "Ritchie2006"
  10158. literal "false"
  10159. \end_inset
  10160. .
  10161. The remainder of the analysis proceeded as in analysis A.
  10162. For analysis C, the voom method was adapted to run on methylation array
  10163. data and used to model and correct for the mean-variance trend using individual
  10164. observation weights
  10165. \begin_inset CommandInset citation
  10166. LatexCommand cite
  10167. key "Law2014"
  10168. literal "false"
  10169. \end_inset
  10170. , which were combined with the sample weights
  10171. \begin_inset CommandInset citation
  10172. LatexCommand cite
  10173. key "Liu2015,Ritchie2006"
  10174. literal "false"
  10175. \end_inset
  10176. .
  10177. Each time weights were used, they were estimated once before estimating
  10178. the random effect correlation value, and then the weights were re-estimated
  10179. taking the random effect into account.
  10180. The remainder of the analysis proceeded as in analysis B.
  10181. \end_layout
  10182. \begin_layout Section
  10183. Results
  10184. \end_layout
  10185. \begin_layout Standard
  10186. \begin_inset Flex TODO Note (inline)
  10187. status open
  10188. \begin_layout Plain Layout
  10189. Improve subsection titles in this section.
  10190. \end_layout
  10191. \end_inset
  10192. \end_layout
  10193. \begin_layout Standard
  10194. \begin_inset Flex TODO Note (inline)
  10195. status open
  10196. \begin_layout Plain Layout
  10197. Reconsider subsection organization?
  10198. \end_layout
  10199. \end_inset
  10200. \end_layout
  10201. \begin_layout Subsection
  10202. Separate normalization with RMA introduces unwanted biases in classification
  10203. \end_layout
  10204. \begin_layout Standard
  10205. To demonstrate the problem with non-single-channel normalization methods,
  10206. we considered the problem of training a classifier to distinguish
  10207. \begin_inset Flex Glossary Term
  10208. status open
  10209. \begin_layout Plain Layout
  10210. TX
  10211. \end_layout
  10212. \end_inset
  10213. from
  10214. \begin_inset Flex Glossary Term
  10215. status open
  10216. \begin_layout Plain Layout
  10217. AR
  10218. \end_layout
  10219. \end_inset
  10220. using the samples from the internal set as training data, evaluating performanc
  10221. e on the external set.
  10222. First, training and evaluation were performed after normalizing all array
  10223. samples together as a single set using
  10224. \begin_inset Flex Glossary Term
  10225. status open
  10226. \begin_layout Plain Layout
  10227. RMA
  10228. \end_layout
  10229. \end_inset
  10230. , and second, the internal samples were normalized separately from the external
  10231. samples and the training and evaluation were repeated.
  10232. For each sample in the validation set, the classifier probabilities from
  10233. both classifiers were plotted against each other (Fig.
  10234. \begin_inset CommandInset ref
  10235. LatexCommand ref
  10236. reference "fig:Classifier-probabilities-RMA"
  10237. plural "false"
  10238. caps "false"
  10239. noprefix "false"
  10240. \end_inset
  10241. ).
  10242. As expected, separate normalization biases the classifier probabilities,
  10243. resulting in several misclassifications.
  10244. In this case, the bias from separate normalization causes the classifier
  10245. to assign a lower probability of
  10246. \begin_inset Flex Glossary Term
  10247. status open
  10248. \begin_layout Plain Layout
  10249. AR
  10250. \end_layout
  10251. \end_inset
  10252. to every sample.
  10253. \end_layout
  10254. \begin_layout Standard
  10255. \begin_inset Float figure
  10256. wide false
  10257. sideways false
  10258. status collapsed
  10259. \begin_layout Plain Layout
  10260. \align center
  10261. \begin_inset Graphics
  10262. filename graphics/PAM/predplot.pdf
  10263. lyxscale 50
  10264. width 60col%
  10265. groupId colwidth
  10266. \end_inset
  10267. \end_layout
  10268. \begin_layout Plain Layout
  10269. \begin_inset Caption Standard
  10270. \begin_layout Plain Layout
  10271. \begin_inset Argument 1
  10272. status collapsed
  10273. \begin_layout Plain Layout
  10274. Classifier probabilities on validation samples when normalized with RMA
  10275. together vs.
  10276. separately.
  10277. \end_layout
  10278. \end_inset
  10279. \begin_inset CommandInset label
  10280. LatexCommand label
  10281. name "fig:Classifier-probabilities-RMA"
  10282. \end_inset
  10283. \series bold
  10284. Classifier probabilities on validation samples when normalized with RMA
  10285. together vs.
  10286. separately.
  10287. \series default
  10288. The PAM classifier algorithm was trained on the training set of arrays to
  10289. distinguish AR from TX and then used to assign class probabilities to the
  10290. validation set.
  10291. The process was performed after normalizing all samples together and after
  10292. normalizing the training and test sets separately, and the class probabilities
  10293. assigned to each sample in the validation set were plotted against each
  10294. other.
  10295. Each axis indicates the posterior probability of AR assigned to a sample
  10296. by the classifier in the specified analysis.
  10297. The color of each point indicates the true classification of that sample.
  10298. \end_layout
  10299. \end_inset
  10300. \end_layout
  10301. \end_inset
  10302. \end_layout
  10303. \begin_layout Subsection
  10304. fRMA and SCAN maintain classification performance while eliminating dependence
  10305. on normalization strategy
  10306. \end_layout
  10307. \begin_layout Standard
  10308. For internal validation, the 6 methods' AUC values ranged from 0.816 to 0.891,
  10309. as shown in Table
  10310. \begin_inset CommandInset ref
  10311. LatexCommand ref
  10312. reference "tab:AUC-PAM"
  10313. plural "false"
  10314. caps "false"
  10315. noprefix "false"
  10316. \end_inset
  10317. .
  10318. Among the non-single-channel normalizations, dChip outperformed
  10319. \begin_inset Flex Glossary Term
  10320. status open
  10321. \begin_layout Plain Layout
  10322. RMA
  10323. \end_layout
  10324. \end_inset
  10325. , while
  10326. \begin_inset Flex Glossary Term
  10327. status open
  10328. \begin_layout Plain Layout
  10329. GRSN
  10330. \end_layout
  10331. \end_inset
  10332. reduced the
  10333. \begin_inset Flex Glossary Term
  10334. status open
  10335. \begin_layout Plain Layout
  10336. AUC
  10337. \end_layout
  10338. \end_inset
  10339. values for both dChip and
  10340. \begin_inset Flex Glossary Term
  10341. status open
  10342. \begin_layout Plain Layout
  10343. RMA
  10344. \end_layout
  10345. \end_inset
  10346. .
  10347. Both single-channel methods,
  10348. \begin_inset Flex Glossary Term
  10349. status open
  10350. \begin_layout Plain Layout
  10351. fRMA
  10352. \end_layout
  10353. \end_inset
  10354. and
  10355. \begin_inset Flex Glossary Term
  10356. status open
  10357. \begin_layout Plain Layout
  10358. SCAN
  10359. \end_layout
  10360. \end_inset
  10361. , slightly outperformed
  10362. \begin_inset Flex Glossary Term
  10363. status open
  10364. \begin_layout Plain Layout
  10365. RMA
  10366. \end_layout
  10367. \end_inset
  10368. , with
  10369. \begin_inset Flex Glossary Term
  10370. status open
  10371. \begin_layout Plain Layout
  10372. fRMA
  10373. \end_layout
  10374. \end_inset
  10375. ahead of
  10376. \begin_inset Flex Glossary Term
  10377. status open
  10378. \begin_layout Plain Layout
  10379. SCAN
  10380. \end_layout
  10381. \end_inset
  10382. .
  10383. However, the difference between
  10384. \begin_inset Flex Glossary Term
  10385. status open
  10386. \begin_layout Plain Layout
  10387. RMA
  10388. \end_layout
  10389. \end_inset
  10390. and
  10391. \begin_inset Flex Glossary Term
  10392. status open
  10393. \begin_layout Plain Layout
  10394. fRMA
  10395. \end_layout
  10396. \end_inset
  10397. is still quite small.
  10398. Figure
  10399. \begin_inset CommandInset ref
  10400. LatexCommand ref
  10401. reference "fig:ROC-PAM-int"
  10402. plural "false"
  10403. caps "false"
  10404. noprefix "false"
  10405. \end_inset
  10406. shows that the
  10407. \begin_inset Flex Glossary Term
  10408. status open
  10409. \begin_layout Plain Layout
  10410. ROC
  10411. \end_layout
  10412. \end_inset
  10413. curves for
  10414. \begin_inset Flex Glossary Term
  10415. status open
  10416. \begin_layout Plain Layout
  10417. RMA
  10418. \end_layout
  10419. \end_inset
  10420. , dChip, and
  10421. \begin_inset Flex Glossary Term
  10422. status open
  10423. \begin_layout Plain Layout
  10424. fRMA
  10425. \end_layout
  10426. \end_inset
  10427. look very similar and relatively smooth, while both
  10428. \begin_inset Flex Glossary Term
  10429. status open
  10430. \begin_layout Plain Layout
  10431. GRSN
  10432. \end_layout
  10433. \end_inset
  10434. curves and the curve for
  10435. \begin_inset Flex Glossary Term
  10436. status open
  10437. \begin_layout Plain Layout
  10438. SCAN
  10439. \end_layout
  10440. \end_inset
  10441. have a more jagged appearance.
  10442. \end_layout
  10443. \begin_layout Standard
  10444. \begin_inset Float figure
  10445. wide false
  10446. sideways false
  10447. status collapsed
  10448. \begin_layout Plain Layout
  10449. \align center
  10450. \begin_inset Float figure
  10451. placement tb
  10452. wide false
  10453. sideways false
  10454. status open
  10455. \begin_layout Plain Layout
  10456. \align center
  10457. \begin_inset Graphics
  10458. filename graphics/PAM/ROC-TXvsAR-internal.pdf
  10459. lyxscale 50
  10460. height 40theight%
  10461. groupId roc-pam
  10462. \end_inset
  10463. \end_layout
  10464. \begin_layout Plain Layout
  10465. \begin_inset Caption Standard
  10466. \begin_layout Plain Layout
  10467. \begin_inset CommandInset label
  10468. LatexCommand label
  10469. name "fig:ROC-PAM-int"
  10470. \end_inset
  10471. ROC curves for PAM on internal validation data
  10472. \end_layout
  10473. \end_inset
  10474. \end_layout
  10475. \end_inset
  10476. \end_layout
  10477. \begin_layout Plain Layout
  10478. \align center
  10479. \begin_inset Float figure
  10480. placement tb
  10481. wide false
  10482. sideways false
  10483. status open
  10484. \begin_layout Plain Layout
  10485. \align center
  10486. \begin_inset Graphics
  10487. filename graphics/PAM/ROC-TXvsAR-external.pdf
  10488. lyxscale 50
  10489. height 40theight%
  10490. groupId roc-pam
  10491. \end_inset
  10492. \end_layout
  10493. \begin_layout Plain Layout
  10494. \begin_inset Caption Standard
  10495. \begin_layout Plain Layout
  10496. \begin_inset CommandInset label
  10497. LatexCommand label
  10498. name "fig:ROC-PAM-ext"
  10499. \end_inset
  10500. ROC curves for PAM on external validation data
  10501. \end_layout
  10502. \end_inset
  10503. \end_layout
  10504. \end_inset
  10505. \end_layout
  10506. \begin_layout Plain Layout
  10507. \begin_inset Caption Standard
  10508. \begin_layout Plain Layout
  10509. \begin_inset Argument 1
  10510. status collapsed
  10511. \begin_layout Plain Layout
  10512. ROC curves for PAM using different normalization strategies.
  10513. \end_layout
  10514. \end_inset
  10515. \begin_inset CommandInset label
  10516. LatexCommand label
  10517. name "fig:ROC-PAM-main"
  10518. \end_inset
  10519. \series bold
  10520. ROC curves for PAM using different normalization strategies.
  10521. \series default
  10522. ROC curves were generated for PAM classification of AR vs TX after 6 different
  10523. normalization strategies applied to the same data sets.
  10524. Only fRMA and SCAN are single-channel normalizations.
  10525. The other normalizations are for comparison.
  10526. \end_layout
  10527. \end_inset
  10528. \end_layout
  10529. \end_inset
  10530. \end_layout
  10531. \begin_layout Standard
  10532. \begin_inset Float table
  10533. wide false
  10534. sideways false
  10535. status collapsed
  10536. \begin_layout Plain Layout
  10537. \align center
  10538. \begin_inset Tabular
  10539. <lyxtabular version="3" rows="7" columns="4">
  10540. <features tabularvalignment="middle">
  10541. <column alignment="center" valignment="top">
  10542. <column alignment="center" valignment="top">
  10543. <column alignment="center" valignment="top">
  10544. <column alignment="center" valignment="top">
  10545. <row>
  10546. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  10547. \begin_inset Text
  10548. \begin_layout Plain Layout
  10549. \family roman
  10550. \series medium
  10551. \shape up
  10552. \size normal
  10553. \emph off
  10554. \bar no
  10555. \strikeout off
  10556. \xout off
  10557. \uuline off
  10558. \uwave off
  10559. \noun off
  10560. \color none
  10561. Normalization
  10562. \end_layout
  10563. \end_inset
  10564. </cell>
  10565. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  10566. \begin_inset Text
  10567. \begin_layout Plain Layout
  10568. Single-channel?
  10569. \end_layout
  10570. \end_inset
  10571. </cell>
  10572. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  10573. \begin_inset Text
  10574. \begin_layout Plain Layout
  10575. \family roman
  10576. \series medium
  10577. \shape up
  10578. \size normal
  10579. \emph off
  10580. \bar no
  10581. \strikeout off
  10582. \xout off
  10583. \uuline off
  10584. \uwave off
  10585. \noun off
  10586. \color none
  10587. Internal Val.
  10588. AUC
  10589. \end_layout
  10590. \end_inset
  10591. </cell>
  10592. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  10593. \begin_inset Text
  10594. \begin_layout Plain Layout
  10595. External Val.
  10596. AUC
  10597. \end_layout
  10598. \end_inset
  10599. </cell>
  10600. </row>
  10601. <row>
  10602. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  10603. \begin_inset Text
  10604. \begin_layout Plain Layout
  10605. \family roman
  10606. \series medium
  10607. \shape up
  10608. \size normal
  10609. \emph off
  10610. \bar no
  10611. \strikeout off
  10612. \xout off
  10613. \uuline off
  10614. \uwave off
  10615. \noun off
  10616. \color none
  10617. RMA
  10618. \end_layout
  10619. \end_inset
  10620. </cell>
  10621. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  10622. \begin_inset Text
  10623. \begin_layout Plain Layout
  10624. No
  10625. \end_layout
  10626. \end_inset
  10627. </cell>
  10628. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  10629. \begin_inset Text
  10630. \begin_layout Plain Layout
  10631. \family roman
  10632. \series medium
  10633. \shape up
  10634. \size normal
  10635. \emph off
  10636. \bar no
  10637. \strikeout off
  10638. \xout off
  10639. \uuline off
  10640. \uwave off
  10641. \noun off
  10642. \color none
  10643. 0.852
  10644. \end_layout
  10645. \end_inset
  10646. </cell>
  10647. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  10648. \begin_inset Text
  10649. \begin_layout Plain Layout
  10650. \family roman
  10651. \series medium
  10652. \shape up
  10653. \size normal
  10654. \emph off
  10655. \bar no
  10656. \strikeout off
  10657. \xout off
  10658. \uuline off
  10659. \uwave off
  10660. \noun off
  10661. \color none
  10662. 0.713
  10663. \end_layout
  10664. \end_inset
  10665. </cell>
  10666. </row>
  10667. <row>
  10668. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  10669. \begin_inset Text
  10670. \begin_layout Plain Layout
  10671. \family roman
  10672. \series medium
  10673. \shape up
  10674. \size normal
  10675. \emph off
  10676. \bar no
  10677. \strikeout off
  10678. \xout off
  10679. \uuline off
  10680. \uwave off
  10681. \noun off
  10682. \color none
  10683. dChip
  10684. \end_layout
  10685. \end_inset
  10686. </cell>
  10687. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  10688. \begin_inset Text
  10689. \begin_layout Plain Layout
  10690. No
  10691. \end_layout
  10692. \end_inset
  10693. </cell>
  10694. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  10695. \begin_inset Text
  10696. \begin_layout Plain Layout
  10697. \family roman
  10698. \series medium
  10699. \shape up
  10700. \size normal
  10701. \emph off
  10702. \bar no
  10703. \strikeout off
  10704. \xout off
  10705. \uuline off
  10706. \uwave off
  10707. \noun off
  10708. \color none
  10709. 0.891
  10710. \end_layout
  10711. \end_inset
  10712. </cell>
  10713. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  10714. \begin_inset Text
  10715. \begin_layout Plain Layout
  10716. \family roman
  10717. \series medium
  10718. \shape up
  10719. \size normal
  10720. \emph off
  10721. \bar no
  10722. \strikeout off
  10723. \xout off
  10724. \uuline off
  10725. \uwave off
  10726. \noun off
  10727. \color none
  10728. 0.657
  10729. \end_layout
  10730. \end_inset
  10731. </cell>
  10732. </row>
  10733. <row>
  10734. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  10735. \begin_inset Text
  10736. \begin_layout Plain Layout
  10737. \family roman
  10738. \series medium
  10739. \shape up
  10740. \size normal
  10741. \emph off
  10742. \bar no
  10743. \strikeout off
  10744. \xout off
  10745. \uuline off
  10746. \uwave off
  10747. \noun off
  10748. \color none
  10749. RMA + GRSN
  10750. \end_layout
  10751. \end_inset
  10752. </cell>
  10753. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  10754. \begin_inset Text
  10755. \begin_layout Plain Layout
  10756. No
  10757. \end_layout
  10758. \end_inset
  10759. </cell>
  10760. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  10761. \begin_inset Text
  10762. \begin_layout Plain Layout
  10763. \family roman
  10764. \series medium
  10765. \shape up
  10766. \size normal
  10767. \emph off
  10768. \bar no
  10769. \strikeout off
  10770. \xout off
  10771. \uuline off
  10772. \uwave off
  10773. \noun off
  10774. \color none
  10775. 0.816
  10776. \end_layout
  10777. \end_inset
  10778. </cell>
  10779. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  10780. \begin_inset Text
  10781. \begin_layout Plain Layout
  10782. \family roman
  10783. \series medium
  10784. \shape up
  10785. \size normal
  10786. \emph off
  10787. \bar no
  10788. \strikeout off
  10789. \xout off
  10790. \uuline off
  10791. \uwave off
  10792. \noun off
  10793. \color none
  10794. 0.750
  10795. \end_layout
  10796. \end_inset
  10797. </cell>
  10798. </row>
  10799. <row>
  10800. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  10801. \begin_inset Text
  10802. \begin_layout Plain Layout
  10803. \family roman
  10804. \series medium
  10805. \shape up
  10806. \size normal
  10807. \emph off
  10808. \bar no
  10809. \strikeout off
  10810. \xout off
  10811. \uuline off
  10812. \uwave off
  10813. \noun off
  10814. \color none
  10815. dChip + GRSN
  10816. \end_layout
  10817. \end_inset
  10818. </cell>
  10819. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  10820. \begin_inset Text
  10821. \begin_layout Plain Layout
  10822. No
  10823. \end_layout
  10824. \end_inset
  10825. </cell>
  10826. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  10827. \begin_inset Text
  10828. \begin_layout Plain Layout
  10829. \family roman
  10830. \series medium
  10831. \shape up
  10832. \size normal
  10833. \emph off
  10834. \bar no
  10835. \strikeout off
  10836. \xout off
  10837. \uuline off
  10838. \uwave off
  10839. \noun off
  10840. \color none
  10841. 0.875
  10842. \end_layout
  10843. \end_inset
  10844. </cell>
  10845. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  10846. \begin_inset Text
  10847. \begin_layout Plain Layout
  10848. \family roman
  10849. \series medium
  10850. \shape up
  10851. \size normal
  10852. \emph off
  10853. \bar no
  10854. \strikeout off
  10855. \xout off
  10856. \uuline off
  10857. \uwave off
  10858. \noun off
  10859. \color none
  10860. 0.642
  10861. \end_layout
  10862. \end_inset
  10863. </cell>
  10864. </row>
  10865. <row>
  10866. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  10867. \begin_inset Text
  10868. \begin_layout Plain Layout
  10869. \family roman
  10870. \series medium
  10871. \shape up
  10872. \size normal
  10873. \emph off
  10874. \bar no
  10875. \strikeout off
  10876. \xout off
  10877. \uuline off
  10878. \uwave off
  10879. \noun off
  10880. \color none
  10881. fRMA
  10882. \end_layout
  10883. \end_inset
  10884. </cell>
  10885. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  10886. \begin_inset Text
  10887. \begin_layout Plain Layout
  10888. Yes
  10889. \end_layout
  10890. \end_inset
  10891. </cell>
  10892. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  10893. \begin_inset Text
  10894. \begin_layout Plain Layout
  10895. \family roman
  10896. \series medium
  10897. \shape up
  10898. \size normal
  10899. \emph off
  10900. \bar no
  10901. \strikeout off
  10902. \xout off
  10903. \uuline off
  10904. \uwave off
  10905. \noun off
  10906. \color none
  10907. 0.863
  10908. \end_layout
  10909. \end_inset
  10910. </cell>
  10911. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  10912. \begin_inset Text
  10913. \begin_layout Plain Layout
  10914. \family roman
  10915. \series medium
  10916. \shape up
  10917. \size normal
  10918. \emph off
  10919. \bar no
  10920. \strikeout off
  10921. \xout off
  10922. \uuline off
  10923. \uwave off
  10924. \noun off
  10925. \color none
  10926. 0.718
  10927. \end_layout
  10928. \end_inset
  10929. </cell>
  10930. </row>
  10931. <row>
  10932. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  10933. \begin_inset Text
  10934. \begin_layout Plain Layout
  10935. \family roman
  10936. \series medium
  10937. \shape up
  10938. \size normal
  10939. \emph off
  10940. \bar no
  10941. \strikeout off
  10942. \xout off
  10943. \uuline off
  10944. \uwave off
  10945. \noun off
  10946. \color none
  10947. SCAN
  10948. \end_layout
  10949. \end_inset
  10950. </cell>
  10951. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  10952. \begin_inset Text
  10953. \begin_layout Plain Layout
  10954. Yes
  10955. \end_layout
  10956. \end_inset
  10957. </cell>
  10958. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  10959. \begin_inset Text
  10960. \begin_layout Plain Layout
  10961. \family roman
  10962. \series medium
  10963. \shape up
  10964. \size normal
  10965. \emph off
  10966. \bar no
  10967. \strikeout off
  10968. \xout off
  10969. \uuline off
  10970. \uwave off
  10971. \noun off
  10972. \color none
  10973. 0.853
  10974. \end_layout
  10975. \end_inset
  10976. </cell>
  10977. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  10978. \begin_inset Text
  10979. \begin_layout Plain Layout
  10980. \family roman
  10981. \series medium
  10982. \shape up
  10983. \size normal
  10984. \emph off
  10985. \bar no
  10986. \strikeout off
  10987. \xout off
  10988. \uuline off
  10989. \uwave off
  10990. \noun off
  10991. \color none
  10992. 0.689
  10993. \end_layout
  10994. \end_inset
  10995. </cell>
  10996. </row>
  10997. </lyxtabular>
  10998. \end_inset
  10999. \end_layout
  11000. \begin_layout Plain Layout
  11001. \begin_inset Caption Standard
  11002. \begin_layout Plain Layout
  11003. \begin_inset Argument 1
  11004. status collapsed
  11005. \begin_layout Plain Layout
  11006. ROC curve AUC values for internal and external validation with 6 different
  11007. normalization strategies.
  11008. \end_layout
  11009. \end_inset
  11010. \begin_inset CommandInset label
  11011. LatexCommand label
  11012. name "tab:AUC-PAM"
  11013. \end_inset
  11014. \series bold
  11015. ROC curve AUC values for internal and external validation with 6 different
  11016. normalization strategies.
  11017. \series default
  11018. These AUC values correspond to the ROC curves in Figure
  11019. \begin_inset CommandInset ref
  11020. LatexCommand ref
  11021. reference "fig:ROC-PAM-main"
  11022. plural "false"
  11023. caps "false"
  11024. noprefix "false"
  11025. \end_inset
  11026. .
  11027. \end_layout
  11028. \end_inset
  11029. \end_layout
  11030. \end_inset
  11031. \end_layout
  11032. \begin_layout Standard
  11033. For external validation, as expected, all the
  11034. \begin_inset Flex Glossary Term
  11035. status open
  11036. \begin_layout Plain Layout
  11037. AUC
  11038. \end_layout
  11039. \end_inset
  11040. values are lower than the internal validations, ranging from 0.642 to 0.750
  11041. (Table
  11042. \begin_inset CommandInset ref
  11043. LatexCommand ref
  11044. reference "tab:AUC-PAM"
  11045. plural "false"
  11046. caps "false"
  11047. noprefix "false"
  11048. \end_inset
  11049. ).
  11050. With or without
  11051. \begin_inset Flex Glossary Term
  11052. status open
  11053. \begin_layout Plain Layout
  11054. GRSN
  11055. \end_layout
  11056. \end_inset
  11057. ,
  11058. \begin_inset Flex Glossary Term
  11059. status open
  11060. \begin_layout Plain Layout
  11061. RMA
  11062. \end_layout
  11063. \end_inset
  11064. shows its dominance over dChip in this more challenging test.
  11065. Unlike in the internal validation,
  11066. \begin_inset Flex Glossary Term
  11067. status open
  11068. \begin_layout Plain Layout
  11069. GRSN
  11070. \end_layout
  11071. \end_inset
  11072. actually improves the classifier performance for
  11073. \begin_inset Flex Glossary Term
  11074. status open
  11075. \begin_layout Plain Layout
  11076. RMA
  11077. \end_layout
  11078. \end_inset
  11079. , although it does not for dChip.
  11080. Once again, both single-channel methods perform about on par with
  11081. \begin_inset Flex Glossary Term
  11082. status open
  11083. \begin_layout Plain Layout
  11084. RMA
  11085. \end_layout
  11086. \end_inset
  11087. , with
  11088. \begin_inset Flex Glossary Term
  11089. status open
  11090. \begin_layout Plain Layout
  11091. fRMA
  11092. \end_layout
  11093. \end_inset
  11094. performing slightly better and
  11095. \begin_inset Flex Glossary Term
  11096. status open
  11097. \begin_layout Plain Layout
  11098. SCAN
  11099. \end_layout
  11100. \end_inset
  11101. performing a bit worse.
  11102. Figure
  11103. \begin_inset CommandInset ref
  11104. LatexCommand ref
  11105. reference "fig:ROC-PAM-ext"
  11106. plural "false"
  11107. caps "false"
  11108. noprefix "false"
  11109. \end_inset
  11110. shows the
  11111. \begin_inset Flex Glossary Term
  11112. status open
  11113. \begin_layout Plain Layout
  11114. ROC
  11115. \end_layout
  11116. \end_inset
  11117. curves for the external validation test.
  11118. As expected, none of them are as clean-looking as the internal validation
  11119. \begin_inset Flex Glossary Term
  11120. status open
  11121. \begin_layout Plain Layout
  11122. ROC
  11123. \end_layout
  11124. \end_inset
  11125. curves.
  11126. The curves for
  11127. \begin_inset Flex Glossary Term
  11128. status open
  11129. \begin_layout Plain Layout
  11130. RMA
  11131. \end_layout
  11132. \end_inset
  11133. , RMA+GRSN, and
  11134. \begin_inset Flex Glossary Term
  11135. status open
  11136. \begin_layout Plain Layout
  11137. fRMA
  11138. \end_layout
  11139. \end_inset
  11140. all look similar, while the other curves look more divergent.
  11141. \end_layout
  11142. \begin_layout Subsection
  11143. fRMA with custom-generated vectors enables single-channel normalization
  11144. on hthgu133pluspm platform
  11145. \end_layout
  11146. \begin_layout Standard
  11147. In order to enable use of
  11148. \begin_inset Flex Glossary Term
  11149. status open
  11150. \begin_layout Plain Layout
  11151. fRMA
  11152. \end_layout
  11153. \end_inset
  11154. to normalize hthgu133pluspm, a custom set of
  11155. \begin_inset Flex Glossary Term
  11156. status open
  11157. \begin_layout Plain Layout
  11158. fRMA
  11159. \end_layout
  11160. \end_inset
  11161. vectors was created.
  11162. First, an appropriate batch size was chosen by looking at the number of
  11163. batches and number of samples included as a function of batch size (Figure
  11164. \begin_inset CommandInset ref
  11165. LatexCommand ref
  11166. reference "fig:frmatools-batch-size"
  11167. plural "false"
  11168. caps "false"
  11169. noprefix "false"
  11170. \end_inset
  11171. ).
  11172. For a given batch size, all batches with fewer samples that the chosen
  11173. size must be ignored during training, while larger batches must be randomly
  11174. downsampled to the chosen size.
  11175. Hence, the number of samples included for a given batch size equals the
  11176. batch size times the number of batches with at least that many samples.
  11177. From Figure
  11178. \begin_inset CommandInset ref
  11179. LatexCommand ref
  11180. reference "fig:batch-size-samples"
  11181. plural "false"
  11182. caps "false"
  11183. noprefix "false"
  11184. \end_inset
  11185. , it is apparent that a batch size of 8 maximizes the number of samples
  11186. included in training.
  11187. Increasing the batch size beyond this causes too many smaller batches to
  11188. be excluded, reducing the total number of samples for both tissue types.
  11189. However, a batch size of 8 is not necessarily optimal.
  11190. The article introducing frmaTools concluded that it was highly advantageous
  11191. to use a smaller batch size in order to include more batches, even at the
  11192. cost of including fewer total samples in training
  11193. \begin_inset CommandInset citation
  11194. LatexCommand cite
  11195. key "McCall2011"
  11196. literal "false"
  11197. \end_inset
  11198. .
  11199. To strike an appropriate balance between more batches and more samples,
  11200. a batch size of 5 was chosen.
  11201. For both blood and biopsy samples, this increased the number of batches
  11202. included by 10, with only a modest reduction in the number of samples compared
  11203. to a batch size of 8.
  11204. With a batch size of 5, 26 batches of biopsy samples and 46 batches of
  11205. blood samples were available.
  11206. \end_layout
  11207. \begin_layout Standard
  11208. \begin_inset Float figure
  11209. wide false
  11210. sideways false
  11211. status collapsed
  11212. \begin_layout Plain Layout
  11213. \align center
  11214. \begin_inset Float figure
  11215. placement tb
  11216. wide false
  11217. sideways false
  11218. status collapsed
  11219. \begin_layout Plain Layout
  11220. \align center
  11221. \begin_inset Graphics
  11222. filename graphics/frma-pax-bx/batchsize_batches.pdf
  11223. lyxscale 50
  11224. height 35theight%
  11225. groupId frmatools-subfig
  11226. \end_inset
  11227. \end_layout
  11228. \begin_layout Plain Layout
  11229. \begin_inset Caption Standard
  11230. \begin_layout Plain Layout
  11231. \begin_inset CommandInset label
  11232. LatexCommand label
  11233. name "fig:batch-size-batches"
  11234. \end_inset
  11235. \series bold
  11236. Number of batches usable in fRMA probe weight learning as a function of
  11237. batch size.
  11238. \end_layout
  11239. \end_inset
  11240. \end_layout
  11241. \end_inset
  11242. \end_layout
  11243. \begin_layout Plain Layout
  11244. \align center
  11245. \begin_inset Float figure
  11246. placement tb
  11247. wide false
  11248. sideways false
  11249. status collapsed
  11250. \begin_layout Plain Layout
  11251. \align center
  11252. \begin_inset Graphics
  11253. filename graphics/frma-pax-bx/batchsize_samples.pdf
  11254. lyxscale 50
  11255. height 35theight%
  11256. groupId frmatools-subfig
  11257. \end_inset
  11258. \end_layout
  11259. \begin_layout Plain Layout
  11260. \begin_inset Caption Standard
  11261. \begin_layout Plain Layout
  11262. \begin_inset CommandInset label
  11263. LatexCommand label
  11264. name "fig:batch-size-samples"
  11265. \end_inset
  11266. \series bold
  11267. Number of samples usable in fRMA probe weight learning as a function of
  11268. batch size.
  11269. \end_layout
  11270. \end_inset
  11271. \end_layout
  11272. \end_inset
  11273. \end_layout
  11274. \begin_layout Plain Layout
  11275. \begin_inset Caption Standard
  11276. \begin_layout Plain Layout
  11277. \begin_inset Argument 1
  11278. status collapsed
  11279. \begin_layout Plain Layout
  11280. Effect of batch size selection on number of batches and number of samples
  11281. included in fRMA probe weight learning.
  11282. \end_layout
  11283. \end_inset
  11284. \begin_inset CommandInset label
  11285. LatexCommand label
  11286. name "fig:frmatools-batch-size"
  11287. \end_inset
  11288. \series bold
  11289. Effect of batch size selection on number of batches and number of samples
  11290. included in fRMA probe weight learning.
  11291. \series default
  11292. For batch sizes ranging from 3 to 15, the number of batches (a) and samples
  11293. (b) included in probe weight training were plotted for biopsy (BX) and
  11294. blood (PAX) samples.
  11295. The selected batch size, 5, is marked with a dotted vertical line.
  11296. \end_layout
  11297. \end_inset
  11298. \end_layout
  11299. \end_inset
  11300. \end_layout
  11301. \begin_layout Standard
  11302. Since
  11303. \begin_inset Flex Glossary Term
  11304. status open
  11305. \begin_layout Plain Layout
  11306. fRMA
  11307. \end_layout
  11308. \end_inset
  11309. training requires equal-size batches, larger batches are downsampled randomly.
  11310. This introduces a nondeterministic step in the generation of normalization
  11311. vectors.
  11312. To show that this randomness does not substantially change the outcome,
  11313. the random downsampling and subsequent vector learning was repeated 5 times,
  11314. with a different random seed each time.
  11315. 20 samples were selected at random as a test set and normalized with each
  11316. of the 5 sets of
  11317. \begin_inset Flex Glossary Term
  11318. status open
  11319. \begin_layout Plain Layout
  11320. fRMA
  11321. \end_layout
  11322. \end_inset
  11323. normalization vectors as well as ordinary RMA, and the normalized expression
  11324. values were compared across normalizations.
  11325. Figure
  11326. \begin_inset CommandInset ref
  11327. LatexCommand ref
  11328. reference "fig:m-bx-violin"
  11329. plural "false"
  11330. caps "false"
  11331. noprefix "false"
  11332. \end_inset
  11333. shows a summary of these comparisons for biopsy samples.
  11334. Comparing RMA to each of the 5
  11335. \begin_inset Flex Glossary Term
  11336. status open
  11337. \begin_layout Plain Layout
  11338. fRMA
  11339. \end_layout
  11340. \end_inset
  11341. normalizations, the distribution of log ratios is somewhat wide, indicating
  11342. that the normalizations disagree on the expression values of a fair number
  11343. of probe sets.
  11344. In contrast, comparisons of
  11345. \begin_inset Flex Glossary Term
  11346. status open
  11347. \begin_layout Plain Layout
  11348. fRMA
  11349. \end_layout
  11350. \end_inset
  11351. against
  11352. \begin_inset Flex Glossary Term
  11353. status open
  11354. \begin_layout Plain Layout
  11355. fRMA
  11356. \end_layout
  11357. \end_inset
  11358. , the vast majority of probe sets have very small log ratios, indicating
  11359. a very high agreement between the normalized values generated by the two
  11360. normalizations.
  11361. This shows that the
  11362. \begin_inset Flex Glossary Term
  11363. status open
  11364. \begin_layout Plain Layout
  11365. fRMA
  11366. \end_layout
  11367. \end_inset
  11368. normalization's behavior is not very sensitive to the random downsampling
  11369. of larger batches during training.
  11370. \end_layout
  11371. \begin_layout Standard
  11372. \begin_inset Float figure
  11373. wide false
  11374. sideways false
  11375. status collapsed
  11376. \begin_layout Plain Layout
  11377. \align center
  11378. \begin_inset Graphics
  11379. filename graphics/frma-pax-bx/M-BX-violin.pdf
  11380. lyxscale 40
  11381. height 90theight%
  11382. groupId m-violin
  11383. \end_inset
  11384. \end_layout
  11385. \begin_layout Plain Layout
  11386. \begin_inset Caption Standard
  11387. \begin_layout Plain Layout
  11388. \begin_inset Argument 1
  11389. status collapsed
  11390. \begin_layout Plain Layout
  11391. Violin plot of log ratios between normalizations for 20 biopsy samples.
  11392. \end_layout
  11393. \end_inset
  11394. \begin_inset CommandInset label
  11395. LatexCommand label
  11396. name "fig:m-bx-violin"
  11397. \end_inset
  11398. \series bold
  11399. Violin plot of log ratios between normalizations for 20 biopsy samples.
  11400. \series default
  11401. Each of 20 randomly selected samples was normalized with RMA and with 5
  11402. different sets of fRMA vectors.
  11403. The distribution of log ratios between normalized expression values, aggregated
  11404. across all 20 arrays, was plotted for each pair of normalizations.
  11405. \end_layout
  11406. \end_inset
  11407. \end_layout
  11408. \end_inset
  11409. \end_layout
  11410. \begin_layout Standard
  11411. \begin_inset Float figure
  11412. wide false
  11413. sideways false
  11414. status collapsed
  11415. \begin_layout Plain Layout
  11416. \align center
  11417. \begin_inset Graphics
  11418. filename graphics/frma-pax-bx/M-PAX-violin.pdf
  11419. lyxscale 40
  11420. height 90theight%
  11421. groupId m-violin
  11422. \end_inset
  11423. \end_layout
  11424. \begin_layout Plain Layout
  11425. \begin_inset Caption Standard
  11426. \begin_layout Plain Layout
  11427. \begin_inset CommandInset label
  11428. LatexCommand label
  11429. name "fig:m-pax-violin"
  11430. \end_inset
  11431. \begin_inset Argument 1
  11432. status open
  11433. \begin_layout Plain Layout
  11434. Violin plot of log ratios between normalizations for 20 blood samples.
  11435. \end_layout
  11436. \end_inset
  11437. \series bold
  11438. Violin plot of log ratios between normalizations for 20 blood samples.
  11439. \series default
  11440. Each of 20 randomly selected samples was normalized with RMA and with 5
  11441. different sets of fRMA vectors.
  11442. The distribution of log ratios between normalized expression values, aggregated
  11443. across all 20 arrays, was plotted for each pair of normalizations.
  11444. \end_layout
  11445. \end_inset
  11446. \end_layout
  11447. \end_inset
  11448. \end_layout
  11449. \begin_layout Standard
  11450. Figure
  11451. \begin_inset CommandInset ref
  11452. LatexCommand ref
  11453. reference "fig:ma-bx-rma-frma"
  11454. plural "false"
  11455. caps "false"
  11456. noprefix "false"
  11457. \end_inset
  11458. shows an MA plot of the RMA-normalized values against the fRMA-normalized
  11459. values for the same probe sets and arrays, corresponding to the first row
  11460. of Figure
  11461. \begin_inset CommandInset ref
  11462. LatexCommand ref
  11463. reference "fig:m-bx-violin"
  11464. plural "false"
  11465. caps "false"
  11466. noprefix "false"
  11467. \end_inset
  11468. .
  11469. This MA plot shows that not only is there a wide distribution of
  11470. \begin_inset Flex Glossary Term (pl)
  11471. status open
  11472. \begin_layout Plain Layout
  11473. M-value
  11474. \end_layout
  11475. \end_inset
  11476. , but the trend of
  11477. \begin_inset Flex Glossary Term (pl)
  11478. status open
  11479. \begin_layout Plain Layout
  11480. M-value
  11481. \end_layout
  11482. \end_inset
  11483. is dependent on the average normalized intensity.
  11484. This is expected, since the overall trend represents the differences in
  11485. the quantile normalization step.
  11486. When running
  11487. \begin_inset Flex Glossary Term
  11488. status open
  11489. \begin_layout Plain Layout
  11490. RMA
  11491. \end_layout
  11492. \end_inset
  11493. , only the quantiles for these specific 20 arrays are used, while for
  11494. \begin_inset Flex Glossary Term
  11495. status open
  11496. \begin_layout Plain Layout
  11497. fRMA
  11498. \end_layout
  11499. \end_inset
  11500. the quantile distribution is taking from all arrays used in training.
  11501. Figure
  11502. \begin_inset CommandInset ref
  11503. LatexCommand ref
  11504. reference "fig:ma-bx-frma-frma"
  11505. plural "false"
  11506. caps "false"
  11507. noprefix "false"
  11508. \end_inset
  11509. shows a similar MA plot comparing 2 different
  11510. \begin_inset Flex Glossary Term
  11511. status open
  11512. \begin_layout Plain Layout
  11513. fRMA
  11514. \end_layout
  11515. \end_inset
  11516. normalizations, corresponding to the 6th row of Figure
  11517. \begin_inset CommandInset ref
  11518. LatexCommand ref
  11519. reference "fig:m-bx-violin"
  11520. plural "false"
  11521. caps "false"
  11522. noprefix "false"
  11523. \end_inset
  11524. .
  11525. The MA plot is very tightly centered around zero with no visible trend.
  11526. Figures
  11527. \begin_inset CommandInset ref
  11528. LatexCommand ref
  11529. reference "fig:m-pax-violin"
  11530. plural "false"
  11531. caps "false"
  11532. noprefix "false"
  11533. \end_inset
  11534. ,
  11535. \begin_inset CommandInset ref
  11536. LatexCommand ref
  11537. reference "fig:MA-PAX-rma-frma"
  11538. plural "false"
  11539. caps "false"
  11540. noprefix "false"
  11541. \end_inset
  11542. , and
  11543. \begin_inset CommandInset ref
  11544. LatexCommand ref
  11545. reference "fig:ma-bx-frma-frma"
  11546. plural "false"
  11547. caps "false"
  11548. noprefix "false"
  11549. \end_inset
  11550. show exactly the same information for the blood samples, once again comparing
  11551. the normalized expression values between normalizations for all probe sets
  11552. across 20 randomly selected test arrays.
  11553. Once again, there is a wider distribution of log ratios between RMA-normalized
  11554. values and fRMA-normalized, and a much tighter distribution when comparing
  11555. different
  11556. \begin_inset Flex Glossary Term
  11557. status open
  11558. \begin_layout Plain Layout
  11559. fRMA
  11560. \end_layout
  11561. \end_inset
  11562. normalizations to each other, indicating that the
  11563. \begin_inset Flex Glossary Term
  11564. status open
  11565. \begin_layout Plain Layout
  11566. fRMA
  11567. \end_layout
  11568. \end_inset
  11569. training process is robust to random batch sub-sampling for the blood samples
  11570. as well.
  11571. \end_layout
  11572. \begin_layout Standard
  11573. \begin_inset Float figure
  11574. wide false
  11575. sideways false
  11576. status collapsed
  11577. \begin_layout Plain Layout
  11578. \align center
  11579. \begin_inset Float figure
  11580. wide false
  11581. sideways false
  11582. status open
  11583. \begin_layout Plain Layout
  11584. \align center
  11585. \begin_inset Graphics
  11586. filename graphics/frma-pax-bx/MA-BX-RMA.fRMA-RASTER.png
  11587. lyxscale 10
  11588. width 45col%
  11589. groupId ma-frma
  11590. \end_inset
  11591. \end_layout
  11592. \begin_layout Plain Layout
  11593. \begin_inset Caption Standard
  11594. \begin_layout Plain Layout
  11595. \begin_inset CommandInset label
  11596. LatexCommand label
  11597. name "fig:ma-bx-rma-frma"
  11598. \end_inset
  11599. RMA vs.
  11600. fRMA for biopsy samples.
  11601. \end_layout
  11602. \end_inset
  11603. \end_layout
  11604. \end_inset
  11605. \begin_inset space \hfill{}
  11606. \end_inset
  11607. \begin_inset Float figure
  11608. wide false
  11609. sideways false
  11610. status collapsed
  11611. \begin_layout Plain Layout
  11612. \align center
  11613. \begin_inset Graphics
  11614. filename graphics/frma-pax-bx/MA-BX-fRMA.fRMA-RASTER.png
  11615. lyxscale 10
  11616. width 45col%
  11617. groupId ma-frma
  11618. \end_inset
  11619. \end_layout
  11620. \begin_layout Plain Layout
  11621. \begin_inset Caption Standard
  11622. \begin_layout Plain Layout
  11623. \begin_inset CommandInset label
  11624. LatexCommand label
  11625. name "fig:ma-bx-frma-frma"
  11626. \end_inset
  11627. fRMA vs fRMA for biopsy samples.
  11628. \end_layout
  11629. \end_inset
  11630. \end_layout
  11631. \end_inset
  11632. \end_layout
  11633. \begin_layout Plain Layout
  11634. \align center
  11635. \begin_inset Float figure
  11636. wide false
  11637. sideways false
  11638. status collapsed
  11639. \begin_layout Plain Layout
  11640. \align center
  11641. \begin_inset Graphics
  11642. filename graphics/frma-pax-bx/MA-PAX-RMA.fRMA-RASTER.png
  11643. lyxscale 10
  11644. width 45col%
  11645. groupId ma-frma
  11646. \end_inset
  11647. \end_layout
  11648. \begin_layout Plain Layout
  11649. \begin_inset Caption Standard
  11650. \begin_layout Plain Layout
  11651. \begin_inset CommandInset label
  11652. LatexCommand label
  11653. name "fig:MA-PAX-rma-frma"
  11654. \end_inset
  11655. RMA vs.
  11656. fRMA for blood samples.
  11657. \end_layout
  11658. \end_inset
  11659. \end_layout
  11660. \end_inset
  11661. \begin_inset space \hfill{}
  11662. \end_inset
  11663. \begin_inset Float figure
  11664. wide false
  11665. sideways false
  11666. status collapsed
  11667. \begin_layout Plain Layout
  11668. \align center
  11669. \begin_inset Graphics
  11670. filename graphics/frma-pax-bx/MA-PAX-fRMA.fRMA-RASTER.png
  11671. lyxscale 10
  11672. width 45col%
  11673. groupId ma-frma
  11674. \end_inset
  11675. \end_layout
  11676. \begin_layout Plain Layout
  11677. \begin_inset Caption Standard
  11678. \begin_layout Plain Layout
  11679. \begin_inset CommandInset label
  11680. LatexCommand label
  11681. name "fig:MA-PAX-frma-frma"
  11682. \end_inset
  11683. fRMA vs fRMA for blood samples.
  11684. \end_layout
  11685. \end_inset
  11686. \end_layout
  11687. \end_inset
  11688. \end_layout
  11689. \begin_layout Plain Layout
  11690. \begin_inset Caption Standard
  11691. \begin_layout Plain Layout
  11692. \begin_inset Argument 1
  11693. status collapsed
  11694. \begin_layout Plain Layout
  11695. Representative MA plots comparing RMA and custom fRMA normalizations.
  11696. \end_layout
  11697. \end_inset
  11698. \begin_inset CommandInset label
  11699. LatexCommand label
  11700. name "fig:Representative-MA-plots"
  11701. \end_inset
  11702. \series bold
  11703. Representative MA plots comparing RMA and custom fRMA normalizations.
  11704. \series default
  11705. For each plot, 20 samples were normalized using 2 different normalizations,
  11706. and then averages (A) and log ratios (M) were plotted between the two different
  11707. normalizations for every probe.
  11708. For the
  11709. \begin_inset Quotes eld
  11710. \end_inset
  11711. fRMA vs fRMA
  11712. \begin_inset Quotes erd
  11713. \end_inset
  11714. plots (b & d), two different fRMA normalizations using vectors from two
  11715. independent batch samplings were compared.
  11716. Density of points is represented by blue shading, and individual outlier
  11717. points are plotted.
  11718. \end_layout
  11719. \end_inset
  11720. \end_layout
  11721. \end_inset
  11722. \end_layout
  11723. \begin_layout Subsection
  11724. SVA, voom, and array weights improve model fit for methylation array data
  11725. \end_layout
  11726. \begin_layout Standard
  11727. Figure
  11728. \begin_inset CommandInset ref
  11729. LatexCommand ref
  11730. reference "fig:meanvar-basic"
  11731. plural "false"
  11732. caps "false"
  11733. noprefix "false"
  11734. \end_inset
  11735. shows the relationship between the mean
  11736. \begin_inset Flex Glossary Term
  11737. status open
  11738. \begin_layout Plain Layout
  11739. M-value
  11740. \end_layout
  11741. \end_inset
  11742. and the standard deviation calculated for each probe in the methylation
  11743. array data set.
  11744. A few features of the data are apparent.
  11745. First, the data are very strongly bimodal, with peaks in the density around
  11746. \begin_inset Flex Glossary Term (pl)
  11747. status open
  11748. \begin_layout Plain Layout
  11749. M-value
  11750. \end_layout
  11751. \end_inset
  11752. of +4 and -4.
  11753. These modes correspond to methylation sites that are nearly 100% methylated
  11754. and nearly 100% unmethylated, respectively.
  11755. The strong bimodality indicates that a majority of probes interrogate sites
  11756. that fall into one of these two categories.
  11757. The points in between these modes represent sites that are either partially
  11758. methylated in many samples, or are fully methylated in some samples and
  11759. fully unmethylated in other samples, or some combination.
  11760. The next visible feature of the data is the W-shaped variance trend.
  11761. The upticks in the variance trend on either side are expected, based on
  11762. the sigmoid transformation exaggerating small differences at extreme
  11763. \begin_inset Flex Glossary Term (pl)
  11764. status open
  11765. \begin_layout Plain Layout
  11766. M-value
  11767. \end_layout
  11768. \end_inset
  11769. (Figure
  11770. \begin_inset CommandInset ref
  11771. LatexCommand ref
  11772. reference "fig:Sigmoid-beta-m-mapping"
  11773. plural "false"
  11774. caps "false"
  11775. noprefix "false"
  11776. \end_inset
  11777. ).
  11778. However, the uptick in the center is interesting: it indicates that sites
  11779. that are not constitutively methylated or unmethylated have a higher variance.
  11780. This could be a genuine biological effect, or it could be spurious noise
  11781. that is only observable at sites with varying methylation.
  11782. \end_layout
  11783. \begin_layout Standard
  11784. \begin_inset ERT
  11785. status open
  11786. \begin_layout Plain Layout
  11787. \backslash
  11788. afterpage{
  11789. \end_layout
  11790. \begin_layout Plain Layout
  11791. \backslash
  11792. begin{landscape}
  11793. \end_layout
  11794. \end_inset
  11795. \end_layout
  11796. \begin_layout Standard
  11797. \begin_inset Float figure
  11798. wide false
  11799. sideways false
  11800. status open
  11801. \begin_layout Plain Layout
  11802. \begin_inset Flex TODO Note (inline)
  11803. status open
  11804. \begin_layout Plain Layout
  11805. Fix axis labels:
  11806. \begin_inset Quotes eld
  11807. \end_inset
  11808. log2 M-value
  11809. \begin_inset Quotes erd
  11810. \end_inset
  11811. is redundant because M-values are already log scale
  11812. \end_layout
  11813. \end_inset
  11814. \end_layout
  11815. \begin_layout Plain Layout
  11816. \begin_inset Float figure
  11817. wide false
  11818. sideways false
  11819. status collapsed
  11820. \begin_layout Plain Layout
  11821. \align center
  11822. \begin_inset Graphics
  11823. filename graphics/methylvoom/unadj.dupcor/meanvar-trends-PAGE1-CROP-RASTER.png
  11824. lyxscale 15
  11825. width 30col%
  11826. groupId voomaw-subfig
  11827. \end_inset
  11828. \end_layout
  11829. \begin_layout Plain Layout
  11830. \begin_inset Caption Standard
  11831. \begin_layout Plain Layout
  11832. \begin_inset CommandInset label
  11833. LatexCommand label
  11834. name "fig:meanvar-basic"
  11835. \end_inset
  11836. Mean-variance trend for analysis A.
  11837. \end_layout
  11838. \end_inset
  11839. \end_layout
  11840. \end_inset
  11841. \begin_inset space \hfill{}
  11842. \end_inset
  11843. \begin_inset Float figure
  11844. wide false
  11845. sideways false
  11846. status collapsed
  11847. \begin_layout Plain Layout
  11848. \align center
  11849. \begin_inset Graphics
  11850. filename graphics/methylvoom/unadj.dupcor.sva.aw/meanvar-trends-PAGE1-CROP-RASTER.png
  11851. lyxscale 15
  11852. width 30col%
  11853. groupId voomaw-subfig
  11854. \end_inset
  11855. \end_layout
  11856. \begin_layout Plain Layout
  11857. \begin_inset Caption Standard
  11858. \begin_layout Plain Layout
  11859. \begin_inset CommandInset label
  11860. LatexCommand label
  11861. name "fig:meanvar-sva-aw"
  11862. \end_inset
  11863. Mean-variance trend for analysis B.
  11864. \end_layout
  11865. \end_inset
  11866. \end_layout
  11867. \end_inset
  11868. \begin_inset space \hfill{}
  11869. \end_inset
  11870. \begin_inset Float figure
  11871. wide false
  11872. sideways false
  11873. status collapsed
  11874. \begin_layout Plain Layout
  11875. \align center
  11876. \begin_inset Graphics
  11877. filename graphics/methylvoom/unadj.dupcor.sva.voomaw/meanvar-trends-PAGE2-CROP-RASTER.png
  11878. lyxscale 15
  11879. width 30col%
  11880. groupId voomaw-subfig
  11881. \end_inset
  11882. \end_layout
  11883. \begin_layout Plain Layout
  11884. \begin_inset Caption Standard
  11885. \begin_layout Plain Layout
  11886. \begin_inset CommandInset label
  11887. LatexCommand label
  11888. name "fig:meanvar-sva-voomaw"
  11889. \end_inset
  11890. Mean-variance trend after voom modeling in analysis C.
  11891. \end_layout
  11892. \end_inset
  11893. \end_layout
  11894. \end_inset
  11895. \end_layout
  11896. \begin_layout Plain Layout
  11897. \begin_inset Caption Standard
  11898. \begin_layout Plain Layout
  11899. \begin_inset Argument 1
  11900. status collapsed
  11901. \begin_layout Plain Layout
  11902. Mean-variance trend modeling in methylation array data.
  11903. \end_layout
  11904. \end_inset
  11905. \begin_inset CommandInset label
  11906. LatexCommand label
  11907. name "fig:-Meanvar-trend-methyl"
  11908. \end_inset
  11909. \series bold
  11910. Mean-variance trend modeling in methylation array data.
  11911. \series default
  11912. The estimated
  11913. \begin_inset Formula $\log_{2}$
  11914. \end_inset
  11915. (standard deviation) for each probe is plotted against the probe's average
  11916. M-value across all samples as a black point, with some transparency to
  11917. make over-plotting more visible, since there are about 450,000 points.
  11918. Density of points is also indicated by the dark blue contour lines.
  11919. The prior variance trend estimated by eBayes is shown in light blue, while
  11920. the lowess trend of the points is shown in red.
  11921. \end_layout
  11922. \end_inset
  11923. \end_layout
  11924. \end_inset
  11925. \end_layout
  11926. \begin_layout Standard
  11927. \begin_inset ERT
  11928. status open
  11929. \begin_layout Plain Layout
  11930. \backslash
  11931. end{landscape}
  11932. \end_layout
  11933. \begin_layout Plain Layout
  11934. }
  11935. \end_layout
  11936. \end_inset
  11937. \end_layout
  11938. \begin_layout Standard
  11939. In Figure
  11940. \begin_inset CommandInset ref
  11941. LatexCommand ref
  11942. reference "fig:meanvar-sva-aw"
  11943. plural "false"
  11944. caps "false"
  11945. noprefix "false"
  11946. \end_inset
  11947. , we see the mean-variance trend for the same methylation array data, this
  11948. time with surrogate variables and sample quality weights estimated from
  11949. the data and included in the model.
  11950. As expected, the overall average variance is smaller, since the surrogate
  11951. variables account for some of the variance.
  11952. In addition, the uptick in variance in the middle of the
  11953. \begin_inset Flex Glossary Term
  11954. status open
  11955. \begin_layout Plain Layout
  11956. M-value
  11957. \end_layout
  11958. \end_inset
  11959. range has disappeared, turning the W shape into a wide U shape.
  11960. This indicates that the excess variance in the probes with intermediate
  11961. \begin_inset Flex Glossary Term (pl)
  11962. status open
  11963. \begin_layout Plain Layout
  11964. M-value
  11965. \end_layout
  11966. \end_inset
  11967. was explained by systematic variations not correlated with known covariates,
  11968. and these variations were modeled by the surrogate variables.
  11969. The result is a nearly flat variance trend for the entire intermediate
  11970. \begin_inset Flex Glossary Term
  11971. status open
  11972. \begin_layout Plain Layout
  11973. M-value
  11974. \end_layout
  11975. \end_inset
  11976. range from about -3 to +3.
  11977. Note that this corresponds closely to the range within which the
  11978. \begin_inset Flex Glossary Term
  11979. status open
  11980. \begin_layout Plain Layout
  11981. M-value
  11982. \end_layout
  11983. \end_inset
  11984. transformation shown in Figure
  11985. \begin_inset CommandInset ref
  11986. LatexCommand ref
  11987. reference "fig:Sigmoid-beta-m-mapping"
  11988. plural "false"
  11989. caps "false"
  11990. noprefix "false"
  11991. \end_inset
  11992. is nearly linear.
  11993. In contrast, the excess variance at the extremes (greater than +3 and less
  11994. than -3) was not
  11995. \begin_inset Quotes eld
  11996. \end_inset
  11997. absorbed
  11998. \begin_inset Quotes erd
  11999. \end_inset
  12000. by the surrogate variables and remains in the plot, indicating that this
  12001. variation has no systematic component: probes with extreme
  12002. \begin_inset Flex Glossary Term (pl)
  12003. status open
  12004. \begin_layout Plain Layout
  12005. M-value
  12006. \end_layout
  12007. \end_inset
  12008. are uniformly more variable across all samples, as expected.
  12009. \end_layout
  12010. \begin_layout Standard
  12011. Figure
  12012. \begin_inset CommandInset ref
  12013. LatexCommand ref
  12014. reference "fig:meanvar-sva-voomaw"
  12015. plural "false"
  12016. caps "false"
  12017. noprefix "false"
  12018. \end_inset
  12019. shows the mean-variance trend after fitting the model with the observation
  12020. weights assigned by voom based on the mean-variance trend shown in Figure
  12021. \begin_inset CommandInset ref
  12022. LatexCommand ref
  12023. reference "fig:meanvar-sva-aw"
  12024. plural "false"
  12025. caps "false"
  12026. noprefix "false"
  12027. \end_inset
  12028. .
  12029. As expected, the weights exactly counteract the trend in the data, resulting
  12030. in a nearly flat trend centered vertically at 1 (i.e.
  12031. 0 on the log scale).
  12032. This shows that the observations with extreme
  12033. \begin_inset Flex Glossary Term (pl)
  12034. status open
  12035. \begin_layout Plain Layout
  12036. M-value
  12037. \end_layout
  12038. \end_inset
  12039. have been appropriately down-weighted to account for the fact that the
  12040. noise in those observations has been amplified by the non-linear
  12041. \begin_inset Flex Glossary Term
  12042. status open
  12043. \begin_layout Plain Layout
  12044. M-value
  12045. \end_layout
  12046. \end_inset
  12047. transformation.
  12048. In turn, this gives relatively more weight to observations in the middle
  12049. region, which are more likely to correspond to probes measuring interesting
  12050. biology (not constitutively methylated or unmethylated).
  12051. \end_layout
  12052. \begin_layout Standard
  12053. To determine whether any of the known experimental factors had an impact
  12054. on data quality, the sample quality weights estimated from the data were
  12055. tested for association with each of the experimental factors (Table
  12056. \begin_inset CommandInset ref
  12057. LatexCommand ref
  12058. reference "tab:weight-covariate-tests"
  12059. plural "false"
  12060. caps "false"
  12061. noprefix "false"
  12062. \end_inset
  12063. ).
  12064. Diabetes diagnosis was found to have a potentially significant association
  12065. with the sample weights, with a t-test p-value of
  12066. \begin_inset Formula $1.06\times10^{-3}$
  12067. \end_inset
  12068. .
  12069. Figure
  12070. \begin_inset CommandInset ref
  12071. LatexCommand ref
  12072. reference "fig:diabetes-sample-weights"
  12073. plural "false"
  12074. caps "false"
  12075. noprefix "false"
  12076. \end_inset
  12077. shows the distribution of sample weights grouped by diabetes diagnosis.
  12078. The samples from patients with
  12079. \begin_inset Flex Glossary Term
  12080. status open
  12081. \begin_layout Plain Layout
  12082. T2D
  12083. \end_layout
  12084. \end_inset
  12085. were assigned significantly lower weights than those from patients with
  12086. \begin_inset Flex Glossary Term
  12087. status open
  12088. \begin_layout Plain Layout
  12089. T1D
  12090. \end_layout
  12091. \end_inset
  12092. .
  12093. This indicates that the
  12094. \begin_inset Flex Glossary Term
  12095. status open
  12096. \begin_layout Plain Layout
  12097. T2D
  12098. \end_layout
  12099. \end_inset
  12100. samples had an overall higher variance on average across all probes.
  12101. \end_layout
  12102. \begin_layout Standard
  12103. \begin_inset Float table
  12104. wide false
  12105. sideways false
  12106. status collapsed
  12107. \begin_layout Plain Layout
  12108. \align center
  12109. \begin_inset Tabular
  12110. <lyxtabular version="3" rows="5" columns="3">
  12111. <features tabularvalignment="middle">
  12112. <column alignment="center" valignment="top">
  12113. <column alignment="center" valignment="top">
  12114. <column alignment="center" valignment="top">
  12115. <row>
  12116. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  12117. \begin_inset Text
  12118. \begin_layout Plain Layout
  12119. Covariate
  12120. \end_layout
  12121. \end_inset
  12122. </cell>
  12123. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  12124. \begin_inset Text
  12125. \begin_layout Plain Layout
  12126. Test used
  12127. \end_layout
  12128. \end_inset
  12129. </cell>
  12130. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  12131. \begin_inset Text
  12132. \begin_layout Plain Layout
  12133. p-value
  12134. \end_layout
  12135. \end_inset
  12136. </cell>
  12137. </row>
  12138. <row>
  12139. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  12140. \begin_inset Text
  12141. \begin_layout Plain Layout
  12142. Transplant Status
  12143. \end_layout
  12144. \end_inset
  12145. </cell>
  12146. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  12147. \begin_inset Text
  12148. \begin_layout Plain Layout
  12149. F-test
  12150. \end_layout
  12151. \end_inset
  12152. </cell>
  12153. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  12154. \begin_inset Text
  12155. \begin_layout Plain Layout
  12156. 0.404
  12157. \end_layout
  12158. \end_inset
  12159. </cell>
  12160. </row>
  12161. <row>
  12162. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  12163. \begin_inset Text
  12164. \begin_layout Plain Layout
  12165. Diabetes Diagnosis
  12166. \end_layout
  12167. \end_inset
  12168. </cell>
  12169. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  12170. \begin_inset Text
  12171. \begin_layout Plain Layout
  12172. \emph on
  12173. t
  12174. \emph default
  12175. -test
  12176. \end_layout
  12177. \end_inset
  12178. </cell>
  12179. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  12180. \begin_inset Text
  12181. \begin_layout Plain Layout
  12182. 0.00106
  12183. \end_layout
  12184. \end_inset
  12185. </cell>
  12186. </row>
  12187. <row>
  12188. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  12189. \begin_inset Text
  12190. \begin_layout Plain Layout
  12191. Sex
  12192. \end_layout
  12193. \end_inset
  12194. </cell>
  12195. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  12196. \begin_inset Text
  12197. \begin_layout Plain Layout
  12198. \emph on
  12199. t
  12200. \emph default
  12201. -test
  12202. \end_layout
  12203. \end_inset
  12204. </cell>
  12205. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  12206. \begin_inset Text
  12207. \begin_layout Plain Layout
  12208. 0.148
  12209. \end_layout
  12210. \end_inset
  12211. </cell>
  12212. </row>
  12213. <row>
  12214. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  12215. \begin_inset Text
  12216. \begin_layout Plain Layout
  12217. Age
  12218. \end_layout
  12219. \end_inset
  12220. </cell>
  12221. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  12222. \begin_inset Text
  12223. \begin_layout Plain Layout
  12224. linear regression
  12225. \end_layout
  12226. \end_inset
  12227. </cell>
  12228. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  12229. \begin_inset Text
  12230. \begin_layout Plain Layout
  12231. 0.212
  12232. \end_layout
  12233. \end_inset
  12234. </cell>
  12235. </row>
  12236. </lyxtabular>
  12237. \end_inset
  12238. \end_layout
  12239. \begin_layout Plain Layout
  12240. \begin_inset Caption Standard
  12241. \begin_layout Plain Layout
  12242. \begin_inset Argument 1
  12243. status collapsed
  12244. \begin_layout Plain Layout
  12245. Association of sample weights with clinical covariates in methylation array
  12246. data.
  12247. \end_layout
  12248. \end_inset
  12249. \begin_inset CommandInset label
  12250. LatexCommand label
  12251. name "tab:weight-covariate-tests"
  12252. \end_inset
  12253. \series bold
  12254. Association of sample weights with clinical covariates in methylation array
  12255. data.
  12256. \series default
  12257. Computed sample quality log weights were tested for significant association
  12258. with each of the variables in the model (1st column).
  12259. An appropriate test was selected for each variable based on whether the
  12260. variable had 2 categories (
  12261. \emph on
  12262. t
  12263. \emph default
  12264. -test), had more than 2 categories (F-test), or was numeric (linear regression).
  12265. The test selected is shown in the 2nd column.
  12266. P-values for association with the log weights are shown in the 3rd column.
  12267. No multiple testing adjustment was performed for these p-values.
  12268. \end_layout
  12269. \end_inset
  12270. \end_layout
  12271. \end_inset
  12272. \end_layout
  12273. \begin_layout Standard
  12274. \begin_inset Float figure
  12275. wide false
  12276. sideways false
  12277. status collapsed
  12278. \begin_layout Plain Layout
  12279. \begin_inset Flex TODO Note (inline)
  12280. status open
  12281. \begin_layout Plain Layout
  12282. Redo the sample weight boxplot with notches, and remove fill colors
  12283. \end_layout
  12284. \end_inset
  12285. \end_layout
  12286. \begin_layout Plain Layout
  12287. \align center
  12288. \begin_inset Graphics
  12289. filename graphics/methylvoom/unadj.dupcor.sva.voomaw/sample-weights-PAGE3-CROP.pdf
  12290. lyxscale 50
  12291. width 60col%
  12292. groupId colwidth
  12293. \end_inset
  12294. \end_layout
  12295. \begin_layout Plain Layout
  12296. \begin_inset Caption Standard
  12297. \begin_layout Plain Layout
  12298. \begin_inset Argument 1
  12299. status collapsed
  12300. \begin_layout Plain Layout
  12301. Box-and-whiskers plot of sample quality weights grouped by diabetes diagnosis.
  12302. \end_layout
  12303. \end_inset
  12304. \begin_inset CommandInset label
  12305. LatexCommand label
  12306. name "fig:diabetes-sample-weights"
  12307. \end_inset
  12308. \series bold
  12309. Box-and-whiskers plot of sample quality weights grouped by diabetes diagnosis.
  12310. \series default
  12311. Samples were grouped based on diabetes diagnosis, and the distribution of
  12312. sample quality weights for each diagnosis was plotted as a box-and-whiskers
  12313. plot
  12314. \begin_inset CommandInset citation
  12315. LatexCommand cite
  12316. key "McGill1978"
  12317. literal "false"
  12318. \end_inset
  12319. .
  12320. \end_layout
  12321. \end_inset
  12322. \end_layout
  12323. \end_inset
  12324. \end_layout
  12325. \begin_layout Standard
  12326. Table
  12327. \begin_inset CommandInset ref
  12328. LatexCommand ref
  12329. reference "tab:methyl-num-signif"
  12330. plural "false"
  12331. caps "false"
  12332. noprefix "false"
  12333. \end_inset
  12334. shows the number of significantly differentially methylated probes reported
  12335. by each analysis for each comparison of interest at an
  12336. \begin_inset Flex Glossary Term
  12337. status open
  12338. \begin_layout Plain Layout
  12339. FDR
  12340. \end_layout
  12341. \end_inset
  12342. of 10%.
  12343. As expected, the more elaborate analyses, B and C, report more significant
  12344. probes than the more basic analysis A, consistent with the conclusions
  12345. above that the data contain hidden systematic variations that must be modeled.
  12346. Table
  12347. \begin_inset CommandInset ref
  12348. LatexCommand ref
  12349. reference "tab:methyl-est-nonnull"
  12350. plural "false"
  12351. caps "false"
  12352. noprefix "false"
  12353. \end_inset
  12354. shows the estimated number differentially methylated probes for each test
  12355. from each analysis.
  12356. This was computed by estimating the proportion of null hypotheses that
  12357. were true using the method of
  12358. \begin_inset CommandInset citation
  12359. LatexCommand cite
  12360. key "Phipson2013Thesis"
  12361. literal "false"
  12362. \end_inset
  12363. and subtracting that fraction from the total number of probes, yielding
  12364. an estimate of the number of null hypotheses that are false based on the
  12365. distribution of p-values across the entire dataset.
  12366. Note that this does not identify which null hypotheses should be rejected
  12367. (i.e.
  12368. which probes are significant); it only estimates the true number of such
  12369. probes.
  12370. Once again, analyses B and C result it much larger estimates for the number
  12371. of differentially methylated probes.
  12372. In this case, analysis C, the only analysis that includes voom, estimates
  12373. the largest number of differentially methylated probes for all 3 contrasts.
  12374. If the assumptions of all the methods employed hold, then this represents
  12375. a gain in statistical power over the simpler analysis A.
  12376. Figure
  12377. \begin_inset CommandInset ref
  12378. LatexCommand ref
  12379. reference "fig:meth-p-value-histograms"
  12380. plural "false"
  12381. caps "false"
  12382. noprefix "false"
  12383. \end_inset
  12384. shows the p-value distributions for each test, from which the numbers in
  12385. Table
  12386. \begin_inset CommandInset ref
  12387. LatexCommand ref
  12388. reference "tab:methyl-est-nonnull"
  12389. plural "false"
  12390. caps "false"
  12391. noprefix "false"
  12392. \end_inset
  12393. were generated.
  12394. The distributions for analysis A all have a dip in density near zero, which
  12395. is a strong sign of a poor model fit.
  12396. The histograms for analyses B and C are more well-behaved, with a uniform
  12397. component stretching all the way from 0 to 1 representing the probes for
  12398. which the null hypotheses is true (no differential methylation), and a
  12399. zero-biased component representing the probes for which the null hypothesis
  12400. is false (differentially methylated).
  12401. These histograms do not indicate any major issues with the model fit.
  12402. \end_layout
  12403. \begin_layout Standard
  12404. \begin_inset Float table
  12405. wide false
  12406. sideways false
  12407. status collapsed
  12408. \begin_layout Plain Layout
  12409. \align center
  12410. \begin_inset Flex TODO Note (inline)
  12411. status open
  12412. \begin_layout Plain Layout
  12413. Consider transposing these tables
  12414. \end_layout
  12415. \end_inset
  12416. \end_layout
  12417. \begin_layout Plain Layout
  12418. \begin_inset Float table
  12419. wide false
  12420. sideways false
  12421. status open
  12422. \begin_layout Plain Layout
  12423. \align center
  12424. \begin_inset Tabular
  12425. <lyxtabular version="3" rows="5" columns="4">
  12426. <features tabularvalignment="middle">
  12427. <column alignment="center" valignment="top">
  12428. <column alignment="center" valignment="top">
  12429. <column alignment="center" valignment="top">
  12430. <column alignment="center" valignment="top">
  12431. <row>
  12432. <cell alignment="center" valignment="top" usebox="none">
  12433. \begin_inset Text
  12434. \begin_layout Plain Layout
  12435. \end_layout
  12436. \end_inset
  12437. </cell>
  12438. <cell multicolumn="1" alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  12439. \begin_inset Text
  12440. \begin_layout Plain Layout
  12441. Analysis
  12442. \end_layout
  12443. \end_inset
  12444. </cell>
  12445. <cell multicolumn="2" alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  12446. \begin_inset Text
  12447. \begin_layout Plain Layout
  12448. \end_layout
  12449. \end_inset
  12450. </cell>
  12451. <cell multicolumn="2" alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  12452. \begin_inset Text
  12453. \begin_layout Plain Layout
  12454. \end_layout
  12455. \end_inset
  12456. </cell>
  12457. </row>
  12458. <row>
  12459. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  12460. \begin_inset Text
  12461. \begin_layout Plain Layout
  12462. Contrast
  12463. \end_layout
  12464. \end_inset
  12465. </cell>
  12466. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  12467. \begin_inset Text
  12468. \begin_layout Plain Layout
  12469. A
  12470. \end_layout
  12471. \end_inset
  12472. </cell>
  12473. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  12474. \begin_inset Text
  12475. \begin_layout Plain Layout
  12476. B
  12477. \end_layout
  12478. \end_inset
  12479. </cell>
  12480. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  12481. \begin_inset Text
  12482. \begin_layout Plain Layout
  12483. C
  12484. \end_layout
  12485. \end_inset
  12486. </cell>
  12487. </row>
  12488. <row>
  12489. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  12490. \begin_inset Text
  12491. \begin_layout Plain Layout
  12492. TX vs AR
  12493. \end_layout
  12494. \end_inset
  12495. </cell>
  12496. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  12497. \begin_inset Text
  12498. \begin_layout Plain Layout
  12499. 0
  12500. \end_layout
  12501. \end_inset
  12502. </cell>
  12503. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  12504. \begin_inset Text
  12505. \begin_layout Plain Layout
  12506. 25
  12507. \end_layout
  12508. \end_inset
  12509. </cell>
  12510. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  12511. \begin_inset Text
  12512. \begin_layout Plain Layout
  12513. 22
  12514. \end_layout
  12515. \end_inset
  12516. </cell>
  12517. </row>
  12518. <row>
  12519. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  12520. \begin_inset Text
  12521. \begin_layout Plain Layout
  12522. TX vs ADNR
  12523. \end_layout
  12524. \end_inset
  12525. </cell>
  12526. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  12527. \begin_inset Text
  12528. \begin_layout Plain Layout
  12529. 7
  12530. \end_layout
  12531. \end_inset
  12532. </cell>
  12533. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  12534. \begin_inset Text
  12535. \begin_layout Plain Layout
  12536. 338
  12537. \end_layout
  12538. \end_inset
  12539. </cell>
  12540. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  12541. \begin_inset Text
  12542. \begin_layout Plain Layout
  12543. 369
  12544. \end_layout
  12545. \end_inset
  12546. </cell>
  12547. </row>
  12548. <row>
  12549. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  12550. \begin_inset Text
  12551. \begin_layout Plain Layout
  12552. TX vs CAN
  12553. \end_layout
  12554. \end_inset
  12555. </cell>
  12556. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  12557. \begin_inset Text
  12558. \begin_layout Plain Layout
  12559. 0
  12560. \end_layout
  12561. \end_inset
  12562. </cell>
  12563. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  12564. \begin_inset Text
  12565. \begin_layout Plain Layout
  12566. 231
  12567. \end_layout
  12568. \end_inset
  12569. </cell>
  12570. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  12571. \begin_inset Text
  12572. \begin_layout Plain Layout
  12573. 278
  12574. \end_layout
  12575. \end_inset
  12576. </cell>
  12577. </row>
  12578. </lyxtabular>
  12579. \end_inset
  12580. \end_layout
  12581. \begin_layout Plain Layout
  12582. \begin_inset Caption Standard
  12583. \begin_layout Plain Layout
  12584. \begin_inset CommandInset label
  12585. LatexCommand label
  12586. name "tab:methyl-num-signif"
  12587. \end_inset
  12588. Number of probes significant at 10% FDR.
  12589. \end_layout
  12590. \end_inset
  12591. \end_layout
  12592. \end_inset
  12593. \begin_inset space \hfill{}
  12594. \end_inset
  12595. \begin_inset Float table
  12596. wide false
  12597. sideways false
  12598. status open
  12599. \begin_layout Plain Layout
  12600. \align center
  12601. \begin_inset Tabular
  12602. <lyxtabular version="3" rows="5" columns="4">
  12603. <features tabularvalignment="middle">
  12604. <column alignment="center" valignment="top">
  12605. <column alignment="center" valignment="top">
  12606. <column alignment="center" valignment="top">
  12607. <column alignment="center" valignment="top">
  12608. <row>
  12609. <cell alignment="center" valignment="top" usebox="none">
  12610. \begin_inset Text
  12611. \begin_layout Plain Layout
  12612. \end_layout
  12613. \end_inset
  12614. </cell>
  12615. <cell multicolumn="1" alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  12616. \begin_inset Text
  12617. \begin_layout Plain Layout
  12618. Analysis
  12619. \end_layout
  12620. \end_inset
  12621. </cell>
  12622. <cell multicolumn="2" alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  12623. \begin_inset Text
  12624. \begin_layout Plain Layout
  12625. \end_layout
  12626. \end_inset
  12627. </cell>
  12628. <cell multicolumn="2" alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  12629. \begin_inset Text
  12630. \begin_layout Plain Layout
  12631. \end_layout
  12632. \end_inset
  12633. </cell>
  12634. </row>
  12635. <row>
  12636. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  12637. \begin_inset Text
  12638. \begin_layout Plain Layout
  12639. Contrast
  12640. \end_layout
  12641. \end_inset
  12642. </cell>
  12643. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  12644. \begin_inset Text
  12645. \begin_layout Plain Layout
  12646. A
  12647. \end_layout
  12648. \end_inset
  12649. </cell>
  12650. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  12651. \begin_inset Text
  12652. \begin_layout Plain Layout
  12653. B
  12654. \end_layout
  12655. \end_inset
  12656. </cell>
  12657. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  12658. \begin_inset Text
  12659. \begin_layout Plain Layout
  12660. C
  12661. \end_layout
  12662. \end_inset
  12663. </cell>
  12664. </row>
  12665. <row>
  12666. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  12667. \begin_inset Text
  12668. \begin_layout Plain Layout
  12669. TX vs AR
  12670. \end_layout
  12671. \end_inset
  12672. </cell>
  12673. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  12674. \begin_inset Text
  12675. \begin_layout Plain Layout
  12676. 0
  12677. \end_layout
  12678. \end_inset
  12679. </cell>
  12680. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  12681. \begin_inset Text
  12682. \begin_layout Plain Layout
  12683. 10,063
  12684. \end_layout
  12685. \end_inset
  12686. </cell>
  12687. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  12688. \begin_inset Text
  12689. \begin_layout Plain Layout
  12690. 11,225
  12691. \end_layout
  12692. \end_inset
  12693. </cell>
  12694. </row>
  12695. <row>
  12696. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  12697. \begin_inset Text
  12698. \begin_layout Plain Layout
  12699. TX vs ADNR
  12700. \end_layout
  12701. \end_inset
  12702. </cell>
  12703. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  12704. \begin_inset Text
  12705. \begin_layout Plain Layout
  12706. 27
  12707. \end_layout
  12708. \end_inset
  12709. </cell>
  12710. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  12711. \begin_inset Text
  12712. \begin_layout Plain Layout
  12713. 12,674
  12714. \end_layout
  12715. \end_inset
  12716. </cell>
  12717. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  12718. \begin_inset Text
  12719. \begin_layout Plain Layout
  12720. 13,086
  12721. \end_layout
  12722. \end_inset
  12723. </cell>
  12724. </row>
  12725. <row>
  12726. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  12727. \begin_inset Text
  12728. \begin_layout Plain Layout
  12729. TX vs CAN
  12730. \end_layout
  12731. \end_inset
  12732. </cell>
  12733. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  12734. \begin_inset Text
  12735. \begin_layout Plain Layout
  12736. 966
  12737. \end_layout
  12738. \end_inset
  12739. </cell>
  12740. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  12741. \begin_inset Text
  12742. \begin_layout Plain Layout
  12743. 20,039
  12744. \end_layout
  12745. \end_inset
  12746. </cell>
  12747. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  12748. \begin_inset Text
  12749. \begin_layout Plain Layout
  12750. 20,955
  12751. \end_layout
  12752. \end_inset
  12753. </cell>
  12754. </row>
  12755. </lyxtabular>
  12756. \end_inset
  12757. \end_layout
  12758. \begin_layout Plain Layout
  12759. \begin_inset Caption Standard
  12760. \begin_layout Plain Layout
  12761. \begin_inset CommandInset label
  12762. LatexCommand label
  12763. name "tab:methyl-est-nonnull"
  12764. \end_inset
  12765. Estimated number of non-null tests, using the method of averaging local
  12766. FDR values
  12767. \begin_inset CommandInset citation
  12768. LatexCommand cite
  12769. key "Phipson2013Thesis"
  12770. literal "false"
  12771. \end_inset
  12772. .
  12773. \end_layout
  12774. \end_inset
  12775. \end_layout
  12776. \end_inset
  12777. \end_layout
  12778. \begin_layout Plain Layout
  12779. \begin_inset Caption Standard
  12780. \begin_layout Plain Layout
  12781. \begin_inset Argument 1
  12782. status collapsed
  12783. \begin_layout Plain Layout
  12784. Estimates of degree of differential methylation in for each contrast in
  12785. each analysis.
  12786. \end_layout
  12787. \end_inset
  12788. \series bold
  12789. Estimates of degree of differential methylation in for each contrast in
  12790. each analysis.
  12791. \series default
  12792. For each of the analyses in Table
  12793. \begin_inset CommandInset ref
  12794. LatexCommand ref
  12795. reference "tab:Summary-of-meth-analysis"
  12796. plural "false"
  12797. caps "false"
  12798. noprefix "false"
  12799. \end_inset
  12800. , these tables show the number of probes called significantly differentially
  12801. methylated at a threshold of 10% FDR for each comparison between TX and
  12802. the other 3 transplant statuses (a) and the estimated total number of probes
  12803. that are differentially methylated (b).
  12804. \end_layout
  12805. \end_inset
  12806. \end_layout
  12807. \end_inset
  12808. \end_layout
  12809. \begin_layout Standard
  12810. \begin_inset Float figure
  12811. wide false
  12812. sideways false
  12813. status collapsed
  12814. \begin_layout Plain Layout
  12815. \align center
  12816. \series bold
  12817. \begin_inset Float figure
  12818. wide false
  12819. sideways false
  12820. status collapsed
  12821. \begin_layout Plain Layout
  12822. \align center
  12823. \begin_inset Graphics
  12824. filename graphics/methylvoom/unadj.dupcor/pval-histograms-PAGE1.pdf
  12825. lyxscale 33
  12826. width 30col%
  12827. groupId meth-pval-hist
  12828. \end_inset
  12829. \end_layout
  12830. \begin_layout Plain Layout
  12831. \series bold
  12832. \begin_inset Caption Standard
  12833. \begin_layout Plain Layout
  12834. AR vs.
  12835. TX, Analysis A
  12836. \end_layout
  12837. \end_inset
  12838. \end_layout
  12839. \end_inset
  12840. \begin_inset space \hfill{}
  12841. \end_inset
  12842. \begin_inset Float figure
  12843. wide false
  12844. sideways false
  12845. status collapsed
  12846. \begin_layout Plain Layout
  12847. \align center
  12848. \begin_inset Graphics
  12849. filename graphics/methylvoom/unadj.dupcor/pval-histograms-PAGE2.pdf
  12850. lyxscale 33
  12851. width 30col%
  12852. groupId meth-pval-hist
  12853. \end_inset
  12854. \end_layout
  12855. \begin_layout Plain Layout
  12856. \series bold
  12857. \begin_inset Caption Standard
  12858. \begin_layout Plain Layout
  12859. ADNR vs.
  12860. TX, Analysis A
  12861. \end_layout
  12862. \end_inset
  12863. \end_layout
  12864. \end_inset
  12865. \begin_inset space \hfill{}
  12866. \end_inset
  12867. \begin_inset Float figure
  12868. wide false
  12869. sideways false
  12870. status collapsed
  12871. \begin_layout Plain Layout
  12872. \align center
  12873. \begin_inset Graphics
  12874. filename graphics/methylvoom/unadj.dupcor/pval-histograms-PAGE3.pdf
  12875. lyxscale 33
  12876. width 30col%
  12877. groupId meth-pval-hist
  12878. \end_inset
  12879. \end_layout
  12880. \begin_layout Plain Layout
  12881. \series bold
  12882. \begin_inset Caption Standard
  12883. \begin_layout Plain Layout
  12884. CAN vs.
  12885. TX, Analysis A
  12886. \end_layout
  12887. \end_inset
  12888. \end_layout
  12889. \end_inset
  12890. \end_layout
  12891. \begin_layout Plain Layout
  12892. \align center
  12893. \series bold
  12894. \begin_inset Float figure
  12895. wide false
  12896. sideways false
  12897. status collapsed
  12898. \begin_layout Plain Layout
  12899. \align center
  12900. \begin_inset Graphics
  12901. filename graphics/methylvoom/unadj.dupcor.sva.aw/pval-histograms-PAGE1.pdf
  12902. lyxscale 33
  12903. width 30col%
  12904. groupId meth-pval-hist
  12905. \end_inset
  12906. \end_layout
  12907. \begin_layout Plain Layout
  12908. \series bold
  12909. \begin_inset Caption Standard
  12910. \begin_layout Plain Layout
  12911. AR vs.
  12912. TX, Analysis B
  12913. \end_layout
  12914. \end_inset
  12915. \end_layout
  12916. \end_inset
  12917. \begin_inset space \hfill{}
  12918. \end_inset
  12919. \begin_inset Float figure
  12920. wide false
  12921. sideways false
  12922. status collapsed
  12923. \begin_layout Plain Layout
  12924. \align center
  12925. \begin_inset Graphics
  12926. filename graphics/methylvoom/unadj.dupcor.sva.aw/pval-histograms-PAGE2.pdf
  12927. lyxscale 33
  12928. width 30col%
  12929. groupId meth-pval-hist
  12930. \end_inset
  12931. \end_layout
  12932. \begin_layout Plain Layout
  12933. \series bold
  12934. \begin_inset Caption Standard
  12935. \begin_layout Plain Layout
  12936. ADNR vs.
  12937. TX, Analysis B
  12938. \end_layout
  12939. \end_inset
  12940. \end_layout
  12941. \end_inset
  12942. \begin_inset space \hfill{}
  12943. \end_inset
  12944. \begin_inset Float figure
  12945. wide false
  12946. sideways false
  12947. status collapsed
  12948. \begin_layout Plain Layout
  12949. \align center
  12950. \begin_inset Graphics
  12951. filename graphics/methylvoom/unadj.dupcor.sva.aw/pval-histograms-PAGE3.pdf
  12952. lyxscale 33
  12953. width 30col%
  12954. groupId meth-pval-hist
  12955. \end_inset
  12956. \end_layout
  12957. \begin_layout Plain Layout
  12958. \series bold
  12959. \begin_inset Caption Standard
  12960. \begin_layout Plain Layout
  12961. CAN vs.
  12962. TX, Analysis B
  12963. \end_layout
  12964. \end_inset
  12965. \end_layout
  12966. \end_inset
  12967. \end_layout
  12968. \begin_layout Plain Layout
  12969. \align center
  12970. \series bold
  12971. \begin_inset Float figure
  12972. wide false
  12973. sideways false
  12974. status collapsed
  12975. \begin_layout Plain Layout
  12976. \align center
  12977. \begin_inset Graphics
  12978. filename graphics/methylvoom/unadj.dupcor.sva.voomaw/pval-histograms-PAGE1.pdf
  12979. lyxscale 33
  12980. width 30col%
  12981. groupId meth-pval-hist
  12982. \end_inset
  12983. \end_layout
  12984. \begin_layout Plain Layout
  12985. \series bold
  12986. \begin_inset Caption Standard
  12987. \begin_layout Plain Layout
  12988. AR vs.
  12989. TX, Analysis C
  12990. \end_layout
  12991. \end_inset
  12992. \end_layout
  12993. \end_inset
  12994. \begin_inset space \hfill{}
  12995. \end_inset
  12996. \begin_inset Float figure
  12997. wide false
  12998. sideways false
  12999. status collapsed
  13000. \begin_layout Plain Layout
  13001. \align center
  13002. \begin_inset Graphics
  13003. filename graphics/methylvoom/unadj.dupcor.sva.voomaw/pval-histograms-PAGE2.pdf
  13004. lyxscale 33
  13005. width 30col%
  13006. groupId meth-pval-hist
  13007. \end_inset
  13008. \end_layout
  13009. \begin_layout Plain Layout
  13010. \series bold
  13011. \begin_inset Caption Standard
  13012. \begin_layout Plain Layout
  13013. ADNR vs.
  13014. TX, Analysis C
  13015. \end_layout
  13016. \end_inset
  13017. \end_layout
  13018. \end_inset
  13019. \begin_inset space \hfill{}
  13020. \end_inset
  13021. \begin_inset Float figure
  13022. wide false
  13023. sideways false
  13024. status collapsed
  13025. \begin_layout Plain Layout
  13026. \align center
  13027. \begin_inset Graphics
  13028. filename graphics/methylvoom/unadj.dupcor.sva.voomaw/pval-histograms-PAGE3.pdf
  13029. lyxscale 33
  13030. width 30col%
  13031. groupId meth-pval-hist
  13032. \end_inset
  13033. \end_layout
  13034. \begin_layout Plain Layout
  13035. \series bold
  13036. \begin_inset Caption Standard
  13037. \begin_layout Plain Layout
  13038. CAN vs.
  13039. TX, Analysis C
  13040. \end_layout
  13041. \end_inset
  13042. \end_layout
  13043. \end_inset
  13044. \end_layout
  13045. \begin_layout Plain Layout
  13046. \begin_inset Caption Standard
  13047. \begin_layout Plain Layout
  13048. \begin_inset Argument 1
  13049. status collapsed
  13050. \begin_layout Plain Layout
  13051. Probe p-value histograms for each contrast in each analysis.
  13052. \end_layout
  13053. \end_inset
  13054. \begin_inset CommandInset label
  13055. LatexCommand label
  13056. name "fig:meth-p-value-histograms"
  13057. \end_inset
  13058. \series bold
  13059. Probe p-value histograms for each contrast in each analysis.
  13060. \series default
  13061. For each differential methylation test of interest, the distribution of
  13062. p-values across all probes is plotted as a histogram.
  13063. The red solid line indicates the density that would be expected under the
  13064. null hypothesis for all probes (a
  13065. \begin_inset Formula $\mathrm{Uniform}(0,1)$
  13066. \end_inset
  13067. distribution), while the blue dotted line indicates the fraction of p-values
  13068. that actually follow the null hypothesis (
  13069. \begin_inset Formula $\hat{\pi}_{0}$
  13070. \end_inset
  13071. ) estimated using the method of averaging local FDR values
  13072. \begin_inset CommandInset citation
  13073. LatexCommand cite
  13074. key "Phipson2013Thesis"
  13075. literal "false"
  13076. \end_inset
  13077. .
  13078. A blue line is only shown in each plot if the estimate of
  13079. \begin_inset Formula $\hat{\pi}_{0}$
  13080. \end_inset
  13081. for that p-value distribution is smaller than 1.
  13082. \end_layout
  13083. \end_inset
  13084. \end_layout
  13085. \end_inset
  13086. \end_layout
  13087. \begin_layout Standard
  13088. \begin_inset Flex TODO Note (inline)
  13089. status open
  13090. \begin_layout Plain Layout
  13091. If time allows, maybe generate the PCA plots before/after SVA effect subtraction
  13092. ?
  13093. \end_layout
  13094. \end_inset
  13095. \end_layout
  13096. \begin_layout Section
  13097. Discussion
  13098. \end_layout
  13099. \begin_layout Subsection
  13100. fRMA achieves clinically applicable normalization without sacrificing classifica
  13101. tion performance
  13102. \end_layout
  13103. \begin_layout Standard
  13104. As shown in Figure
  13105. \begin_inset CommandInset ref
  13106. LatexCommand ref
  13107. reference "fig:Classifier-probabilities-RMA"
  13108. plural "false"
  13109. caps "false"
  13110. noprefix "false"
  13111. \end_inset
  13112. , improper normalization, particularly separate normalization of training
  13113. and test samples, leads to unwanted biases in classification.
  13114. In a controlled experimental context, it is always possible to correct
  13115. this issue by normalizing all experimental samples together.
  13116. However, because it is not feasible to normalize all samples together in
  13117. a clinical context, a single-channel normalization is required.
  13118. \end_layout
  13119. \begin_layout Standard
  13120. The major concern in using a single-channel normalization is that non-single-cha
  13121. nnel methods can share information between arrays to improve the normalization,
  13122. and single-channel methods risk sacrificing the gains in normalization
  13123. accuracy that come from this information sharing.
  13124. In the case of
  13125. \begin_inset Flex Glossary Term
  13126. status open
  13127. \begin_layout Plain Layout
  13128. RMA
  13129. \end_layout
  13130. \end_inset
  13131. , this information sharing is accomplished through quantile normalization
  13132. and median polish steps.
  13133. The need for information sharing in quantile normalization can easily be
  13134. removed by learning a fixed set of quantiles from external data and normalizing
  13135. each array to these fixed quantiles, instead of the quantiles of the data
  13136. itself.
  13137. As long as the fixed quantiles are reasonable, the result will be similar
  13138. to standard
  13139. \begin_inset Flex Glossary Term
  13140. status open
  13141. \begin_layout Plain Layout
  13142. RMA
  13143. \end_layout
  13144. \end_inset
  13145. .
  13146. However, there is no analogous way to eliminate cross-array information
  13147. sharing in the median polish step, so
  13148. \begin_inset Flex Glossary Term
  13149. status open
  13150. \begin_layout Plain Layout
  13151. fRMA
  13152. \end_layout
  13153. \end_inset
  13154. replaces this with a weighted average of probes on each array, with the
  13155. weights learned from external data.
  13156. This step of
  13157. \begin_inset Flex Glossary Term
  13158. status open
  13159. \begin_layout Plain Layout
  13160. fRMA
  13161. \end_layout
  13162. \end_inset
  13163. has the greatest potential to diverge from RMA in undesirable ways.
  13164. \end_layout
  13165. \begin_layout Standard
  13166. However, when run on real data,
  13167. \begin_inset Flex Glossary Term
  13168. status open
  13169. \begin_layout Plain Layout
  13170. fRMA
  13171. \end_layout
  13172. \end_inset
  13173. performed at least as well as
  13174. \begin_inset Flex Glossary Term
  13175. status open
  13176. \begin_layout Plain Layout
  13177. RMA
  13178. \end_layout
  13179. \end_inset
  13180. in both the internal validation and external validation tests.
  13181. This shows that
  13182. \begin_inset Flex Glossary Term
  13183. status open
  13184. \begin_layout Plain Layout
  13185. fRMA
  13186. \end_layout
  13187. \end_inset
  13188. can be used to normalize individual clinical samples in a class prediction
  13189. context without sacrificing the classifier performance that would be obtained
  13190. by using the more well-established
  13191. \begin_inset Flex Glossary Term
  13192. status open
  13193. \begin_layout Plain Layout
  13194. RMA
  13195. \end_layout
  13196. \end_inset
  13197. for normalization.
  13198. The other single-channel normalization method considered,
  13199. \begin_inset Flex Glossary Term
  13200. status open
  13201. \begin_layout Plain Layout
  13202. SCAN
  13203. \end_layout
  13204. \end_inset
  13205. , showed some loss of
  13206. \begin_inset Flex Glossary Term
  13207. status open
  13208. \begin_layout Plain Layout
  13209. AUC
  13210. \end_layout
  13211. \end_inset
  13212. in the external validation test.
  13213. Based on these results,
  13214. \begin_inset Flex Glossary Term
  13215. status open
  13216. \begin_layout Plain Layout
  13217. fRMA
  13218. \end_layout
  13219. \end_inset
  13220. is the preferred normalization for clinical samples in a class prediction
  13221. context.
  13222. \end_layout
  13223. \begin_layout Subsection
  13224. Robust fRMA vectors can be generated for new array platforms
  13225. \end_layout
  13226. \begin_layout Standard
  13227. \begin_inset Flex TODO Note (inline)
  13228. status open
  13229. \begin_layout Plain Layout
  13230. Look up the exact numbers, do a find & replace for
  13231. \begin_inset Quotes eld
  13232. \end_inset
  13233. 850
  13234. \begin_inset Quotes erd
  13235. \end_inset
  13236. \end_layout
  13237. \end_inset
  13238. \end_layout
  13239. \begin_layout Standard
  13240. The published
  13241. \begin_inset Flex Glossary Term
  13242. status open
  13243. \begin_layout Plain Layout
  13244. fRMA
  13245. \end_layout
  13246. \end_inset
  13247. normalization vectors for the hgu133plus2 platform were generated from
  13248. a set of about 850 samples chosen from a wide range of tissues, which the
  13249. authors determined was sufficient to generate a robust set of normalization
  13250. vectors that could be applied across all tissues
  13251. \begin_inset CommandInset citation
  13252. LatexCommand cite
  13253. key "McCall2010"
  13254. literal "false"
  13255. \end_inset
  13256. .
  13257. Since we only had hthgu133pluspm for 2 tissues of interest, our needs were
  13258. more modest.
  13259. Even using only 130 samples in 26 batches of 5 samples each for kidney
  13260. biopsies, we were able to train a robust set of
  13261. \begin_inset Flex Glossary Term
  13262. status open
  13263. \begin_layout Plain Layout
  13264. fRMA
  13265. \end_layout
  13266. \end_inset
  13267. normalization vectors that were not meaningfully affected by the random
  13268. selection of 5 samples from each batch.
  13269. As expected, the training process was just as robust for the blood samples
  13270. with 230 samples in 46 batches of 5 samples each.
  13271. Because these vectors were each generated using training samples from a
  13272. single tissue, they are not suitable for general use, unlike the vectors
  13273. provided with
  13274. \begin_inset Flex Glossary Term
  13275. status open
  13276. \begin_layout Plain Layout
  13277. fRMA
  13278. \end_layout
  13279. \end_inset
  13280. itself.
  13281. They are purpose-built for normalizing a specific type of sample on a specific
  13282. platform.
  13283. This is a mostly acceptable limitation in the context of developing a machine
  13284. learning classifier for diagnosing a disease based on samples of a specific
  13285. tissue.
  13286. \end_layout
  13287. \begin_layout Standard
  13288. \begin_inset Flex TODO Note (inline)
  13289. status open
  13290. \begin_layout Plain Layout
  13291. Talk about how these vectors can be used for any data from these tissues
  13292. on this platform even though they were custom made for this data set.
  13293. \end_layout
  13294. \end_inset
  13295. \end_layout
  13296. \begin_layout Standard
  13297. \begin_inset Flex TODO Note (inline)
  13298. status open
  13299. \begin_layout Plain Layout
  13300. How to bring up that these custom vectors were used in another project by
  13301. someone else that was never published?
  13302. \end_layout
  13303. \end_inset
  13304. \end_layout
  13305. \begin_layout Subsection
  13306. Methylation array data can be successfully analyzed using existing techniques,
  13307. but machine learning poses additional challenges
  13308. \end_layout
  13309. \begin_layout Standard
  13310. Both analysis strategies B and C both yield a reasonable analysis, with
  13311. a mean-variance trend that matches the expected behavior for the non-linear
  13312. \begin_inset Flex Glossary Term
  13313. status open
  13314. \begin_layout Plain Layout
  13315. M-value
  13316. \end_layout
  13317. \end_inset
  13318. transformation (Figure
  13319. \begin_inset CommandInset ref
  13320. LatexCommand ref
  13321. reference "fig:meanvar-sva-aw"
  13322. plural "false"
  13323. caps "false"
  13324. noprefix "false"
  13325. \end_inset
  13326. ) and well-behaved p-value distributions (Figure
  13327. \begin_inset CommandInset ref
  13328. LatexCommand ref
  13329. reference "fig:meth-p-value-histograms"
  13330. plural "false"
  13331. caps "false"
  13332. noprefix "false"
  13333. \end_inset
  13334. ).
  13335. These two analyses also yield similar numbers of significant probes (Table
  13336. \begin_inset CommandInset ref
  13337. LatexCommand ref
  13338. reference "tab:methyl-num-signif"
  13339. plural "false"
  13340. caps "false"
  13341. noprefix "false"
  13342. \end_inset
  13343. ) and similar estimates of the number of differentially methylated probes
  13344. (Table
  13345. \begin_inset CommandInset ref
  13346. LatexCommand ref
  13347. reference "tab:methyl-est-nonnull"
  13348. plural "false"
  13349. caps "false"
  13350. noprefix "false"
  13351. \end_inset
  13352. ).
  13353. The main difference between these two analyses is the method used to account
  13354. for the mean-variance trend.
  13355. In analysis B, the trend is estimated and applied at the probe level: each
  13356. probe's estimated variance is squeezed toward the trend using an empirical
  13357. Bayes procedure (Figure
  13358. \begin_inset CommandInset ref
  13359. LatexCommand ref
  13360. reference "fig:meanvar-sva-aw"
  13361. plural "false"
  13362. caps "false"
  13363. noprefix "false"
  13364. \end_inset
  13365. ).
  13366. In analysis C, the trend is still estimated at the probe level, but instead
  13367. of estimating a single variance value shared across all observations for
  13368. a given probe, the voom method computes an initial estimate of the variance
  13369. for each observation individually based on where its model-fitted
  13370. \begin_inset Flex Glossary Term
  13371. status open
  13372. \begin_layout Plain Layout
  13373. M-value
  13374. \end_layout
  13375. \end_inset
  13376. falls on the trend line and then assigns inverse-variance weights to model
  13377. the difference in variance between observations.
  13378. An overall variance is still estimated for each probe using the same empirical
  13379. Bayes method, but now the residual trend is flat (Figure
  13380. \begin_inset CommandInset ref
  13381. LatexCommand ref
  13382. reference "fig:meanvar-sva-voomaw"
  13383. plural "false"
  13384. caps "false"
  13385. noprefix "false"
  13386. \end_inset
  13387. ), indicating that the mean-variance trend is adequately modeled by scaling
  13388. the estimated variance for each observation using the weights computed
  13389. by voom.
  13390. \end_layout
  13391. \begin_layout Standard
  13392. The difference between the standard empirical Bayes trended variance modeling
  13393. (analysis B) and voom (analysis C) is analogous to the difference between
  13394. a t-test with equal variance and a t-test with unequal variance, except
  13395. that the unequal group variances used in the latter test are estimated
  13396. based on the mean-variance trend from all the probes rather than the data
  13397. for the specific probe being tested, thus stabilizing the group variance
  13398. estimates by sharing information between probes.
  13399. Allowing voom to model the variance using observation weights in this manner
  13400. allows the linear model fit to concentrate statistical power where it will
  13401. do the most good.
  13402. For example, if a particular probe's
  13403. \begin_inset Flex Glossary Term (pl)
  13404. status open
  13405. \begin_layout Plain Layout
  13406. M-value
  13407. \end_layout
  13408. \end_inset
  13409. are always at the extreme of the
  13410. \begin_inset Flex Glossary Term
  13411. status open
  13412. \begin_layout Plain Layout
  13413. M-value
  13414. \end_layout
  13415. \end_inset
  13416. range (e.g.
  13417. less than -4) for
  13418. \begin_inset Flex Glossary Term
  13419. status open
  13420. \begin_layout Plain Layout
  13421. ADNR
  13422. \end_layout
  13423. \end_inset
  13424. samples, but the
  13425. \begin_inset Flex Glossary Term (pl)
  13426. status open
  13427. \begin_layout Plain Layout
  13428. M-value
  13429. \end_layout
  13430. \end_inset
  13431. for that probe in
  13432. \begin_inset Flex Glossary Term
  13433. status open
  13434. \begin_layout Plain Layout
  13435. TX
  13436. \end_layout
  13437. \end_inset
  13438. and
  13439. \begin_inset Flex Glossary Term
  13440. status open
  13441. \begin_layout Plain Layout
  13442. CAN
  13443. \end_layout
  13444. \end_inset
  13445. samples are within the flat region of the mean-variance trend (between
  13446. \begin_inset Formula $-3$
  13447. \end_inset
  13448. and
  13449. \begin_inset Formula $+3$
  13450. \end_inset
  13451. ), voom is able to down-weight the contribution of the high-variance
  13452. \begin_inset Flex Glossary Term (pl)
  13453. status open
  13454. \begin_layout Plain Layout
  13455. M-value
  13456. \end_layout
  13457. \end_inset
  13458. from the
  13459. \begin_inset Flex Glossary Term
  13460. status open
  13461. \begin_layout Plain Layout
  13462. ADNR
  13463. \end_layout
  13464. \end_inset
  13465. samples in order to gain more statistical power while testing for differential
  13466. methylation between
  13467. \begin_inset Flex Glossary Term
  13468. status open
  13469. \begin_layout Plain Layout
  13470. TX
  13471. \end_layout
  13472. \end_inset
  13473. and
  13474. \begin_inset Flex Glossary Term
  13475. status open
  13476. \begin_layout Plain Layout
  13477. CAN
  13478. \end_layout
  13479. \end_inset
  13480. .
  13481. In contrast, modeling the mean-variance trend only at the probe level would
  13482. combine the high-variance
  13483. \begin_inset Flex Glossary Term
  13484. status open
  13485. \begin_layout Plain Layout
  13486. ADNR
  13487. \end_layout
  13488. \end_inset
  13489. samples and lower-variance samples from other conditions and estimate an
  13490. intermediate variance for this probe.
  13491. In practice, analysis B shows that this approach is adequate, but the voom
  13492. approach in analysis C performs at least as well on all model fit criteria
  13493. and yields a larger estimate for the number of differentially methylated
  13494. genes,
  13495. \emph on
  13496. and
  13497. \emph default
  13498. it matches up slightly better with the theoretical properties of the data.
  13499. \end_layout
  13500. \begin_layout Standard
  13501. The significant association of diabetes diagnosis with sample quality is
  13502. interesting.
  13503. The samples with
  13504. \begin_inset Flex Glossary Term
  13505. status open
  13506. \begin_layout Plain Layout
  13507. T2D
  13508. \end_layout
  13509. \end_inset
  13510. tended to have more variation, averaged across all probes, than those with
  13511. \begin_inset Flex Glossary Term
  13512. status open
  13513. \begin_layout Plain Layout
  13514. T1D
  13515. \end_layout
  13516. \end_inset
  13517. .
  13518. This is consistent with the consensus that
  13519. \begin_inset Flex Glossary Term
  13520. status open
  13521. \begin_layout Plain Layout
  13522. T2D
  13523. \end_layout
  13524. \end_inset
  13525. and the associated metabolic syndrome represent a broad dysregulation of
  13526. the body's endocrine signaling related to metabolism
  13527. \begin_inset CommandInset citation
  13528. LatexCommand cite
  13529. key "Volkmar2012,Hall2018,Yokoi2018"
  13530. literal "false"
  13531. \end_inset
  13532. .
  13533. This dysregulation could easily manifest as a greater degree of variation
  13534. in the DNA methylation patterns of affected tissues.
  13535. In contrast,
  13536. \begin_inset Flex Glossary Term
  13537. status open
  13538. \begin_layout Plain Layout
  13539. T1D
  13540. \end_layout
  13541. \end_inset
  13542. has a more specific cause and effect, so a less variable methylation signature
  13543. is expected.
  13544. \end_layout
  13545. \begin_layout Standard
  13546. This preliminary analysis suggests that some degree of differential methylation
  13547. exists between
  13548. \begin_inset Flex Glossary Term
  13549. status open
  13550. \begin_layout Plain Layout
  13551. TX
  13552. \end_layout
  13553. \end_inset
  13554. and each of the three types of transplant disfunction studied.
  13555. Hence, it may be feasible to train a classifier to diagnose transplant
  13556. disfunction from DNA methylation array data.
  13557. However, the major importance of both
  13558. \begin_inset Flex Glossary Term
  13559. status open
  13560. \begin_layout Plain Layout
  13561. SVA
  13562. \end_layout
  13563. \end_inset
  13564. and sample quality weighting for proper modeling of this data poses significant
  13565. challenges for any attempt at a machine learning on data of similar quality.
  13566. While these are easily used in a modeling context with full sample information,
  13567. neither of these methods is directly applicable in a machine learning context,
  13568. where the diagnosis is not known ahead of time.
  13569. If a machine learning approach for methylation-based diagnosis is to be
  13570. pursued, it will either require machine-learning-friendly methods to address
  13571. the same systematic trends in the data that
  13572. \begin_inset Flex Glossary Term
  13573. status open
  13574. \begin_layout Plain Layout
  13575. SVA
  13576. \end_layout
  13577. \end_inset
  13578. and sample quality weighting address, or it will require higher quality
  13579. data with substantially less systematic perturbation of the data.
  13580. \end_layout
  13581. \begin_layout Section
  13582. Future Directions
  13583. \end_layout
  13584. \begin_layout Standard
  13585. \begin_inset Flex TODO Note (inline)
  13586. status open
  13587. \begin_layout Plain Layout
  13588. Some work was already being done with the existing fRMA vectors.
  13589. Do I mention that here?
  13590. \end_layout
  13591. \end_inset
  13592. \end_layout
  13593. \begin_layout Subsection
  13594. Improving fRMA to allow training from batches of unequal size
  13595. \end_layout
  13596. \begin_layout Standard
  13597. Because the tools for building
  13598. \begin_inset Flex Glossary Term
  13599. status open
  13600. \begin_layout Plain Layout
  13601. fRMA
  13602. \end_layout
  13603. \end_inset
  13604. normalization vectors require equal-size batches, many samples must be
  13605. discarded from the training data.
  13606. This is undesirable for a few reasons.
  13607. First, more data is simply better, all other things being equal.
  13608. In this case,
  13609. \begin_inset Quotes eld
  13610. \end_inset
  13611. better
  13612. \begin_inset Quotes erd
  13613. \end_inset
  13614. means a more precise estimate of normalization parameters.
  13615. In addition, the samples to be discarded must be chosen arbitrarily, which
  13616. introduces an unnecessary element of randomness into the estimation process.
  13617. While the randomness can be made deterministic by setting a consistent
  13618. random seed, the need for equal size batches also introduces a need for
  13619. the analyst to decide on the appropriate trade-off between batch size and
  13620. the number of batches.
  13621. This introduces an unnecessary and undesirable
  13622. \begin_inset Quotes eld
  13623. \end_inset
  13624. researcher degree of freedom
  13625. \begin_inset Quotes erd
  13626. \end_inset
  13627. into the analysis, since the generated normalization vectors now depend
  13628. on the choice of batch size based on vague selection criteria and instinct,
  13629. which can unintentionally introduce bias if the researcher chooses a batch
  13630. size based on what seems to yield the most favorable downstream results
  13631. \begin_inset CommandInset citation
  13632. LatexCommand cite
  13633. key "Simmons2011"
  13634. literal "false"
  13635. \end_inset
  13636. .
  13637. \end_layout
  13638. \begin_layout Standard
  13639. Fortunately, the requirement for equal-size batches is not inherent to the
  13640. \begin_inset Flex Glossary Term
  13641. status open
  13642. \begin_layout Plain Layout
  13643. fRMA
  13644. \end_layout
  13645. \end_inset
  13646. algorithm but rather a limitation of the implementation in the
  13647. \begin_inset Flex Code
  13648. status open
  13649. \begin_layout Plain Layout
  13650. frmaTools
  13651. \end_layout
  13652. \end_inset
  13653. package.
  13654. In personal communication, the package's author, Matthew McCall, has indicated
  13655. that with some work, it should be possible to improve the implementation
  13656. to work with batches of unequal sizes.
  13657. The current implementation ignores the batch size when calculating with-batch
  13658. and between-batch residual variances, since the batch size constant cancels
  13659. out later in the calculations as long as all batches are of equal size.
  13660. Hence, the calculations of these parameters would need to be modified to
  13661. remove this optimization and properly calculate the variances using the
  13662. full formula.
  13663. Once this modification is made, a new strategy would need to be developed
  13664. for assessing the stability of parameter estimates, since the random sub-sampli
  13665. ng step is eliminated, meaning that different sub-samplings can no longer
  13666. be compared as in Figures
  13667. \begin_inset CommandInset ref
  13668. LatexCommand ref
  13669. reference "fig:frma-violin"
  13670. plural "false"
  13671. caps "false"
  13672. noprefix "false"
  13673. \end_inset
  13674. and
  13675. \begin_inset CommandInset ref
  13676. LatexCommand ref
  13677. reference "fig:Representative-MA-plots"
  13678. plural "false"
  13679. caps "false"
  13680. noprefix "false"
  13681. \end_inset
  13682. .
  13683. Bootstrap resampling is likely a good candidate here: sample many training
  13684. sets of equal size from the existing training set with replacement, estimate
  13685. parameters from each resampled training set, and compare the estimated
  13686. parameters between bootstraps in order to quantify the variability in each
  13687. parameter's estimation.
  13688. \end_layout
  13689. \begin_layout Subsection
  13690. Developing methylation arrays as a diagnostic tool for kidney transplant
  13691. rejection
  13692. \end_layout
  13693. \begin_layout Standard
  13694. The current study has showed that DNA methylation, as assayed by Illumina
  13695. 450k methylation arrays, has some potential for diagnosing transplant dysfuncti
  13696. ons, including rejection.
  13697. However, very few probes could be confidently identified as differentially
  13698. methylated between healthy and dysfunctional transplants.
  13699. One likely explanation for this is the predominant influence of unobserved
  13700. confounding factors.
  13701. \begin_inset Flex Glossary Term
  13702. status open
  13703. \begin_layout Plain Layout
  13704. SVA
  13705. \end_layout
  13706. \end_inset
  13707. can model and correct for such factors, but the correction can never be
  13708. perfect, so some degree of unwanted systematic variation will always remain
  13709. after
  13710. \begin_inset Flex Glossary Term
  13711. status open
  13712. \begin_layout Plain Layout
  13713. SVA
  13714. \end_layout
  13715. \end_inset
  13716. correction.
  13717. If the effect size of the confounding factors was similar to that of the
  13718. factor of interest (in this case, transplant status), this would be an
  13719. acceptable limitation, since removing most of the confounding factors'
  13720. effects would allow the main effect to stand out.
  13721. However, in this data set, the confounding factors have a much larger effect
  13722. size than transplant status, which means that the small degree of remaining
  13723. variation not removed by
  13724. \begin_inset Flex Glossary Term
  13725. status open
  13726. \begin_layout Plain Layout
  13727. SVA
  13728. \end_layout
  13729. \end_inset
  13730. can still swamp the effect of interest, making it difficult to detect.
  13731. This is, of course, a major issue when the end goal is to develop a classifier
  13732. to diagnose transplant rejection from methylation data, since batch-correction
  13733. methods like
  13734. \begin_inset Flex Glossary Term
  13735. status open
  13736. \begin_layout Plain Layout
  13737. SVA
  13738. \end_layout
  13739. \end_inset
  13740. that work in a linear modeling context cannot be applied in a machine learning
  13741. context.
  13742. \end_layout
  13743. \begin_layout Standard
  13744. Currently, the source of these unwanted systematic variations in the data
  13745. is unknown.
  13746. The best solution would be to determine the cause of the variation and
  13747. eliminate it, thereby eliminating the need to model and remove that variation.
  13748. However, if this proves impractical, another option is to use
  13749. \begin_inset Flex Glossary Term
  13750. status open
  13751. \begin_layout Plain Layout
  13752. SVA
  13753. \end_layout
  13754. \end_inset
  13755. to identify probes that are highly associated with the surrogate variables
  13756. that describe the unwanted variation in the data.
  13757. These probes could be discarded prior to classifier training, in order
  13758. to maximize the chance that the training algorithm will be able to identify
  13759. highly predictive probes from those remaining.
  13760. Lastly, it is possible that some of this unwanted variation is a result
  13761. of the array-based assay being used and would be eliminated by switching
  13762. to assaying DNA methylation using bisulphite sequencing.
  13763. However, this carries the risk that the sequencing assay will have its
  13764. own set of biases that must be corrected for in a different way.
  13765. \end_layout
  13766. \begin_layout Chapter
  13767. \begin_inset CommandInset label
  13768. LatexCommand label
  13769. name "chap:Globin-blocking-cyno"
  13770. \end_inset
  13771. Globin-blocking for more effective blood RNA-seq analysis in primate animal
  13772. model
  13773. \end_layout
  13774. \begin_layout Standard
  13775. \size large
  13776. Ryan C.
  13777. Thompson, Terri Gelbart, Steven R.
  13778. Head, Phillip Ordoukhanian, Courtney Mullen, Dongmei Han, Dora Berman,
  13779. Amelia Bartholomew, Norma Kenyon, Daniel R.
  13780. Salomon
  13781. \end_layout
  13782. \begin_layout Standard
  13783. \begin_inset ERT
  13784. status collapsed
  13785. \begin_layout Plain Layout
  13786. \backslash
  13787. glsresetall
  13788. \end_layout
  13789. \end_inset
  13790. \begin_inset Note Note
  13791. status collapsed
  13792. \begin_layout Plain Layout
  13793. Reintroduce all abbreviations
  13794. \end_layout
  13795. \end_inset
  13796. \end_layout
  13797. \begin_layout Standard
  13798. \begin_inset Flex TODO Note (inline)
  13799. status open
  13800. \begin_layout Plain Layout
  13801. Choose between above and the paper title: Optimizing yield of deep RNA sequencin
  13802. g for gene expression profiling by globin reduction of peripheral blood
  13803. samples from cynomolgus monkeys (
  13804. \emph on
  13805. Macaca fascicularis
  13806. \emph default
  13807. ).
  13808. \end_layout
  13809. \end_inset
  13810. \end_layout
  13811. \begin_layout Section*
  13812. Abstract
  13813. \end_layout
  13814. \begin_layout Paragraph
  13815. Background
  13816. \end_layout
  13817. \begin_layout Standard
  13818. Primate blood contains high concentrations of globin
  13819. \begin_inset Flex Glossary Term
  13820. status open
  13821. \begin_layout Plain Layout
  13822. mRNA
  13823. \end_layout
  13824. \end_inset
  13825. .
  13826. Globin reduction is a standard technique used to improve the expression
  13827. results obtained by DNA microarrays on RNA from blood samples.
  13828. However, with
  13829. \begin_inset Flex Glossary Term
  13830. status open
  13831. \begin_layout Plain Layout
  13832. RNA-seq
  13833. \end_layout
  13834. \end_inset
  13835. quickly replacing microarrays for many applications, the impact of globin
  13836. reduction for
  13837. \begin_inset Flex Glossary Term
  13838. status open
  13839. \begin_layout Plain Layout
  13840. RNA-seq
  13841. \end_layout
  13842. \end_inset
  13843. is less well-studied.
  13844. Moreover, no off-the-shelf kits are available for globin reduction in nonhuman
  13845. primates.
  13846. \end_layout
  13847. \begin_layout Paragraph
  13848. Results
  13849. \end_layout
  13850. \begin_layout Standard
  13851. Here we report a protocol for
  13852. \begin_inset Flex Glossary Term
  13853. status open
  13854. \begin_layout Plain Layout
  13855. RNA-seq
  13856. \end_layout
  13857. \end_inset
  13858. in primate blood samples that uses complimentary
  13859. \begin_inset Flex Glossary Term (pl)
  13860. status open
  13861. \begin_layout Plain Layout
  13862. oligo
  13863. \end_layout
  13864. \end_inset
  13865. to block reverse transcription of the alpha and beta globin genes.
  13866. In test samples from cynomolgus monkeys (
  13867. \emph on
  13868. Macaca fascicularis
  13869. \emph default
  13870. ), this
  13871. \begin_inset Flex Glossary Term
  13872. status open
  13873. \begin_layout Plain Layout
  13874. GB
  13875. \end_layout
  13876. \end_inset
  13877. protocol approximately doubles the yield of informative (non-globin) reads
  13878. by greatly reducing the fraction of globin reads, while also improving
  13879. the consistency in sequencing depth between samples.
  13880. The increased yield enables detection of about 2000 more genes, significantly
  13881. increases the correlation in measured gene expression levels between samples,
  13882. and increases the sensitivity of differential gene expression tests.
  13883. \end_layout
  13884. \begin_layout Paragraph
  13885. Conclusions
  13886. \end_layout
  13887. \begin_layout Standard
  13888. These results show that
  13889. \begin_inset Flex Glossary Term
  13890. status open
  13891. \begin_layout Plain Layout
  13892. GB
  13893. \end_layout
  13894. \end_inset
  13895. significantly improves the cost-effectiveness of
  13896. \begin_inset Flex Glossary Term
  13897. status open
  13898. \begin_layout Plain Layout
  13899. RNA-seq
  13900. \end_layout
  13901. \end_inset
  13902. in primate blood samples by doubling the yield of useful reads, allowing
  13903. detection of more genes, and improving the precision of gene expression
  13904. measurements.
  13905. Based on these results, a globin reducing or blocking protocol is recommended
  13906. for all
  13907. \begin_inset Flex Glossary Term
  13908. status open
  13909. \begin_layout Plain Layout
  13910. RNA-seq
  13911. \end_layout
  13912. \end_inset
  13913. studies of primate blood samples.
  13914. \end_layout
  13915. \begin_layout Standard
  13916. \begin_inset ERT
  13917. status collapsed
  13918. \begin_layout Plain Layout
  13919. \backslash
  13920. glsresetall
  13921. \end_layout
  13922. \end_inset
  13923. \end_layout
  13924. \begin_layout Section
  13925. Introduction
  13926. \end_layout
  13927. \begin_layout Standard
  13928. As part of a multi-lab PO1 grant to study
  13929. \begin_inset Flex Glossary Term
  13930. status open
  13931. \begin_layout Plain Layout
  13932. MSC
  13933. \end_layout
  13934. \end_inset
  13935. infusion as a treatment for graft rejection in cynomolgus monkeys (
  13936. \emph on
  13937. Macaca fascicularis
  13938. \emph default
  13939. ), a large number of serial blood draws from cynomolgus monkeys were planned
  13940. in order to monitor the progress of graft healing and eventual rejection
  13941. after transplantation.
  13942. In order to streamline the process of performing
  13943. \begin_inset Flex Glossary Term
  13944. status open
  13945. \begin_layout Plain Layout
  13946. RNA-seq
  13947. \end_layout
  13948. \end_inset
  13949. on these blood samples, we developed a custom sequencing protocol.
  13950. In the developement of this protocol, we required a solution for the problem
  13951. of excess globin reads.
  13952. High fractions of globin
  13953. \begin_inset Flex Glossary Term
  13954. status open
  13955. \begin_layout Plain Layout
  13956. mRNA
  13957. \end_layout
  13958. \end_inset
  13959. are naturally present in mammalian peripheral blood samples (up to 70%
  13960. of total
  13961. \begin_inset Flex Glossary Term
  13962. status open
  13963. \begin_layout Plain Layout
  13964. mRNA
  13965. \end_layout
  13966. \end_inset
  13967. ) and these are known to interfere with the results of array-based expression
  13968. profiling
  13969. \begin_inset CommandInset citation
  13970. LatexCommand cite
  13971. key "Winn2010"
  13972. literal "false"
  13973. \end_inset
  13974. .
  13975. Globin reduction is also necessary for
  13976. \begin_inset Flex Glossary Term
  13977. status open
  13978. \begin_layout Plain Layout
  13979. RNA-seq
  13980. \end_layout
  13981. \end_inset
  13982. of blood samples, though for unrelated reasons: without globin reduction,
  13983. many
  13984. \begin_inset Flex Glossary Term
  13985. status open
  13986. \begin_layout Plain Layout
  13987. RNA-seq
  13988. \end_layout
  13989. \end_inset
  13990. reads will be derived from the globin genes, leaving fewer for the remainder
  13991. of the genes in the transcriptome.
  13992. However, existing strategies for globin reduction require an additional
  13993. step during sample preparation to deplete the population of globin transcripts
  13994. from the sample prior to reverse transcription
  13995. \begin_inset CommandInset citation
  13996. LatexCommand cite
  13997. key "Mastrokolias2012,Choi2014,Shin2014"
  13998. literal "false"
  13999. \end_inset
  14000. .
  14001. Furthermore, off-the-shelf globin reduction kits are generally targeted
  14002. at human or mouse globin, not cynomolgus monkey, and sequence identity
  14003. between human and cyno globin genes cannot be automatically assumed.
  14004. Hence, we sought to incorporate a custom globin reduction method into our
  14005. \begin_inset Flex Glossary Term
  14006. status open
  14007. \begin_layout Plain Layout
  14008. RNA-seq
  14009. \end_layout
  14010. \end_inset
  14011. protocol purely by adding additional reagents to an existing step in the
  14012. sample preparation.
  14013. \end_layout
  14014. \begin_layout Section
  14015. Approach
  14016. \end_layout
  14017. \begin_layout Standard
  14018. \begin_inset Note Note
  14019. status collapsed
  14020. \begin_layout Plain Layout
  14021. Consider putting some of this in the Intro chapter
  14022. \end_layout
  14023. \begin_layout Itemize
  14024. Cynomolgus monkeys as a model organism
  14025. \end_layout
  14026. \begin_deeper
  14027. \begin_layout Itemize
  14028. Highly related to humans
  14029. \end_layout
  14030. \begin_layout Itemize
  14031. Small size and short life cycle - good research animal
  14032. \end_layout
  14033. \begin_layout Itemize
  14034. Genomics resources still in development
  14035. \end_layout
  14036. \end_deeper
  14037. \begin_layout Itemize
  14038. Inadequacy of existing blood RNA-seq protocols
  14039. \end_layout
  14040. \begin_deeper
  14041. \begin_layout Itemize
  14042. Existing protocols use a separate globin pulldown step, slowing down processing
  14043. \end_layout
  14044. \end_deeper
  14045. \end_inset
  14046. \end_layout
  14047. \begin_layout Standard
  14048. We evaluated globin reduction for
  14049. \begin_inset Flex Glossary Term
  14050. status open
  14051. \begin_layout Plain Layout
  14052. RNA-seq
  14053. \end_layout
  14054. \end_inset
  14055. by blocking reverse transcription of globin transcripts using custom blocking
  14056. \begin_inset Flex Glossary Term (pl)
  14057. status open
  14058. \begin_layout Plain Layout
  14059. oligo
  14060. \end_layout
  14061. \end_inset
  14062. .
  14063. We demonstrate that
  14064. \begin_inset Flex Glossary Term
  14065. status open
  14066. \begin_layout Plain Layout
  14067. GB
  14068. \end_layout
  14069. \end_inset
  14070. significantly improves the cost-effectiveness of
  14071. \begin_inset Flex Glossary Term
  14072. status open
  14073. \begin_layout Plain Layout
  14074. RNA-seq
  14075. \end_layout
  14076. \end_inset
  14077. in blood samples.
  14078. Thus, our protocol offers a significant advantage to any investigator planning
  14079. to use
  14080. \begin_inset Flex Glossary Term
  14081. status open
  14082. \begin_layout Plain Layout
  14083. RNA-seq
  14084. \end_layout
  14085. \end_inset
  14086. for gene expression profiling of nonhuman primate blood samples.
  14087. Our method can be generally applied to any species by designing complementary
  14088. \begin_inset Flex Glossary Term
  14089. status open
  14090. \begin_layout Plain Layout
  14091. oligo
  14092. \end_layout
  14093. \end_inset
  14094. blocking probes to the globin gene sequences of that species.
  14095. Indeed, any highly expressed but biologically uninformative transcripts
  14096. can also be blocked to further increase sequencing efficiency and value
  14097. \begin_inset CommandInset citation
  14098. LatexCommand cite
  14099. key "Arnaud2016"
  14100. literal "false"
  14101. \end_inset
  14102. .
  14103. \end_layout
  14104. \begin_layout Section
  14105. Methods
  14106. \end_layout
  14107. \begin_layout Subsection
  14108. Sample collection
  14109. \end_layout
  14110. \begin_layout Standard
  14111. All research reported here was done under IACUC-approved protocols at the
  14112. University of Miami and complied with all applicable federal and state
  14113. regulations and ethical principles for nonhuman primate research.
  14114. Blood draws occurred between 16
  14115. \begin_inset space ~
  14116. \end_inset
  14117. April
  14118. \begin_inset space ~
  14119. \end_inset
  14120. 2012 and 18
  14121. \begin_inset space ~
  14122. \end_inset
  14123. June
  14124. \begin_inset space ~
  14125. \end_inset
  14126. 2015.
  14127. The experimental system involved intrahepatic pancreatic islet transplantation
  14128. into Cynomolgus monkeys with induced diabetes mellitus with or without
  14129. concomitant infusion of mesenchymal stem cells.
  14130. Blood was collected at serial time points before and after transplantation
  14131. into PAXgene Blood RNA tubes (PreAnalytiX/Qiagen, Valencia, CA) at the
  14132. precise volume:volume ratio of 2.5
  14133. \begin_inset space ~
  14134. \end_inset
  14135. ml whole blood into 6.9
  14136. \begin_inset space ~
  14137. \end_inset
  14138. ml of PAX gene additive.
  14139. \end_layout
  14140. \begin_layout Subsection
  14141. Globin blocking oligonucleotide design
  14142. \end_layout
  14143. \begin_layout Standard
  14144. \begin_inset Flex TODO Note (inline)
  14145. status open
  14146. \begin_layout Plain Layout
  14147. HBA1 and HBA2 is wrong for cyno?
  14148. \end_layout
  14149. \end_inset
  14150. \end_layout
  14151. \begin_layout Standard
  14152. Four
  14153. \begin_inset Flex Glossary Term (pl)
  14154. status open
  14155. \begin_layout Plain Layout
  14156. oligo
  14157. \end_layout
  14158. \end_inset
  14159. were designed to hybridize to the
  14160. \begin_inset Formula $3^{\prime}$
  14161. \end_inset
  14162. end of the transcripts for the Cynomolgus HBA1, HBA2 and HBB genes, with
  14163. two hybridization sites for HBB and 2 sites for HBA (the chosen sites were
  14164. identical in both HBA genes).
  14165. All
  14166. \begin_inset Flex Glossary Term (pl)
  14167. status open
  14168. \begin_layout Plain Layout
  14169. oligo
  14170. \end_layout
  14171. \end_inset
  14172. were purchased from Sigma and were entirely composed of 2
  14173. \begin_inset Formula $^{\prime}$
  14174. \end_inset
  14175. O-Me bases with a C3 spacer positioned at the
  14176. \begin_inset Formula $3^{\prime}$
  14177. \end_inset
  14178. ends to prevent any polymerase mediated primer extension.
  14179. \end_layout
  14180. \begin_layout Description
  14181. HBA1/2
  14182. \begin_inset space ~
  14183. \end_inset
  14184. site
  14185. \begin_inset space ~
  14186. \end_inset
  14187. 1:
  14188. \family typewriter
  14189. GCCCACUCAGACUUUAUUCAAAG-C3spacer
  14190. \end_layout
  14191. \begin_layout Description
  14192. HBA1/2
  14193. \begin_inset space ~
  14194. \end_inset
  14195. site
  14196. \begin_inset space ~
  14197. \end_inset
  14198. 2:
  14199. \family typewriter
  14200. GGUGCAAGGAGGGGAGGAG-C3spacer
  14201. \end_layout
  14202. \begin_layout Description
  14203. HBB
  14204. \begin_inset space ~
  14205. \end_inset
  14206. site
  14207. \begin_inset space ~
  14208. \end_inset
  14209. 1:
  14210. \family typewriter
  14211. AAUGAAAAUAAAUGUUUUUUAUUAG-C3spacer
  14212. \end_layout
  14213. \begin_layout Description
  14214. HBB
  14215. \begin_inset space ~
  14216. \end_inset
  14217. site
  14218. \begin_inset space ~
  14219. \end_inset
  14220. 2:
  14221. \family typewriter
  14222. CUCAAGGCCCUUCAUAAUAUCCC-C3spacer
  14223. \end_layout
  14224. \begin_layout Subsection
  14225. RNA-seq library preparation
  14226. \end_layout
  14227. \begin_layout Standard
  14228. Sequencing libraries were prepared with 200
  14229. \begin_inset space ~
  14230. \end_inset
  14231. ng total RNA from each sample.
  14232. Polyadenylated
  14233. \begin_inset Flex Glossary Term
  14234. status open
  14235. \begin_layout Plain Layout
  14236. mRNA
  14237. \end_layout
  14238. \end_inset
  14239. was selected from 200
  14240. \begin_inset space ~
  14241. \end_inset
  14242. ng aliquots of cynomolgus blood-derived total RNA using Ambion Dynabeads
  14243. Oligo(dT)25 beads (Invitrogen) following the manufacturer’s recommended
  14244. protocol.
  14245. PolyA selected RNA was then combined with 8
  14246. \begin_inset space ~
  14247. \end_inset
  14248. pmol of HBA1/2
  14249. \begin_inset space ~
  14250. \end_inset
  14251. (site
  14252. \begin_inset space ~
  14253. \end_inset
  14254. 1), 8
  14255. \begin_inset space ~
  14256. \end_inset
  14257. pmol of HBA1/2
  14258. \begin_inset space ~
  14259. \end_inset
  14260. (site
  14261. \begin_inset space ~
  14262. \end_inset
  14263. 2), 12
  14264. \begin_inset space ~
  14265. \end_inset
  14266. pmol of HBB
  14267. \begin_inset space ~
  14268. \end_inset
  14269. (site
  14270. \begin_inset space ~
  14271. \end_inset
  14272. 1) and 12
  14273. \begin_inset space ~
  14274. \end_inset
  14275. pmol of HBB
  14276. \begin_inset space ~
  14277. \end_inset
  14278. (site
  14279. \begin_inset space ~
  14280. \end_inset
  14281. 2)
  14282. \begin_inset Flex Glossary Term (pl)
  14283. status open
  14284. \begin_layout Plain Layout
  14285. oligo
  14286. \end_layout
  14287. \end_inset
  14288. .
  14289. In addition, 20
  14290. \begin_inset space ~
  14291. \end_inset
  14292. pmol of RT primer containing a portion of the Illumina adapter sequence
  14293. (B-oligo-dTV: GAGTTCCTTGGCACCCGAGAATTCCATTTTTTTTTTTTTTTTTTTV) and 4
  14294. \begin_inset space ~
  14295. \end_inset
  14296. \emph on
  14297. μ
  14298. \emph default
  14299. L of 5X First Strand buffer (250
  14300. \begin_inset space ~
  14301. \end_inset
  14302. mM Tris-HCl pH
  14303. \begin_inset space ~
  14304. \end_inset
  14305. 8.3, 375
  14306. \begin_inset space ~
  14307. \end_inset
  14308. mM KCl, 15
  14309. \begin_inset space ~
  14310. \end_inset
  14311. mM
  14312. \begin_inset Formula $\textrm{MgCl}_{2}$
  14313. \end_inset
  14314. ) were added in a total volume of 15
  14315. \begin_inset space ~
  14316. \end_inset
  14317. µL.
  14318. The RNA was fragmented by heating this cocktail for 3 minutes at 95°C and
  14319. then placed on ice.
  14320. This was followed by the addition of 2
  14321. \begin_inset space ~
  14322. \end_inset
  14323. µL 0.1
  14324. \begin_inset space ~
  14325. \end_inset
  14326. M DTT, 1
  14327. \begin_inset space ~
  14328. \end_inset
  14329. µL RNaseOUT, 1
  14330. \begin_inset space ~
  14331. \end_inset
  14332. µL 10
  14333. \begin_inset space ~
  14334. \end_inset
  14335. mM dNTPs 10% biotin-16 aminoallyl-
  14336. \begin_inset Formula $2^{\prime}$
  14337. \end_inset
  14338. - dUTP and 10% biotin-16 aminoallyl-
  14339. \begin_inset Formula $2^{\prime}$
  14340. \end_inset
  14341. -dCTP (TriLink Biotech, San Diego, CA), 1
  14342. \begin_inset space ~
  14343. \end_inset
  14344. µL Superscript II (200
  14345. \begin_inset space ~
  14346. \end_inset
  14347. U/µL, Thermo-Fisher).
  14348. A second “unblocked” library was prepared in the same way for each sample
  14349. but replacing the blocking
  14350. \begin_inset Flex Glossary Term (pl)
  14351. status open
  14352. \begin_layout Plain Layout
  14353. oligo
  14354. \end_layout
  14355. \end_inset
  14356. with an equivalent volume of water.
  14357. The reaction was carried out at 25°C for 15 minutes and 42°C for 40 minutes,
  14358. followed by incubation at 75°C for 10 minutes to inactivate the reverse
  14359. transcriptase.
  14360. \end_layout
  14361. \begin_layout Standard
  14362. The cDNA/RNA hybrid molecules were purified using 1.8X Ampure XP beads (Agencourt
  14363. ) following supplier’s recommended protocol.
  14364. The cDNA/RNA hybrid was eluted in 25
  14365. \begin_inset space ~
  14366. \end_inset
  14367. µL of 10
  14368. \begin_inset space ~
  14369. \end_inset
  14370. mM Tris-HCl pH
  14371. \begin_inset space ~
  14372. \end_inset
  14373. 8.0, and then bound to 25
  14374. \begin_inset space ~
  14375. \end_inset
  14376. µL of M280 Magnetic Streptavidin beads washed per recommended protocol (Thermo-F
  14377. isher).
  14378. After 30 minutes of binding, beads were washed one time in 100
  14379. \begin_inset space ~
  14380. \end_inset
  14381. µL 0.1
  14382. \begin_inset space ~
  14383. \end_inset
  14384. N NaOH to denature and remove the bound RNA, followed by two 100
  14385. \begin_inset space ~
  14386. \end_inset
  14387. µL washes with 1X TE buffer.
  14388. \end_layout
  14389. \begin_layout Standard
  14390. Subsequent attachment of the
  14391. \begin_inset Formula $5^{\prime}$
  14392. \end_inset
  14393. Illumina A adapter was performed by on-bead random primer extension of
  14394. the following sequence (A-N8 primer:
  14395. \family typewriter
  14396. TTCAGAGTTCTACAGTCCGACGATCNNNNNNNN
  14397. \family default
  14398. ).
  14399. Briefly, beads were resuspended in a 20
  14400. \begin_inset space ~
  14401. \end_inset
  14402. µL reaction containing 5
  14403. \begin_inset space ~
  14404. \end_inset
  14405. µM A-N8 primer, 40
  14406. \begin_inset space ~
  14407. \end_inset
  14408. mM Tris-HCl pH
  14409. \begin_inset space ~
  14410. \end_inset
  14411. 7.5, 20
  14412. \begin_inset space ~
  14413. \end_inset
  14414. mM
  14415. \begin_inset Formula $\textrm{MgCl}_{2}$
  14416. \end_inset
  14417. , 50
  14418. \begin_inset space ~
  14419. \end_inset
  14420. mM NaCl, 0.325
  14421. \begin_inset space ~
  14422. \end_inset
  14423. U/µL Sequenase
  14424. \begin_inset space ~
  14425. \end_inset
  14426. 2.0 (Affymetrix, Santa Clara, CA), 0.0025
  14427. \begin_inset space ~
  14428. \end_inset
  14429. U/µL inorganic pyrophosphatase (Affymetrix) and 300
  14430. \begin_inset space ~
  14431. \end_inset
  14432. µM each dNTP.
  14433. Reaction was incubated at 22°C for 30 minutes, then beads were washed 2
  14434. times with 1X TE buffer (200
  14435. \begin_inset space ~
  14436. \end_inset
  14437. µL).
  14438. \end_layout
  14439. \begin_layout Standard
  14440. The magnetic streptavidin beads were resuspended in 34
  14441. \begin_inset space ~
  14442. \end_inset
  14443. µL nuclease-free water and added directly to a
  14444. \begin_inset Flex Glossary Term
  14445. status open
  14446. \begin_layout Plain Layout
  14447. PCR
  14448. \end_layout
  14449. \end_inset
  14450. tube.
  14451. The two Illumina protocol-specified
  14452. \begin_inset Flex Glossary Term
  14453. status open
  14454. \begin_layout Plain Layout
  14455. PCR
  14456. \end_layout
  14457. \end_inset
  14458. primers were added at 0.53
  14459. \begin_inset space ~
  14460. \end_inset
  14461. µM (Illumina TruSeq Universal Primer 1 and Illumina TruSeq barcoded
  14462. \begin_inset Flex Glossary Term
  14463. status open
  14464. \begin_layout Plain Layout
  14465. PCR
  14466. \end_layout
  14467. \end_inset
  14468. primer 2), along with 40
  14469. \begin_inset space ~
  14470. \end_inset
  14471. µL 2X KAPA HiFi Hotstart ReadyMix (KAPA, Willmington MA) and thermocycled
  14472. as follows: starting with 98°C (2 min-hold); 15 cycles of 98°C, 20sec;
  14473. 60°C, 30sec; 72°C, 30sec; and finished with a 72°C (2 min-hold).
  14474. \end_layout
  14475. \begin_layout Standard
  14476. \begin_inset Flex Glossary Term
  14477. status open
  14478. \begin_layout Plain Layout
  14479. PCR
  14480. \end_layout
  14481. \end_inset
  14482. products were purified with 1X Ampure Beads following manufacturer’s recommende
  14483. d protocol.
  14484. Libraries were then analyzed using the Agilent TapeStation and quantitation
  14485. of desired size range was performed by “smear analysis”.
  14486. Samples were pooled in equimolar batches of 16 samples.
  14487. Pooled libraries were size selected on 2% agarose gels (E-Gel EX Agarose
  14488. Gels; Thermo-Fisher).
  14489. Products were cut between 250 and 350
  14490. \begin_inset space ~
  14491. \end_inset
  14492. bp (corresponding to insert sizes of 130 to 230
  14493. \begin_inset space ~
  14494. \end_inset
  14495. bp).
  14496. Finished library pools were then sequenced on the Illumina NextSeq500 instrumen
  14497. t with 75
  14498. \begin_inset space ~
  14499. \end_inset
  14500. bp read lengths.
  14501. \end_layout
  14502. \begin_layout Subsection
  14503. Read alignment and counting
  14504. \end_layout
  14505. \begin_layout Standard
  14506. \begin_inset ERT
  14507. status collapsed
  14508. \begin_layout Plain Layout
  14509. \backslash
  14510. emergencystretch 3em
  14511. \end_layout
  14512. \end_inset
  14513. \begin_inset Note Note
  14514. status collapsed
  14515. \begin_layout Plain Layout
  14516. Need to relax the justification parameters just for this paragraph, or else
  14517. featureCounts can break out of the margin.
  14518. \end_layout
  14519. \end_inset
  14520. \end_layout
  14521. \begin_layout Standard
  14522. Reads were aligned to the cynomolgus genome using STAR
  14523. \begin_inset CommandInset citation
  14524. LatexCommand cite
  14525. key "Wilson2013,Dobin2012"
  14526. literal "false"
  14527. \end_inset
  14528. .
  14529. Counts of uniquely mapped reads were obtained for every gene in each sample
  14530. with the
  14531. \begin_inset Flex Code
  14532. status open
  14533. \begin_layout Plain Layout
  14534. featureCounts
  14535. \end_layout
  14536. \end_inset
  14537. function from the
  14538. \begin_inset Flex Code
  14539. status open
  14540. \begin_layout Plain Layout
  14541. Rsubread
  14542. \end_layout
  14543. \end_inset
  14544. package, using each of the three possibilities for the
  14545. \begin_inset Flex Code
  14546. status open
  14547. \begin_layout Plain Layout
  14548. strandSpecific
  14549. \end_layout
  14550. \end_inset
  14551. option: sense, antisense, and unstranded
  14552. \begin_inset CommandInset citation
  14553. LatexCommand cite
  14554. key "Liao2014"
  14555. literal "false"
  14556. \end_inset
  14557. .
  14558. A few artifacts in the cynomolgus genome annotation complicated read counting.
  14559. First, no ortholog is annotated for alpha globin in the cynomolgus genome,
  14560. presumably because the human genome has two alpha globin genes with nearly
  14561. identical sequences, making the orthology relationship ambiguous.
  14562. However, two loci in the cynomolgus genome are annotated as “hemoglobin
  14563. subunit alpha-like” (LOC102136192 and LOC102136846).
  14564. LOC102136192 is annotated as a pseudogene while LOC102136846 is annotated
  14565. as protein-coding.
  14566. Our globin reduction protocol was designed to include blocking of these
  14567. two genes.
  14568. Indeed, these two genes together have almost the same read counts in each
  14569. library as the properly-annotated HBB gene and much larger counts than
  14570. any other gene in the unblocked libraries, giving confidence that reads
  14571. derived from the real alpha globin are mapping to both genes.
  14572. Thus, reads from both of these loci were counted as alpha globin reads
  14573. in all further analyses.
  14574. The second artifact is a small, uncharacterized non-coding RNA gene (LOC1021365
  14575. 91), which overlaps the HBA-like gene (LOC102136192) on the opposite strand.
  14576. If counting is not performed in stranded mode (or if a non-strand-specific
  14577. sequencing protocol is used), many reads mapping to the globin gene will
  14578. be discarded as ambiguous due to their overlap with this
  14579. \begin_inset Flex Glossary Term
  14580. status open
  14581. \begin_layout Plain Layout
  14582. ncRNA
  14583. \end_layout
  14584. \end_inset
  14585. gene, resulting in significant undercounting of globin reads.
  14586. Therefore, stranded sense counts were used for all further analysis in
  14587. the present study to insure that we accurately accounted for globin transcript
  14588. reduction.
  14589. However, we note that stranded reads are not necessary for
  14590. \begin_inset Flex Glossary Term
  14591. status open
  14592. \begin_layout Plain Layout
  14593. RNA-seq
  14594. \end_layout
  14595. \end_inset
  14596. using our protocol in standard practice.
  14597. \end_layout
  14598. \begin_layout Standard
  14599. \begin_inset ERT
  14600. status collapsed
  14601. \begin_layout Plain Layout
  14602. \backslash
  14603. emergencystretch 0em
  14604. \end_layout
  14605. \end_inset
  14606. \end_layout
  14607. \begin_layout Subsection
  14608. Normalization and exploratory data analysis
  14609. \end_layout
  14610. \begin_layout Standard
  14611. Libraries were normalized by computing scaling factors using the
  14612. \begin_inset Flex Code
  14613. status open
  14614. \begin_layout Plain Layout
  14615. edgeR
  14616. \end_layout
  14617. \end_inset
  14618. package's
  14619. \begin_inset Flex Glossary Term
  14620. status open
  14621. \begin_layout Plain Layout
  14622. TMM
  14623. \end_layout
  14624. \end_inset
  14625. method
  14626. \begin_inset CommandInset citation
  14627. LatexCommand cite
  14628. key "Robinson2010"
  14629. literal "false"
  14630. \end_inset
  14631. .
  14632. \begin_inset Flex Glossary Term (Capital)
  14633. status open
  14634. \begin_layout Plain Layout
  14635. logCPM
  14636. \end_layout
  14637. \end_inset
  14638. values were calculated using the
  14639. \begin_inset Flex Code
  14640. status open
  14641. \begin_layout Plain Layout
  14642. cpm
  14643. \end_layout
  14644. \end_inset
  14645. function in
  14646. \begin_inset Flex Code
  14647. status open
  14648. \begin_layout Plain Layout
  14649. edgeR
  14650. \end_layout
  14651. \end_inset
  14652. for individual samples and
  14653. \begin_inset Flex Code
  14654. status open
  14655. \begin_layout Plain Layout
  14656. aveLogCPM
  14657. \end_layout
  14658. \end_inset
  14659. function for averages across groups of samples, using those functions’
  14660. default prior count values to avoid taking the logarithm of 0.
  14661. Genes were considered “present” if their average normalized
  14662. \begin_inset Flex Glossary Term
  14663. status open
  14664. \begin_layout Plain Layout
  14665. logCPM
  14666. \end_layout
  14667. \end_inset
  14668. values across all libraries were at least
  14669. \begin_inset Formula $-1$
  14670. \end_inset
  14671. .
  14672. Normalizing for gene length was unnecessary because the sequencing protocol
  14673. is
  14674. \begin_inset Formula $3^{\prime}$
  14675. \end_inset
  14676. -biased and hence the expected read count for each gene is related to the
  14677. transcript’s copy number but not its length.
  14678. \end_layout
  14679. \begin_layout Standard
  14680. In order to assess the effect of
  14681. \begin_inset Flex Glossary Term
  14682. status open
  14683. \begin_layout Plain Layout
  14684. GB
  14685. \end_layout
  14686. \end_inset
  14687. on reproducibility, Pearson and Spearman correlation coefficients were
  14688. computed between the
  14689. \begin_inset Flex Glossary Term
  14690. status open
  14691. \begin_layout Plain Layout
  14692. logCPM
  14693. \end_layout
  14694. \end_inset
  14695. values for every pair of libraries within the
  14696. \begin_inset Flex Glossary Term
  14697. status open
  14698. \begin_layout Plain Layout
  14699. GB
  14700. \end_layout
  14701. \end_inset
  14702. non-GB groups, and
  14703. \begin_inset Flex Code
  14704. status open
  14705. \begin_layout Plain Layout
  14706. edgeR
  14707. \end_layout
  14708. \end_inset
  14709. 's
  14710. \begin_inset Flex Code
  14711. status open
  14712. \begin_layout Plain Layout
  14713. estimateDisp
  14714. \end_layout
  14715. \end_inset
  14716. function was used to compute
  14717. \begin_inset Flex Glossary Term
  14718. status open
  14719. \begin_layout Plain Layout
  14720. NB
  14721. \end_layout
  14722. \end_inset
  14723. dispersions separately for the two groups
  14724. \begin_inset CommandInset citation
  14725. LatexCommand cite
  14726. key "Chen2014"
  14727. literal "false"
  14728. \end_inset
  14729. .
  14730. \end_layout
  14731. \begin_layout Subsection
  14732. Differential expression analysis
  14733. \end_layout
  14734. \begin_layout Standard
  14735. All tests for differential gene expression were performed using
  14736. \begin_inset Flex Code
  14737. status open
  14738. \begin_layout Plain Layout
  14739. edgeR
  14740. \end_layout
  14741. \end_inset
  14742. , by first fitting a
  14743. \begin_inset Flex Glossary Term
  14744. status open
  14745. \begin_layout Plain Layout
  14746. NB
  14747. \end_layout
  14748. \end_inset
  14749. \begin_inset Flex Glossary Term
  14750. status open
  14751. \begin_layout Plain Layout
  14752. GLM
  14753. \end_layout
  14754. \end_inset
  14755. to the counts and normalization factors and then performing a quasi-likelihood
  14756. F-test with robust estimation of outlier gene dispersions
  14757. \begin_inset CommandInset citation
  14758. LatexCommand cite
  14759. key "Lund2012,Phipson2016"
  14760. literal "false"
  14761. \end_inset
  14762. .
  14763. To investigate the effects of
  14764. \begin_inset Flex Glossary Term
  14765. status open
  14766. \begin_layout Plain Layout
  14767. GB
  14768. \end_layout
  14769. \end_inset
  14770. on each gene, an additive model was fit to the full data with coefficients
  14771. for
  14772. \begin_inset Flex Glossary Term
  14773. status open
  14774. \begin_layout Plain Layout
  14775. GB
  14776. \end_layout
  14777. \end_inset
  14778. and Sample
  14779. \begin_inset Flex Glossary Term
  14780. status open
  14781. \begin_layout Plain Layout
  14782. ID
  14783. \end_layout
  14784. \end_inset
  14785. .
  14786. To test the effect of
  14787. \begin_inset Flex Glossary Term
  14788. status open
  14789. \begin_layout Plain Layout
  14790. GB
  14791. \end_layout
  14792. \end_inset
  14793. on detection of differentially expressed genes, the
  14794. \begin_inset Flex Glossary Term
  14795. status open
  14796. \begin_layout Plain Layout
  14797. GB
  14798. \end_layout
  14799. \end_inset
  14800. samples and non-GB samples were each analyzed independently as follows:
  14801. for each animal with both a pre-transplant and a post-transplant time point
  14802. in the data set, the pre-transplant sample and the earliest post-transplant
  14803. sample were selected, and all others were excluded, yielding a pre-/post-transp
  14804. lant pair of samples for each animal (
  14805. \begin_inset Formula $N=7$
  14806. \end_inset
  14807. animals with paired samples).
  14808. These samples were analyzed for pre-transplant vs.
  14809. post-transplant differential gene expression while controlling for inter-animal
  14810. variation using an additive model with coefficients for transplant and
  14811. animal
  14812. \begin_inset Flex Glossary Term
  14813. status open
  14814. \begin_layout Plain Layout
  14815. ID
  14816. \end_layout
  14817. \end_inset
  14818. .
  14819. In all analyses, p-values were adjusted using the
  14820. \begin_inset Flex Glossary Term
  14821. status open
  14822. \begin_layout Plain Layout
  14823. BH
  14824. \end_layout
  14825. \end_inset
  14826. procedure for
  14827. \begin_inset Flex Glossary Term
  14828. status open
  14829. \begin_layout Plain Layout
  14830. FDR
  14831. \end_layout
  14832. \end_inset
  14833. control
  14834. \begin_inset CommandInset citation
  14835. LatexCommand cite
  14836. key "Benjamini1995"
  14837. literal "false"
  14838. \end_inset
  14839. .
  14840. \end_layout
  14841. \begin_layout Standard
  14842. \begin_inset Note Note
  14843. status open
  14844. \begin_layout Itemize
  14845. New blood RNA-seq protocol to block reverse transcription of globin genes
  14846. \end_layout
  14847. \begin_layout Itemize
  14848. Blood RNA-seq time course after transplants with/without MSC infusion
  14849. \end_layout
  14850. \end_inset
  14851. \end_layout
  14852. \begin_layout Section
  14853. Results
  14854. \end_layout
  14855. \begin_layout Subsection
  14856. Globin blocking yields a larger and more consistent fraction of useful reads
  14857. \end_layout
  14858. \begin_layout Standard
  14859. The objective of the present study was to validate a new protocol for deep
  14860. \begin_inset Flex Glossary Term
  14861. status open
  14862. \begin_layout Plain Layout
  14863. RNA-seq
  14864. \end_layout
  14865. \end_inset
  14866. of whole blood drawn into PaxGene tubes from cynomolgus monkeys undergoing
  14867. islet transplantation, with particular focus on minimizing the loss of
  14868. useful sequencing space to uninformative globin reads.
  14869. The details of the analysis with respect to transplant outcomes and the
  14870. impact of mesenchymal stem cell treatment will be reported in a separate
  14871. manuscript (in preparation).
  14872. To focus on the efficacy of our
  14873. \begin_inset Flex Glossary Term
  14874. status open
  14875. \begin_layout Plain Layout
  14876. GB
  14877. \end_layout
  14878. \end_inset
  14879. protocol, 37 blood samples, 16 from pre-transplant and 21 from post-transplant
  14880. time points, were each prepped once with and once without
  14881. \begin_inset Flex Glossary Term
  14882. status open
  14883. \begin_layout Plain Layout
  14884. GB
  14885. \end_layout
  14886. \end_inset
  14887. \begin_inset Flex Glossary Term (pl)
  14888. status open
  14889. \begin_layout Plain Layout
  14890. oligo
  14891. \end_layout
  14892. \end_inset
  14893. , and were then sequenced on an Illumina NextSeq500 instrument.
  14894. The number of reads aligning to each gene in the cynomolgus genome was
  14895. counted.
  14896. Table
  14897. \begin_inset CommandInset ref
  14898. LatexCommand ref
  14899. reference "tab:Fractions-of-reads"
  14900. plural "false"
  14901. caps "false"
  14902. noprefix "false"
  14903. \end_inset
  14904. summarizes the distribution of read fractions among the
  14905. \begin_inset Flex Glossary Term
  14906. status open
  14907. \begin_layout Plain Layout
  14908. GB
  14909. \end_layout
  14910. \end_inset
  14911. and non-GB libraries.
  14912. In the libraries with no
  14913. \begin_inset Flex Glossary Term
  14914. status open
  14915. \begin_layout Plain Layout
  14916. GB
  14917. \end_layout
  14918. \end_inset
  14919. , globin reads made up an average of 44.6% of total input reads, while reads
  14920. assigned to all other genes made up an average of 26.3%.
  14921. The remaining reads either aligned to intergenic regions (that include
  14922. long non-coding RNAs) or did not align with any annotated transcripts in
  14923. the current build of the cynomolgus genome.
  14924. In the
  14925. \begin_inset Flex Glossary Term
  14926. status open
  14927. \begin_layout Plain Layout
  14928. GB
  14929. \end_layout
  14930. \end_inset
  14931. libraries, globin reads made up only 3.48% and reads assigned to all other
  14932. genes increased to 50.4%.
  14933. Thus,
  14934. \begin_inset Flex Glossary Term
  14935. status open
  14936. \begin_layout Plain Layout
  14937. GB
  14938. \end_layout
  14939. \end_inset
  14940. resulted in a 92.2% reduction in globin reads and a 91.6% increase in yield
  14941. of useful non-globin reads.
  14942. \end_layout
  14943. \begin_layout Standard
  14944. \begin_inset ERT
  14945. status open
  14946. \begin_layout Plain Layout
  14947. \backslash
  14948. afterpage{
  14949. \end_layout
  14950. \begin_layout Plain Layout
  14951. \backslash
  14952. begin{landscape}
  14953. \end_layout
  14954. \end_inset
  14955. \end_layout
  14956. \begin_layout Standard
  14957. \begin_inset Float table
  14958. placement p
  14959. wide false
  14960. sideways false
  14961. status collapsed
  14962. \begin_layout Plain Layout
  14963. \align center
  14964. \begin_inset Tabular
  14965. <lyxtabular version="3" rows="4" columns="7">
  14966. <features tabularvalignment="middle">
  14967. <column alignment="center" valignment="top">
  14968. <column alignment="center" valignment="top">
  14969. <column alignment="center" valignment="top">
  14970. <column alignment="center" valignment="top">
  14971. <column alignment="center" valignment="top">
  14972. <column alignment="center" valignment="top">
  14973. <column alignment="center" valignment="top">
  14974. <row>
  14975. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  14976. \begin_inset Text
  14977. \begin_layout Plain Layout
  14978. \end_layout
  14979. \end_inset
  14980. </cell>
  14981. <cell multicolumn="1" alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  14982. \begin_inset Text
  14983. \begin_layout Plain Layout
  14984. \family roman
  14985. \series medium
  14986. \shape up
  14987. \size normal
  14988. \emph off
  14989. \bar no
  14990. \strikeout off
  14991. \xout off
  14992. \uuline off
  14993. \uwave off
  14994. \noun off
  14995. \color none
  14996. Percent of Total Reads
  14997. \end_layout
  14998. \end_inset
  14999. </cell>
  15000. <cell multicolumn="2" alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  15001. \begin_inset Text
  15002. \begin_layout Plain Layout
  15003. \end_layout
  15004. \end_inset
  15005. </cell>
  15006. <cell multicolumn="2" alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  15007. \begin_inset Text
  15008. \begin_layout Plain Layout
  15009. \end_layout
  15010. \end_inset
  15011. </cell>
  15012. <cell multicolumn="2" alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  15013. \begin_inset Text
  15014. \begin_layout Plain Layout
  15015. \end_layout
  15016. \end_inset
  15017. </cell>
  15018. <cell multicolumn="1" alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  15019. \begin_inset Text
  15020. \begin_layout Plain Layout
  15021. \family roman
  15022. \series medium
  15023. \shape up
  15024. \size normal
  15025. \emph off
  15026. \bar no
  15027. \strikeout off
  15028. \xout off
  15029. \uuline off
  15030. \uwave off
  15031. \noun off
  15032. \color none
  15033. Percent of Genic Reads
  15034. \end_layout
  15035. \end_inset
  15036. </cell>
  15037. <cell multicolumn="2" alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  15038. \begin_inset Text
  15039. \begin_layout Plain Layout
  15040. \end_layout
  15041. \end_inset
  15042. </cell>
  15043. </row>
  15044. <row>
  15045. <cell alignment="center" valignment="top" bottomline="true" leftline="true" usebox="none">
  15046. \begin_inset Text
  15047. \begin_layout Plain Layout
  15048. GB
  15049. \end_layout
  15050. \end_inset
  15051. </cell>
  15052. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  15053. \begin_inset Text
  15054. \begin_layout Plain Layout
  15055. \family roman
  15056. \series medium
  15057. \shape up
  15058. \size normal
  15059. \emph off
  15060. \bar no
  15061. \strikeout off
  15062. \xout off
  15063. \uuline off
  15064. \uwave off
  15065. \noun off
  15066. \color none
  15067. Non-globin Reads
  15068. \end_layout
  15069. \end_inset
  15070. </cell>
  15071. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  15072. \begin_inset Text
  15073. \begin_layout Plain Layout
  15074. \family roman
  15075. \series medium
  15076. \shape up
  15077. \size normal
  15078. \emph off
  15079. \bar no
  15080. \strikeout off
  15081. \xout off
  15082. \uuline off
  15083. \uwave off
  15084. \noun off
  15085. \color none
  15086. Globin Reads
  15087. \end_layout
  15088. \end_inset
  15089. </cell>
  15090. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  15091. \begin_inset Text
  15092. \begin_layout Plain Layout
  15093. \family roman
  15094. \series medium
  15095. \shape up
  15096. \size normal
  15097. \emph off
  15098. \bar no
  15099. \strikeout off
  15100. \xout off
  15101. \uuline off
  15102. \uwave off
  15103. \noun off
  15104. \color none
  15105. All Genic Reads
  15106. \end_layout
  15107. \end_inset
  15108. </cell>
  15109. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  15110. \begin_inset Text
  15111. \begin_layout Plain Layout
  15112. \family roman
  15113. \series medium
  15114. \shape up
  15115. \size normal
  15116. \emph off
  15117. \bar no
  15118. \strikeout off
  15119. \xout off
  15120. \uuline off
  15121. \uwave off
  15122. \noun off
  15123. \color none
  15124. All Aligned Reads
  15125. \end_layout
  15126. \end_inset
  15127. </cell>
  15128. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  15129. \begin_inset Text
  15130. \begin_layout Plain Layout
  15131. \family roman
  15132. \series medium
  15133. \shape up
  15134. \size normal
  15135. \emph off
  15136. \bar no
  15137. \strikeout off
  15138. \xout off
  15139. \uuline off
  15140. \uwave off
  15141. \noun off
  15142. \color none
  15143. Non-globin Reads
  15144. \end_layout
  15145. \end_inset
  15146. </cell>
  15147. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  15148. \begin_inset Text
  15149. \begin_layout Plain Layout
  15150. \family roman
  15151. \series medium
  15152. \shape up
  15153. \size normal
  15154. \emph off
  15155. \bar no
  15156. \strikeout off
  15157. \xout off
  15158. \uuline off
  15159. \uwave off
  15160. \noun off
  15161. \color none
  15162. Globin Reads
  15163. \end_layout
  15164. \end_inset
  15165. </cell>
  15166. </row>
  15167. <row>
  15168. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  15169. \begin_inset Text
  15170. \begin_layout Plain Layout
  15171. \family roman
  15172. \series medium
  15173. \shape up
  15174. \size normal
  15175. \emph off
  15176. \bar no
  15177. \strikeout off
  15178. \xout off
  15179. \uuline off
  15180. \uwave off
  15181. \noun off
  15182. \color none
  15183. Yes
  15184. \end_layout
  15185. \end_inset
  15186. </cell>
  15187. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  15188. \begin_inset Text
  15189. \begin_layout Plain Layout
  15190. \family roman
  15191. \series medium
  15192. \shape up
  15193. \size normal
  15194. \emph off
  15195. \bar no
  15196. \strikeout off
  15197. \xout off
  15198. \uuline off
  15199. \uwave off
  15200. \noun off
  15201. \color none
  15202. 50.4% ± 6.82
  15203. \end_layout
  15204. \end_inset
  15205. </cell>
  15206. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  15207. \begin_inset Text
  15208. \begin_layout Plain Layout
  15209. \family roman
  15210. \series medium
  15211. \shape up
  15212. \size normal
  15213. \emph off
  15214. \bar no
  15215. \strikeout off
  15216. \xout off
  15217. \uuline off
  15218. \uwave off
  15219. \noun off
  15220. \color none
  15221. 3.48% ± 2.94
  15222. \end_layout
  15223. \end_inset
  15224. </cell>
  15225. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  15226. \begin_inset Text
  15227. \begin_layout Plain Layout
  15228. \family roman
  15229. \series medium
  15230. \shape up
  15231. \size normal
  15232. \emph off
  15233. \bar no
  15234. \strikeout off
  15235. \xout off
  15236. \uuline off
  15237. \uwave off
  15238. \noun off
  15239. \color none
  15240. 53.9% ± 6.81
  15241. \end_layout
  15242. \end_inset
  15243. </cell>
  15244. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  15245. \begin_inset Text
  15246. \begin_layout Plain Layout
  15247. \family roman
  15248. \series medium
  15249. \shape up
  15250. \size normal
  15251. \emph off
  15252. \bar no
  15253. \strikeout off
  15254. \xout off
  15255. \uuline off
  15256. \uwave off
  15257. \noun off
  15258. \color none
  15259. 89.7% ± 2.40
  15260. \end_layout
  15261. \end_inset
  15262. </cell>
  15263. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  15264. \begin_inset Text
  15265. \begin_layout Plain Layout
  15266. \family roman
  15267. \series medium
  15268. \shape up
  15269. \size normal
  15270. \emph off
  15271. \bar no
  15272. \strikeout off
  15273. \xout off
  15274. \uuline off
  15275. \uwave off
  15276. \noun off
  15277. \color none
  15278. 93.5% ± 5.25
  15279. \end_layout
  15280. \end_inset
  15281. </cell>
  15282. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  15283. \begin_inset Text
  15284. \begin_layout Plain Layout
  15285. \family roman
  15286. \series medium
  15287. \shape up
  15288. \size normal
  15289. \emph off
  15290. \bar no
  15291. \strikeout off
  15292. \xout off
  15293. \uuline off
  15294. \uwave off
  15295. \noun off
  15296. \color none
  15297. 6.49% ± 5.25
  15298. \end_layout
  15299. \end_inset
  15300. </cell>
  15301. </row>
  15302. <row>
  15303. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  15304. \begin_inset Text
  15305. \begin_layout Plain Layout
  15306. \family roman
  15307. \series medium
  15308. \shape up
  15309. \size normal
  15310. \emph off
  15311. \bar no
  15312. \strikeout off
  15313. \xout off
  15314. \uuline off
  15315. \uwave off
  15316. \noun off
  15317. \color none
  15318. No
  15319. \end_layout
  15320. \end_inset
  15321. </cell>
  15322. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  15323. \begin_inset Text
  15324. \begin_layout Plain Layout
  15325. \family roman
  15326. \series medium
  15327. \shape up
  15328. \size normal
  15329. \emph off
  15330. \bar no
  15331. \strikeout off
  15332. \xout off
  15333. \uuline off
  15334. \uwave off
  15335. \noun off
  15336. \color none
  15337. 26.3% ± 8.95
  15338. \end_layout
  15339. \end_inset
  15340. </cell>
  15341. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  15342. \begin_inset Text
  15343. \begin_layout Plain Layout
  15344. \family roman
  15345. \series medium
  15346. \shape up
  15347. \size normal
  15348. \emph off
  15349. \bar no
  15350. \strikeout off
  15351. \xout off
  15352. \uuline off
  15353. \uwave off
  15354. \noun off
  15355. \color none
  15356. 44.6% ± 16.6
  15357. \end_layout
  15358. \end_inset
  15359. </cell>
  15360. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  15361. \begin_inset Text
  15362. \begin_layout Plain Layout
  15363. \family roman
  15364. \series medium
  15365. \shape up
  15366. \size normal
  15367. \emph off
  15368. \bar no
  15369. \strikeout off
  15370. \xout off
  15371. \uuline off
  15372. \uwave off
  15373. \noun off
  15374. \color none
  15375. 70.1% ± 9.38
  15376. \end_layout
  15377. \end_inset
  15378. </cell>
  15379. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  15380. \begin_inset Text
  15381. \begin_layout Plain Layout
  15382. \family roman
  15383. \series medium
  15384. \shape up
  15385. \size normal
  15386. \emph off
  15387. \bar no
  15388. \strikeout off
  15389. \xout off
  15390. \uuline off
  15391. \uwave off
  15392. \noun off
  15393. \color none
  15394. 90.7% ± 5.16
  15395. \end_layout
  15396. \end_inset
  15397. </cell>
  15398. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  15399. \begin_inset Text
  15400. \begin_layout Plain Layout
  15401. \family roman
  15402. \series medium
  15403. \shape up
  15404. \size normal
  15405. \emph off
  15406. \bar no
  15407. \strikeout off
  15408. \xout off
  15409. \uuline off
  15410. \uwave off
  15411. \noun off
  15412. \color none
  15413. 38.8% ± 17.1
  15414. \end_layout
  15415. \end_inset
  15416. </cell>
  15417. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  15418. \begin_inset Text
  15419. \begin_layout Plain Layout
  15420. \family roman
  15421. \series medium
  15422. \shape up
  15423. \size normal
  15424. \emph off
  15425. \bar no
  15426. \strikeout off
  15427. \xout off
  15428. \uuline off
  15429. \uwave off
  15430. \noun off
  15431. \color none
  15432. 61.2% ± 17.1
  15433. \end_layout
  15434. \end_inset
  15435. </cell>
  15436. </row>
  15437. </lyxtabular>
  15438. \end_inset
  15439. \end_layout
  15440. \begin_layout Plain Layout
  15441. \begin_inset Caption Standard
  15442. \begin_layout Plain Layout
  15443. \begin_inset Argument 1
  15444. status collapsed
  15445. \begin_layout Plain Layout
  15446. Fractions of reads mapping to genomic features in GB and non-GB samples.
  15447. \end_layout
  15448. \end_inset
  15449. \begin_inset CommandInset label
  15450. LatexCommand label
  15451. name "tab:Fractions-of-reads"
  15452. \end_inset
  15453. \series bold
  15454. Fractions of reads mapping to genomic features in GB and non-GB samples.
  15455. \series default
  15456. All values are given as mean ± standard deviation.
  15457. \end_layout
  15458. \end_inset
  15459. \end_layout
  15460. \end_inset
  15461. \end_layout
  15462. \begin_layout Standard
  15463. \begin_inset ERT
  15464. status open
  15465. \begin_layout Plain Layout
  15466. \backslash
  15467. end{landscape}
  15468. \end_layout
  15469. \begin_layout Plain Layout
  15470. }
  15471. \end_layout
  15472. \end_inset
  15473. \end_layout
  15474. \begin_layout Standard
  15475. This reduction is not quite as efficient as the previous analysis showed
  15476. for human samples by DeepSAGE (<0.4% globin reads after globin reduction)
  15477. \begin_inset CommandInset citation
  15478. LatexCommand cite
  15479. key "Mastrokolias2012"
  15480. literal "false"
  15481. \end_inset
  15482. .
  15483. Nonetheless, this degree of globin reduction is sufficient to nearly double
  15484. the yield of useful reads.
  15485. Thus,
  15486. \begin_inset Flex Glossary Term
  15487. status open
  15488. \begin_layout Plain Layout
  15489. GB
  15490. \end_layout
  15491. \end_inset
  15492. cuts the required sequencing effort (and costs) to achieve a target coverage
  15493. depth by almost 50%.
  15494. Consistent with this near doubling of yield, the average difference in
  15495. un-normalized
  15496. \begin_inset Flex Glossary Term
  15497. status open
  15498. \begin_layout Plain Layout
  15499. logCPM
  15500. \end_layout
  15501. \end_inset
  15502. across all genes between the
  15503. \begin_inset Flex Glossary Term
  15504. status open
  15505. \begin_layout Plain Layout
  15506. GB
  15507. \end_layout
  15508. \end_inset
  15509. libraries and non-GB libraries is approximately 1 (mean = 1.01, median =
  15510. 1.08), an overall 2-fold increase.
  15511. Un-normalized values are used here because the
  15512. \begin_inset Flex Glossary Term
  15513. status open
  15514. \begin_layout Plain Layout
  15515. TMM
  15516. \end_layout
  15517. \end_inset
  15518. normalization correctly identifies this 2-fold difference as biologically
  15519. irrelevant and removes it.
  15520. \end_layout
  15521. \begin_layout Standard
  15522. Another important aspect is that the standard deviations in Table
  15523. \begin_inset CommandInset ref
  15524. LatexCommand ref
  15525. reference "tab:Fractions-of-reads"
  15526. plural "false"
  15527. caps "false"
  15528. noprefix "false"
  15529. \end_inset
  15530. are uniformly smaller in the
  15531. \begin_inset Flex Glossary Term
  15532. status open
  15533. \begin_layout Plain Layout
  15534. GB
  15535. \end_layout
  15536. \end_inset
  15537. samples than the non-GB ones, indicating much greater consistency of yield.
  15538. This is best seen in the percentage of non-globin reads as a fraction of
  15539. total reads aligned to annotated genes (genic reads).
  15540. For the non-GB samples, this measure ranges from 10.9% to 80.9%, while for
  15541. the
  15542. \begin_inset Flex Glossary Term
  15543. status open
  15544. \begin_layout Plain Layout
  15545. GB
  15546. \end_layout
  15547. \end_inset
  15548. samples it ranges from 81.9% to 99.9% (Figure
  15549. \begin_inset CommandInset ref
  15550. LatexCommand ref
  15551. reference "fig:Fraction-of-genic-reads"
  15552. plural "false"
  15553. caps "false"
  15554. noprefix "false"
  15555. \end_inset
  15556. \begin_inset Float figure
  15557. wide false
  15558. sideways false
  15559. status collapsed
  15560. \begin_layout Plain Layout
  15561. \align center
  15562. \begin_inset Graphics
  15563. filename graphics/globin-paper/figure1-globin-fractions.pdf
  15564. lyxscale 50
  15565. width 100col%
  15566. groupId colfullwidth
  15567. \end_inset
  15568. \end_layout
  15569. \begin_layout Plain Layout
  15570. \begin_inset Caption Standard
  15571. \begin_layout Plain Layout
  15572. \begin_inset Argument 1
  15573. status collapsed
  15574. \begin_layout Plain Layout
  15575. Fraction of genic reads in each sample aligned to non-globin genes, with
  15576. and without GB.
  15577. \end_layout
  15578. \end_inset
  15579. \begin_inset CommandInset label
  15580. LatexCommand label
  15581. name "fig:Fraction-of-genic-reads"
  15582. \end_inset
  15583. \series bold
  15584. Fraction of genic reads in each sample aligned to non-globin genes, with
  15585. and without GB.
  15586. \series default
  15587. All reads in each sequencing library were aligned to the cyno genome, and
  15588. the number of reads uniquely aligning to each gene was counted.
  15589. For each sample, counts were summed separately for all globin genes and
  15590. for the remainder of the genes (non-globin genes), and the fraction of
  15591. genic reads aligned to non-globin genes was computed.
  15592. Each point represents an individual sample.
  15593. Gray + signs indicate the means for globin-blocked libraries and unblocked
  15594. libraries.
  15595. The overall distribution for each group is represented as a notched box
  15596. plot.
  15597. Points are randomly spread vertically to avoid excessive overlapping.
  15598. \end_layout
  15599. \end_inset
  15600. \end_layout
  15601. \end_inset
  15602. \begin_inset Note Note
  15603. status open
  15604. \begin_layout Plain Layout
  15605. Float lost issues
  15606. \end_layout
  15607. \end_inset
  15608. ).
  15609. This means that for applications where it is critical that each sample
  15610. achieve a specified minimum coverage in order to provide useful information,
  15611. it would be necessary to budget up to 10 times the sequencing depth per
  15612. sample without
  15613. \begin_inset Flex Glossary Term
  15614. status open
  15615. \begin_layout Plain Layout
  15616. GB
  15617. \end_layout
  15618. \end_inset
  15619. , even though the average yield improvement for
  15620. \begin_inset Flex Glossary Term
  15621. status open
  15622. \begin_layout Plain Layout
  15623. GB
  15624. \end_layout
  15625. \end_inset
  15626. is only 2-fold, because every sample has a chance of being 90% globin and
  15627. 10% useful reads.
  15628. Hence, the more consistent behavior of
  15629. \begin_inset Flex Glossary Term
  15630. status open
  15631. \begin_layout Plain Layout
  15632. GB
  15633. \end_layout
  15634. \end_inset
  15635. samples makes planning an experiment easier and more efficient because
  15636. it eliminates the need to over-sequence every sample in order to guard
  15637. against the worst case of a high-globin fraction.
  15638. \end_layout
  15639. \begin_layout Subsection
  15640. Globin blocking lowers the noise floor and allows detection of about 2000
  15641. more low-expression genes
  15642. \end_layout
  15643. \begin_layout Standard
  15644. \begin_inset Flex TODO Note (inline)
  15645. status open
  15646. \begin_layout Plain Layout
  15647. Remove redundant titles from figures
  15648. \end_layout
  15649. \end_inset
  15650. \end_layout
  15651. \begin_layout Standard
  15652. Since
  15653. \begin_inset Flex Glossary Term
  15654. status open
  15655. \begin_layout Plain Layout
  15656. GB
  15657. \end_layout
  15658. \end_inset
  15659. yields more usable sequencing depth, it should also allow detection of
  15660. more genes at any given threshold.
  15661. When we looked at the distribution of average normalized
  15662. \begin_inset Flex Glossary Term
  15663. status open
  15664. \begin_layout Plain Layout
  15665. logCPM
  15666. \end_layout
  15667. \end_inset
  15668. values across all libraries for genes with at least one read assigned to
  15669. them, we observed the expected bimodal distribution, with a high-abundance
  15670. "signal" peak representing detected genes and a low-abundance "noise" peak
  15671. representing genes whose read count did not rise above the noise floor
  15672. (Figure
  15673. \begin_inset CommandInset ref
  15674. LatexCommand ref
  15675. reference "fig:logcpm-dists"
  15676. plural "false"
  15677. caps "false"
  15678. noprefix "false"
  15679. \end_inset
  15680. ).
  15681. Consistent with the 2-fold increase in raw counts assigned to non-globin
  15682. genes, the signal peak for
  15683. \begin_inset Flex Glossary Term
  15684. status open
  15685. \begin_layout Plain Layout
  15686. GB
  15687. \end_layout
  15688. \end_inset
  15689. samples is shifted to the right relative to the non-GB signal peak.
  15690. When all the samples are normalized together, this difference is normalized
  15691. out, lining up the signal peaks, and this reveals that, as expected, the
  15692. noise floor for the
  15693. \begin_inset Flex Glossary Term
  15694. status open
  15695. \begin_layout Plain Layout
  15696. GB
  15697. \end_layout
  15698. \end_inset
  15699. samples is about 2-fold lower.
  15700. This greater separation between signal and noise peaks in the
  15701. \begin_inset Flex Glossary Term
  15702. status open
  15703. \begin_layout Plain Layout
  15704. GB
  15705. \end_layout
  15706. \end_inset
  15707. samples means that low-expression genes should be more easily detected
  15708. and more precisely quantified than in the non-GB samples.
  15709. \end_layout
  15710. \begin_layout Standard
  15711. \begin_inset Float figure
  15712. wide false
  15713. sideways false
  15714. status open
  15715. \begin_layout Plain Layout
  15716. \align center
  15717. \begin_inset Graphics
  15718. filename graphics/globin-paper/figure2-aveLogCPM-colored.pdf
  15719. lyxscale 50
  15720. height 60theight%
  15721. \end_inset
  15722. \end_layout
  15723. \begin_layout Plain Layout
  15724. \begin_inset Caption Standard
  15725. \begin_layout Plain Layout
  15726. \begin_inset Argument 1
  15727. status collapsed
  15728. \begin_layout Plain Layout
  15729. Distributions of average group gene abundances when normalized separately
  15730. or together.
  15731. \end_layout
  15732. \end_inset
  15733. \begin_inset CommandInset label
  15734. LatexCommand label
  15735. name "fig:logcpm-dists"
  15736. \end_inset
  15737. \series bold
  15738. Distributions of average group gene abundances when normalized separately
  15739. or together.
  15740. \series default
  15741. All reads in each sequencing library were aligned to the cyno genome, and
  15742. the number of reads uniquely aligning to each gene was counted.
  15743. Genes with zero counts in all libraries were discarded.
  15744. Libraries were normalized using the TMM method.
  15745. Libraries were split into GB and non-GB groups and the average logCPM was
  15746. computed.
  15747. The distribution of average gene logCPM values was plotted for both groups
  15748. using a kernel density plot to approximate a continuous distribution.
  15749. The GB logCPM distributions are marked in red, non-GB in blue.
  15750. The black vertical line denotes the chosen detection threshold of
  15751. \begin_inset Formula $-1$
  15752. \end_inset
  15753. .
  15754. Top panel: Libraries were split into GB and non-GB groups first and normalized
  15755. separately.
  15756. Bottom panel: Libraries were all normalized together first and then split
  15757. into groups.
  15758. \end_layout
  15759. \end_inset
  15760. \end_layout
  15761. \end_inset
  15762. \end_layout
  15763. \begin_layout Standard
  15764. Based on these distributions, we selected a detection threshold of
  15765. \begin_inset Formula $-1$
  15766. \end_inset
  15767. , which is approximately the leftmost edge of the trough between the signal
  15768. and noise peaks.
  15769. This represents the most liberal possible detection threshold that doesn't
  15770. call substantial numbers of noise genes as detected.
  15771. Among the full dataset, 13429 genes were detected at this threshold, and
  15772. 22276 were not.
  15773. When considering the
  15774. \begin_inset Flex Glossary Term
  15775. status open
  15776. \begin_layout Plain Layout
  15777. GB
  15778. \end_layout
  15779. \end_inset
  15780. libraries and non-GB libraries separately and re-computing normalization
  15781. factors independently within each group, 14535 genes were detected in the
  15782. \begin_inset Flex Glossary Term
  15783. status open
  15784. \begin_layout Plain Layout
  15785. GB
  15786. \end_layout
  15787. \end_inset
  15788. libraries while only 12460 were detected in the non-GB libraries.
  15789. Thus,
  15790. \begin_inset Flex Glossary Term
  15791. status open
  15792. \begin_layout Plain Layout
  15793. GB
  15794. \end_layout
  15795. \end_inset
  15796. allowed the detection of 2000 extra genes that were buried under the noise
  15797. floor without
  15798. \begin_inset Flex Glossary Term
  15799. status open
  15800. \begin_layout Plain Layout
  15801. GB
  15802. \end_layout
  15803. \end_inset
  15804. .
  15805. This pattern of at least 2000 additional genes detected with
  15806. \begin_inset Flex Glossary Term
  15807. status open
  15808. \begin_layout Plain Layout
  15809. GB
  15810. \end_layout
  15811. \end_inset
  15812. was also consistent across a wide range of possible detection thresholds,
  15813. from -2 to 3 (see Figure
  15814. \begin_inset CommandInset ref
  15815. LatexCommand ref
  15816. reference "fig:Gene-detections"
  15817. plural "false"
  15818. caps "false"
  15819. noprefix "false"
  15820. \end_inset
  15821. ).
  15822. \end_layout
  15823. \begin_layout Standard
  15824. \begin_inset Float figure
  15825. wide false
  15826. sideways false
  15827. status open
  15828. \begin_layout Plain Layout
  15829. \align center
  15830. \begin_inset Graphics
  15831. filename graphics/globin-paper/figure3-detection.pdf
  15832. lyxscale 50
  15833. width 70col%
  15834. \end_inset
  15835. \end_layout
  15836. \begin_layout Plain Layout
  15837. \begin_inset Caption Standard
  15838. \begin_layout Plain Layout
  15839. \begin_inset Argument 1
  15840. status collapsed
  15841. \begin_layout Plain Layout
  15842. Gene detections as a function of abundance thresholds in GB and non-GB samples.
  15843. \end_layout
  15844. \end_inset
  15845. \begin_inset CommandInset label
  15846. LatexCommand label
  15847. name "fig:Gene-detections"
  15848. \end_inset
  15849. \series bold
  15850. Gene detections as a function of abundance thresholds in GB and non-GB samples.
  15851. \series default
  15852. Average logCPM was computed by separate group normalization as described
  15853. in Figure
  15854. \begin_inset CommandInset ref
  15855. LatexCommand ref
  15856. reference "fig:logcpm-dists"
  15857. plural "false"
  15858. caps "false"
  15859. noprefix "false"
  15860. \end_inset
  15861. for both the GB and non-GB groups, as well as for all samples considered
  15862. as one large group.
  15863. For each every integer threshold from
  15864. \begin_inset Formula $-2$
  15865. \end_inset
  15866. to 3, the number of genes detected at or above that logCPM threshold was
  15867. plotted for each group.
  15868. \end_layout
  15869. \end_inset
  15870. \end_layout
  15871. \end_inset
  15872. \end_layout
  15873. \begin_layout Subsection
  15874. Globin blocking does not add significant additional noise or decrease sample
  15875. quality
  15876. \end_layout
  15877. \begin_layout Standard
  15878. One potential worry is that the
  15879. \begin_inset Flex Glossary Term
  15880. status open
  15881. \begin_layout Plain Layout
  15882. GB
  15883. \end_layout
  15884. \end_inset
  15885. protocol could perturb the levels of non-globin genes.
  15886. There are two kinds of possible perturbations: systematic and random.
  15887. The former is not a major concern for detection of differential expression,
  15888. since a 2-fold change in every sample has no effect on the relative fold
  15889. change between samples.
  15890. In contrast, random perturbations would increase the noise and obscure
  15891. the signal in the dataset, reducing the capacity to detect differential
  15892. expression.
  15893. \end_layout
  15894. \begin_layout Standard
  15895. \begin_inset Flex TODO Note (inline)
  15896. status open
  15897. \begin_layout Plain Layout
  15898. Standardize on
  15899. \begin_inset Quotes eld
  15900. \end_inset
  15901. log2
  15902. \begin_inset Quotes erd
  15903. \end_inset
  15904. notation
  15905. \end_layout
  15906. \end_inset
  15907. \end_layout
  15908. \begin_layout Standard
  15909. The data do indeed show small systematic perturbations in gene levels (Figure
  15910. \begin_inset CommandInset ref
  15911. LatexCommand ref
  15912. reference "fig:MA-plot"
  15913. plural "false"
  15914. caps "false"
  15915. noprefix "false"
  15916. \end_inset
  15917. ).
  15918. Other than the 3 designated alpha and beta globin genes, two other genes
  15919. stand out as having especially large negative
  15920. \begin_inset Flex Glossary Term (pl)
  15921. status open
  15922. \begin_layout Plain Layout
  15923. logFC
  15924. \end_layout
  15925. \end_inset
  15926. : HBD and LOC1021365.
  15927. HBD, delta globin, is most likely targeted by the blocking
  15928. \begin_inset Flex Glossary Term (pl)
  15929. status open
  15930. \begin_layout Plain Layout
  15931. oligo
  15932. \end_layout
  15933. \end_inset
  15934. due to high sequence homology with the other globin genes.
  15935. LOC1021365 is the aforementioned
  15936. \begin_inset Flex Glossary Term
  15937. status open
  15938. \begin_layout Plain Layout
  15939. ncRNA
  15940. \end_layout
  15941. \end_inset
  15942. that is reverse-complementary to one of the alpha-like genes and that would
  15943. be expected to be removed during the
  15944. \begin_inset Flex Glossary Term
  15945. status open
  15946. \begin_layout Plain Layout
  15947. GB
  15948. \end_layout
  15949. \end_inset
  15950. step.
  15951. All other genes appear in a cluster centered vertically at 0, and the vast
  15952. majority of genes in this cluster show an absolute
  15953. \begin_inset Flex Glossary Term
  15954. status open
  15955. \begin_layout Plain Layout
  15956. logFC
  15957. \end_layout
  15958. \end_inset
  15959. of 0.5 or less.
  15960. Nevertheless, many of these small perturbations are still statistically
  15961. significant, indicating that the
  15962. \begin_inset Flex Glossary Term
  15963. status open
  15964. \begin_layout Plain Layout
  15965. GB
  15966. \end_layout
  15967. \end_inset
  15968. \begin_inset Flex Glossary Term (pl)
  15969. status open
  15970. \begin_layout Plain Layout
  15971. oligo
  15972. \end_layout
  15973. \end_inset
  15974. likely cause very small but non-zero systematic perturbations in measured
  15975. gene expression levels.
  15976. \end_layout
  15977. \begin_layout Standard
  15978. \begin_inset Float figure
  15979. wide false
  15980. sideways false
  15981. status open
  15982. \begin_layout Plain Layout
  15983. \align center
  15984. \begin_inset Graphics
  15985. filename graphics/globin-paper/figure4-maplot-colored.pdf
  15986. lyxscale 50
  15987. width 100col%
  15988. groupId colfullwidth
  15989. \end_inset
  15990. \end_layout
  15991. \begin_layout Plain Layout
  15992. \begin_inset Caption Standard
  15993. \begin_layout Plain Layout
  15994. \begin_inset Argument 1
  15995. status collapsed
  15996. \begin_layout Plain Layout
  15997. MA plot showing effects of GB on each gene's abundance.
  15998. \end_layout
  15999. \end_inset
  16000. \begin_inset CommandInset label
  16001. LatexCommand label
  16002. name "fig:MA-plot"
  16003. \end_inset
  16004. \series bold
  16005. MA plot showing effects of GB on each gene's abundance.
  16006. \series default
  16007. All libraries were normalized together as described in Figure
  16008. \begin_inset CommandInset ref
  16009. LatexCommand ref
  16010. reference "fig:logcpm-dists"
  16011. plural "false"
  16012. caps "false"
  16013. noprefix "false"
  16014. \end_inset
  16015. , and genes with an average logCPM below
  16016. \begin_inset Formula $-1$
  16017. \end_inset
  16018. were filtered out.
  16019. Each remaining gene was tested for differential abundance with respect
  16020. to
  16021. \begin_inset Flex Glossary Term (glstext)
  16022. status open
  16023. \begin_layout Plain Layout
  16024. GB
  16025. \end_layout
  16026. \end_inset
  16027. using
  16028. \begin_inset Flex Code
  16029. status open
  16030. \begin_layout Plain Layout
  16031. edgeR
  16032. \end_layout
  16033. \end_inset
  16034. ’s quasi-likelihood F-test, fitting a NB GLM to table of read counts in
  16035. each library.
  16036. For each gene,
  16037. \begin_inset Flex Code
  16038. status open
  16039. \begin_layout Plain Layout
  16040. edgeR
  16041. \end_layout
  16042. \end_inset
  16043. reported average logCPM, logFC, p-value, and BH-adjusted FDR.
  16044. Each gene's logFC was plotted against its logCPM, colored by FDR.
  16045. Red points are significant at
  16046. \begin_inset Formula $≤10\%$
  16047. \end_inset
  16048. FDR, and blue are not significant at that threshold.
  16049. The alpha and beta globin genes targeted for blocking are marked with large
  16050. triangles, while all other genes are represented as small points.
  16051. \end_layout
  16052. \end_inset
  16053. \end_layout
  16054. \end_inset
  16055. \end_layout
  16056. \begin_layout Standard
  16057. \begin_inset Flex TODO Note (inline)
  16058. status open
  16059. \begin_layout Plain Layout
  16060. Give these numbers the LaTeX math treatment
  16061. \end_layout
  16062. \end_inset
  16063. \end_layout
  16064. \begin_layout Standard
  16065. To evaluate the possibility of
  16066. \begin_inset Flex Glossary Term
  16067. status open
  16068. \begin_layout Plain Layout
  16069. GB
  16070. \end_layout
  16071. \end_inset
  16072. causing random perturbations and reducing sample quality, we computed the
  16073. Pearson correlation between
  16074. \begin_inset Flex Glossary Term
  16075. status open
  16076. \begin_layout Plain Layout
  16077. logCPM
  16078. \end_layout
  16079. \end_inset
  16080. values for every pair of samples with and without
  16081. \begin_inset Flex Glossary Term
  16082. status open
  16083. \begin_layout Plain Layout
  16084. GB
  16085. \end_layout
  16086. \end_inset
  16087. and plotted them against each other (Figure
  16088. \begin_inset CommandInset ref
  16089. LatexCommand ref
  16090. reference "fig:gene-abundance-correlations"
  16091. plural "false"
  16092. caps "false"
  16093. noprefix "false"
  16094. \end_inset
  16095. ).
  16096. The plot indicated that the
  16097. \begin_inset Flex Glossary Term
  16098. status open
  16099. \begin_layout Plain Layout
  16100. GB
  16101. \end_layout
  16102. \end_inset
  16103. libraries have higher sample-to-sample correlations than the non-GB libraries.
  16104. Parametric and nonparametric tests for differences between the correlations
  16105. with and without
  16106. \begin_inset Flex Glossary Term
  16107. status open
  16108. \begin_layout Plain Layout
  16109. GB
  16110. \end_layout
  16111. \end_inset
  16112. both confirmed that this difference was highly significant (2-sided paired
  16113. t-test:
  16114. \begin_inset Formula $t=37.2$
  16115. \end_inset
  16116. ,
  16117. \begin_inset Formula $d.f.=665$
  16118. \end_inset
  16119. ,
  16120. \begin_inset Formula $P\ll2.2\times10^{-16}$
  16121. \end_inset
  16122. ; 2-sided Wilcoxon sign-rank test:
  16123. \begin_inset Formula $V=2195$
  16124. \end_inset
  16125. ,
  16126. \begin_inset Formula $P\ll2.2\times10^{-16}$
  16127. \end_inset
  16128. ).
  16129. Performing the same tests on the Spearman correlations gave the same conclusion
  16130. (t-test:
  16131. \begin_inset Formula $t=26.8$
  16132. \end_inset
  16133. ,
  16134. \begin_inset Formula $d.f.=665$
  16135. \end_inset
  16136. ,
  16137. \begin_inset Formula $P\ll2.2\times10^{-16}$
  16138. \end_inset
  16139. ; sign-rank test:
  16140. \begin_inset Formula $V=8781$
  16141. \end_inset
  16142. ,
  16143. \begin_inset Formula $P\ll2.2\times10^{-16}$
  16144. \end_inset
  16145. ).
  16146. The
  16147. \begin_inset Flex Code
  16148. status open
  16149. \begin_layout Plain Layout
  16150. edgeR
  16151. \end_layout
  16152. \end_inset
  16153. package was used to compute the overall
  16154. \begin_inset Flex Glossary Term
  16155. status open
  16156. \begin_layout Plain Layout
  16157. BCV
  16158. \end_layout
  16159. \end_inset
  16160. for
  16161. \begin_inset Flex Glossary Term
  16162. status open
  16163. \begin_layout Plain Layout
  16164. GB
  16165. \end_layout
  16166. \end_inset
  16167. and non-GB libraries, and found that
  16168. \begin_inset Flex Glossary Term
  16169. status open
  16170. \begin_layout Plain Layout
  16171. GB
  16172. \end_layout
  16173. \end_inset
  16174. resulted in a negligible increase in the
  16175. \begin_inset Flex Glossary Term
  16176. status open
  16177. \begin_layout Plain Layout
  16178. BCV
  16179. \end_layout
  16180. \end_inset
  16181. (0.417 with
  16182. \begin_inset Flex Glossary Term
  16183. status open
  16184. \begin_layout Plain Layout
  16185. GB
  16186. \end_layout
  16187. \end_inset
  16188. vs.
  16189. 0.400 without).
  16190. The near equality of the
  16191. \begin_inset Flex Glossary Term
  16192. status open
  16193. \begin_layout Plain Layout
  16194. BCV
  16195. \end_layout
  16196. \end_inset
  16197. for both sets indicates that the higher correlations in the
  16198. \begin_inset Flex Glossary Term
  16199. status open
  16200. \begin_layout Plain Layout
  16201. GB
  16202. \end_layout
  16203. \end_inset
  16204. libraries are most likely a result of the increased yield of useful reads,
  16205. which reduces the contribution of Poisson counting uncertainty to the overall
  16206. variance of the
  16207. \begin_inset Flex Glossary Term
  16208. status open
  16209. \begin_layout Plain Layout
  16210. logCPM
  16211. \end_layout
  16212. \end_inset
  16213. values
  16214. \begin_inset CommandInset citation
  16215. LatexCommand cite
  16216. key "McCarthy2012"
  16217. literal "false"
  16218. \end_inset
  16219. .
  16220. This improves the precision of expression measurements and more than offsets
  16221. the negligible increase in
  16222. \begin_inset Flex Glossary Term
  16223. status open
  16224. \begin_layout Plain Layout
  16225. BCV
  16226. \end_layout
  16227. \end_inset
  16228. .
  16229. \end_layout
  16230. \begin_layout Standard
  16231. \begin_inset Float figure
  16232. wide false
  16233. sideways false
  16234. status open
  16235. \begin_layout Plain Layout
  16236. \align center
  16237. \begin_inset Graphics
  16238. filename graphics/globin-paper/figure5-corrplot.pdf
  16239. lyxscale 50
  16240. width 100col%
  16241. groupId colfullwidth
  16242. \end_inset
  16243. \end_layout
  16244. \begin_layout Plain Layout
  16245. \begin_inset Caption Standard
  16246. \begin_layout Plain Layout
  16247. \begin_inset Argument 1
  16248. status collapsed
  16249. \begin_layout Plain Layout
  16250. Comparison of inter-sample gene abundance correlations with and without
  16251. GB.
  16252. \end_layout
  16253. \end_inset
  16254. \begin_inset CommandInset label
  16255. LatexCommand label
  16256. name "fig:gene-abundance-correlations"
  16257. \end_inset
  16258. \series bold
  16259. Comparison of inter-sample gene abundance correlations with and without
  16260. GB.
  16261. \series default
  16262. All libraries were normalized together as described in Figure
  16263. \begin_inset CommandInset ref
  16264. LatexCommand ref
  16265. reference "fig:logcpm-dists"
  16266. plural "false"
  16267. caps "false"
  16268. noprefix "false"
  16269. \end_inset
  16270. , and genes with an average logCPM less than
  16271. \begin_inset Formula $-1$
  16272. \end_inset
  16273. were filtered out.
  16274. Each gene’s logCPM was computed in each library using
  16275. \begin_inset Flex Code
  16276. status open
  16277. \begin_layout Plain Layout
  16278. edgeR
  16279. \end_layout
  16280. \end_inset
  16281. 's
  16282. \begin_inset Flex Code
  16283. status open
  16284. \begin_layout Plain Layout
  16285. cpm
  16286. \end_layout
  16287. \end_inset
  16288. function.
  16289. For each pair of biological samples, the Pearson correlation between those
  16290. samples' GB libraries was plotted against the correlation between the same
  16291. samples’ non-GB libraries.
  16292. Each point represents an unique pair of samples.
  16293. The solid gray line shows a quantile-quantile plot of distribution of GB
  16294. correlations vs.
  16295. that of non-GB correlations.
  16296. The thin dashed line is the identity line, provided for reference.
  16297. \end_layout
  16298. \end_inset
  16299. \end_layout
  16300. \end_inset
  16301. \end_layout
  16302. \begin_layout Subsection
  16303. More differentially expressed genes are detected with globin blocking
  16304. \end_layout
  16305. \begin_layout Standard
  16306. To compare performance on differential gene expression tests, we took subsets
  16307. of both the
  16308. \begin_inset Flex Glossary Term
  16309. status open
  16310. \begin_layout Plain Layout
  16311. GB
  16312. \end_layout
  16313. \end_inset
  16314. and non-GB libraries with exactly one pre-transplant and one post-transplant
  16315. sample for each animal that had paired samples available for analysis (
  16316. \begin_inset Formula $N=7$
  16317. \end_inset
  16318. animals,
  16319. \begin_inset Formula $N=14$
  16320. \end_inset
  16321. samples in each subset).
  16322. The same test for pre- vs.
  16323. post-transplant differential gene expression was performed on the same
  16324. 7 pairs of samples from
  16325. \begin_inset Flex Glossary Term
  16326. status open
  16327. \begin_layout Plain Layout
  16328. GB
  16329. \end_layout
  16330. \end_inset
  16331. libraries and non-GB libraries, in each case using an
  16332. \begin_inset Flex Glossary Term
  16333. status open
  16334. \begin_layout Plain Layout
  16335. FDR
  16336. \end_layout
  16337. \end_inset
  16338. of 10% as the threshold of significance.
  16339. Out of 12,954 genes that passed the detection threshold in both subsets,
  16340. 358 were called significantly differentially expressed in the same direction
  16341. in both sets; 1063 were differentially expressed in the
  16342. \begin_inset Flex Glossary Term
  16343. status open
  16344. \begin_layout Plain Layout
  16345. GB
  16346. \end_layout
  16347. \end_inset
  16348. set only; 296 were differentially expressed in the non-GB set only; 2 genes
  16349. were called significantly up in the
  16350. \begin_inset Flex Glossary Term
  16351. status open
  16352. \begin_layout Plain Layout
  16353. GB
  16354. \end_layout
  16355. \end_inset
  16356. set but significantly down in the non-GB set; and the remaining 11,235
  16357. were not called differentially expressed in either set.
  16358. These data are summarized in Table
  16359. \begin_inset CommandInset ref
  16360. LatexCommand ref
  16361. reference "tab:Comparison-of-significant"
  16362. plural "false"
  16363. caps "false"
  16364. noprefix "false"
  16365. \end_inset
  16366. .
  16367. The differences in
  16368. \begin_inset Flex Glossary Term
  16369. status open
  16370. \begin_layout Plain Layout
  16371. BCV
  16372. \end_layout
  16373. \end_inset
  16374. calculated by
  16375. \begin_inset Flex Code
  16376. status open
  16377. \begin_layout Plain Layout
  16378. edgeR
  16379. \end_layout
  16380. \end_inset
  16381. for these subsets of samples were negligible (
  16382. \begin_inset Formula $\textrm{BCV}=0.302$
  16383. \end_inset
  16384. for
  16385. \begin_inset Flex Glossary Term
  16386. status open
  16387. \begin_layout Plain Layout
  16388. GB
  16389. \end_layout
  16390. \end_inset
  16391. and 0.297 for non-GB).
  16392. \end_layout
  16393. \begin_layout Standard
  16394. \begin_inset Float table
  16395. wide false
  16396. sideways false
  16397. status collapsed
  16398. \begin_layout Plain Layout
  16399. \align center
  16400. \begin_inset Tabular
  16401. <lyxtabular version="3" rows="5" columns="5">
  16402. <features tabularvalignment="middle">
  16403. <column alignment="center" valignment="top">
  16404. <column alignment="center" valignment="top">
  16405. <column alignment="center" valignment="top">
  16406. <column alignment="center" valignment="top">
  16407. <column alignment="center" valignment="top">
  16408. <row>
  16409. <cell alignment="center" valignment="top" usebox="none">
  16410. \begin_inset Text
  16411. \begin_layout Plain Layout
  16412. \end_layout
  16413. \end_inset
  16414. </cell>
  16415. <cell alignment="center" valignment="top" usebox="none">
  16416. \begin_inset Text
  16417. \begin_layout Plain Layout
  16418. \end_layout
  16419. \end_inset
  16420. </cell>
  16421. <cell multicolumn="1" alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  16422. \begin_inset Text
  16423. \begin_layout Plain Layout
  16424. \series bold
  16425. No Globin Blocking
  16426. \end_layout
  16427. \end_inset
  16428. </cell>
  16429. <cell multicolumn="2" alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  16430. \begin_inset Text
  16431. \begin_layout Plain Layout
  16432. \end_layout
  16433. \end_inset
  16434. </cell>
  16435. <cell multicolumn="2" alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  16436. \begin_inset Text
  16437. \begin_layout Plain Layout
  16438. \end_layout
  16439. \end_inset
  16440. </cell>
  16441. </row>
  16442. <row>
  16443. <cell alignment="center" valignment="top" usebox="none">
  16444. \begin_inset Text
  16445. \begin_layout Plain Layout
  16446. \end_layout
  16447. \end_inset
  16448. </cell>
  16449. <cell alignment="center" valignment="top" usebox="none">
  16450. \begin_inset Text
  16451. \begin_layout Plain Layout
  16452. \end_layout
  16453. \end_inset
  16454. </cell>
  16455. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  16456. \begin_inset Text
  16457. \begin_layout Plain Layout
  16458. \series bold
  16459. Up
  16460. \end_layout
  16461. \end_inset
  16462. </cell>
  16463. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  16464. \begin_inset Text
  16465. \begin_layout Plain Layout
  16466. \series bold
  16467. NS
  16468. \end_layout
  16469. \end_inset
  16470. </cell>
  16471. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  16472. \begin_inset Text
  16473. \begin_layout Plain Layout
  16474. \series bold
  16475. Down
  16476. \end_layout
  16477. \end_inset
  16478. </cell>
  16479. </row>
  16480. <row>
  16481. <cell multirow="3" alignment="center" valignment="middle" topline="true" bottomline="true" leftline="true" usebox="none">
  16482. \begin_inset Text
  16483. \begin_layout Plain Layout
  16484. \series bold
  16485. Globin-Blocking
  16486. \end_layout
  16487. \end_inset
  16488. </cell>
  16489. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  16490. \begin_inset Text
  16491. \begin_layout Plain Layout
  16492. \series bold
  16493. Up
  16494. \end_layout
  16495. \end_inset
  16496. </cell>
  16497. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  16498. \begin_inset Text
  16499. \begin_layout Plain Layout
  16500. \family roman
  16501. \series medium
  16502. \shape up
  16503. \size normal
  16504. \emph off
  16505. \bar no
  16506. \strikeout off
  16507. \xout off
  16508. \uuline off
  16509. \uwave off
  16510. \noun off
  16511. \color none
  16512. 231
  16513. \end_layout
  16514. \end_inset
  16515. </cell>
  16516. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  16517. \begin_inset Text
  16518. \begin_layout Plain Layout
  16519. \family roman
  16520. \series medium
  16521. \shape up
  16522. \size normal
  16523. \emph off
  16524. \bar no
  16525. \strikeout off
  16526. \xout off
  16527. \uuline off
  16528. \uwave off
  16529. \noun off
  16530. \color none
  16531. 515
  16532. \end_layout
  16533. \end_inset
  16534. </cell>
  16535. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  16536. \begin_inset Text
  16537. \begin_layout Plain Layout
  16538. \family roman
  16539. \series medium
  16540. \shape up
  16541. \size normal
  16542. \emph off
  16543. \bar no
  16544. \strikeout off
  16545. \xout off
  16546. \uuline off
  16547. \uwave off
  16548. \noun off
  16549. \color none
  16550. 2
  16551. \end_layout
  16552. \end_inset
  16553. </cell>
  16554. </row>
  16555. <row>
  16556. <cell multirow="4" alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  16557. \begin_inset Text
  16558. \begin_layout Plain Layout
  16559. \end_layout
  16560. \end_inset
  16561. </cell>
  16562. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  16563. \begin_inset Text
  16564. \begin_layout Plain Layout
  16565. \series bold
  16566. NS
  16567. \end_layout
  16568. \end_inset
  16569. </cell>
  16570. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  16571. \begin_inset Text
  16572. \begin_layout Plain Layout
  16573. \family roman
  16574. \series medium
  16575. \shape up
  16576. \size normal
  16577. \emph off
  16578. \bar no
  16579. \strikeout off
  16580. \xout off
  16581. \uuline off
  16582. \uwave off
  16583. \noun off
  16584. \color none
  16585. 160
  16586. \end_layout
  16587. \end_inset
  16588. </cell>
  16589. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  16590. \begin_inset Text
  16591. \begin_layout Plain Layout
  16592. \family roman
  16593. \series medium
  16594. \shape up
  16595. \size normal
  16596. \emph off
  16597. \bar no
  16598. \strikeout off
  16599. \xout off
  16600. \uuline off
  16601. \uwave off
  16602. \noun off
  16603. \color none
  16604. 11235
  16605. \end_layout
  16606. \end_inset
  16607. </cell>
  16608. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  16609. \begin_inset Text
  16610. \begin_layout Plain Layout
  16611. \family roman
  16612. \series medium
  16613. \shape up
  16614. \size normal
  16615. \emph off
  16616. \bar no
  16617. \strikeout off
  16618. \xout off
  16619. \uuline off
  16620. \uwave off
  16621. \noun off
  16622. \color none
  16623. 136
  16624. \end_layout
  16625. \end_inset
  16626. </cell>
  16627. </row>
  16628. <row>
  16629. <cell multirow="4" alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  16630. \begin_inset Text
  16631. \begin_layout Plain Layout
  16632. \end_layout
  16633. \end_inset
  16634. </cell>
  16635. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  16636. \begin_inset Text
  16637. \begin_layout Plain Layout
  16638. \series bold
  16639. Down
  16640. \end_layout
  16641. \end_inset
  16642. </cell>
  16643. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  16644. \begin_inset Text
  16645. \begin_layout Plain Layout
  16646. \family roman
  16647. \series medium
  16648. \shape up
  16649. \size normal
  16650. \emph off
  16651. \bar no
  16652. \strikeout off
  16653. \xout off
  16654. \uuline off
  16655. \uwave off
  16656. \noun off
  16657. \color none
  16658. 0
  16659. \end_layout
  16660. \end_inset
  16661. </cell>
  16662. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  16663. \begin_inset Text
  16664. \begin_layout Plain Layout
  16665. \family roman
  16666. \series medium
  16667. \shape up
  16668. \size normal
  16669. \emph off
  16670. \bar no
  16671. \strikeout off
  16672. \xout off
  16673. \uuline off
  16674. \uwave off
  16675. \noun off
  16676. \color none
  16677. 548
  16678. \end_layout
  16679. \end_inset
  16680. </cell>
  16681. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  16682. \begin_inset Text
  16683. \begin_layout Plain Layout
  16684. \family roman
  16685. \series medium
  16686. \shape up
  16687. \size normal
  16688. \emph off
  16689. \bar no
  16690. \strikeout off
  16691. \xout off
  16692. \uuline off
  16693. \uwave off
  16694. \noun off
  16695. \color none
  16696. 127
  16697. \end_layout
  16698. \end_inset
  16699. </cell>
  16700. </row>
  16701. </lyxtabular>
  16702. \end_inset
  16703. \end_layout
  16704. \begin_layout Plain Layout
  16705. \begin_inset Caption Standard
  16706. \begin_layout Plain Layout
  16707. \begin_inset Argument 1
  16708. status collapsed
  16709. \begin_layout Plain Layout
  16710. Comparison of significantly differentially expressed genes with and without
  16711. globin blocking.
  16712. \end_layout
  16713. \end_inset
  16714. \begin_inset CommandInset label
  16715. LatexCommand label
  16716. name "tab:Comparison-of-significant"
  16717. \end_inset
  16718. \series bold
  16719. Comparison of significantly differentially expressed genes with and without
  16720. globin blocking.
  16721. \series default
  16722. Up, Down: Genes significantly up/down-regulated in post-transplant samples
  16723. relative to pre-transplant samples, with a false discovery rate of 10%
  16724. or less.
  16725. NS: Non-significant genes (false discovery rate greater than 10%).
  16726. \end_layout
  16727. \end_inset
  16728. \end_layout
  16729. \end_inset
  16730. \end_layout
  16731. \begin_layout Standard
  16732. The key point is that the
  16733. \begin_inset Flex Glossary Term
  16734. status open
  16735. \begin_layout Plain Layout
  16736. GB
  16737. \end_layout
  16738. \end_inset
  16739. data results in substantially more differentially expressed calls than
  16740. the non-GB data.
  16741. Since there is no gold standard for this dataset, it is impossible to be
  16742. certain whether this is due to under-calling of differential expression
  16743. in the non-GB samples or over-calling in the
  16744. \begin_inset Flex Glossary Term
  16745. status open
  16746. \begin_layout Plain Layout
  16747. GB
  16748. \end_layout
  16749. \end_inset
  16750. samples.
  16751. However, given that both datasets are derived from the same biological
  16752. samples and have nearly equal
  16753. \begin_inset Flex Glossary Term (pl)
  16754. status open
  16755. \begin_layout Plain Layout
  16756. BCV
  16757. \end_layout
  16758. \end_inset
  16759. , it is more likely that the larger number of differential expression calls
  16760. in the
  16761. \begin_inset Flex Glossary Term
  16762. status open
  16763. \begin_layout Plain Layout
  16764. GB
  16765. \end_layout
  16766. \end_inset
  16767. samples are genuine detections that were enabled by the higher sequencing
  16768. depth and measurement precision of the
  16769. \begin_inset Flex Glossary Term
  16770. status open
  16771. \begin_layout Plain Layout
  16772. GB
  16773. \end_layout
  16774. \end_inset
  16775. samples.
  16776. Note that the same set of genes was considered in both subsets, so the
  16777. larger number of differentially expressed gene calls in the
  16778. \begin_inset Flex Glossary Term
  16779. status open
  16780. \begin_layout Plain Layout
  16781. GB
  16782. \end_layout
  16783. \end_inset
  16784. data set reflects a greater sensitivity to detect significant differential
  16785. gene expression and not simply the larger total number of detected genes
  16786. in
  16787. \begin_inset Flex Glossary Term
  16788. status open
  16789. \begin_layout Plain Layout
  16790. GB
  16791. \end_layout
  16792. \end_inset
  16793. samples described earlier.
  16794. \end_layout
  16795. \begin_layout Section
  16796. Discussion
  16797. \end_layout
  16798. \begin_layout Standard
  16799. The original experience with whole blood gene expression profiling on DNA
  16800. microarrays demonstrated that the high concentration of globin transcripts
  16801. reduced the sensitivity to detect genes with relatively low expression
  16802. levels, in effect, significantly reducing the sensitivity.
  16803. To address this limitation, commercial protocols for globin reduction were
  16804. developed based on strategies to block globin transcript amplification
  16805. during labeling or physically removing globin transcripts by affinity bead
  16806. methods
  16807. \begin_inset CommandInset citation
  16808. LatexCommand cite
  16809. key "Winn2010"
  16810. literal "false"
  16811. \end_inset
  16812. .
  16813. More recently, using the latest generation of labeling protocols and arrays,
  16814. it was determined that globin reduction was no longer necessary to obtain
  16815. sufficient sensitivity to detect differential transcript expression
  16816. \begin_inset CommandInset citation
  16817. LatexCommand cite
  16818. key "NuGEN2010"
  16819. literal "false"
  16820. \end_inset
  16821. .
  16822. However, we are not aware of any publications using these currently available
  16823. protocols with the latest generation of microarrays that actually compare
  16824. the detection sensitivity with and without globin reduction.
  16825. However, in practice this has now been adopted generally primarily driven
  16826. by concerns for cost control.
  16827. The main objective of our work was to directly test the impact of globin
  16828. gene transcripts and a new
  16829. \begin_inset Flex Glossary Term
  16830. status open
  16831. \begin_layout Plain Layout
  16832. GB
  16833. \end_layout
  16834. \end_inset
  16835. protocol for application to the newest generation of differential gene
  16836. expression profiling determined using next generation sequencing.
  16837. \end_layout
  16838. \begin_layout Standard
  16839. The challenge of doing global gene expression profiling in cynomolgus monkeys
  16840. is that the current available arrays were never designed to comprehensively
  16841. cover this genome and have not been updated since the first assemblies
  16842. of the cynomolgus genome were published.
  16843. Therefore, we determined that the best strategy for peripheral blood profiling
  16844. was to perform deep
  16845. \begin_inset Flex Glossary Term
  16846. status open
  16847. \begin_layout Plain Layout
  16848. RNA-seq
  16849. \end_layout
  16850. \end_inset
  16851. and inform the workflow using the latest available genome assembly and
  16852. annotation
  16853. \begin_inset CommandInset citation
  16854. LatexCommand cite
  16855. key "Wilson2013"
  16856. literal "false"
  16857. \end_inset
  16858. .
  16859. However, it was not immediately clear whether globin reduction was necessary
  16860. for
  16861. \begin_inset Flex Glossary Term
  16862. status open
  16863. \begin_layout Plain Layout
  16864. RNA-seq
  16865. \end_layout
  16866. \end_inset
  16867. or how much improvement in efficiency or sensitivity to detect differential
  16868. gene expression would be achieved for the added cost and effort.
  16869. \end_layout
  16870. \begin_layout Standard
  16871. Existing strategies for globin reduction involve degradation or physical
  16872. removal of globin transcripts in a separate step prior to reverse transcription
  16873. \begin_inset CommandInset citation
  16874. LatexCommand cite
  16875. key "Mastrokolias2012,Choi2014,Shin2014"
  16876. literal "false"
  16877. \end_inset
  16878. .
  16879. This additional step adds significant time, complexity, and cost to sample
  16880. preparation.
  16881. Faced with the need to perform
  16882. \begin_inset Flex Glossary Term
  16883. status open
  16884. \begin_layout Plain Layout
  16885. RNA-seq
  16886. \end_layout
  16887. \end_inset
  16888. on large numbers of blood samples we sought a solution to globin reduction
  16889. that could be achieved purely by adding additional reagents during the
  16890. reverse transcription reaction.
  16891. Furthermore, we needed a globin reduction method specific to cynomolgus
  16892. globin sequences that would work an organism for which no kit is available
  16893. off the shelf.
  16894. \end_layout
  16895. \begin_layout Standard
  16896. As mentioned above, the addition of
  16897. \begin_inset Flex Glossary Term
  16898. status open
  16899. \begin_layout Plain Layout
  16900. GB
  16901. \end_layout
  16902. \end_inset
  16903. \begin_inset Flex Glossary Term (pl)
  16904. status open
  16905. \begin_layout Plain Layout
  16906. oligo
  16907. \end_layout
  16908. \end_inset
  16909. has a very small impact on measured expression levels of gene expression.
  16910. However, this is a non-issue for the purposes of differential expression
  16911. testing, since a systematic change in a gene in all samples does not affect
  16912. relative expression levels between samples.
  16913. However, we must acknowledge that simple comparisons of gene expression
  16914. data obtained by
  16915. \begin_inset Flex Glossary Term
  16916. status open
  16917. \begin_layout Plain Layout
  16918. GB
  16919. \end_layout
  16920. \end_inset
  16921. and non-GB protocols are not possible without additional normalization.
  16922. \end_layout
  16923. \begin_layout Standard
  16924. More importantly,
  16925. \begin_inset Flex Glossary Term
  16926. status open
  16927. \begin_layout Plain Layout
  16928. GB
  16929. \end_layout
  16930. \end_inset
  16931. not only nearly doubles the yield of usable reads, it also increases inter-samp
  16932. le correlation and sensitivity to detect differential gene expression relative
  16933. to the same set of samples profiled without
  16934. \begin_inset Flex Glossary Term
  16935. status open
  16936. \begin_layout Plain Layout
  16937. GB
  16938. \end_layout
  16939. \end_inset
  16940. .
  16941. In addition,
  16942. \begin_inset Flex Glossary Term
  16943. status open
  16944. \begin_layout Plain Layout
  16945. GB
  16946. \end_layout
  16947. \end_inset
  16948. does not add a significant amount of random noise to the data.
  16949. \begin_inset Flex Glossary Term (Capital)
  16950. status open
  16951. \begin_layout Plain Layout
  16952. GB
  16953. \end_layout
  16954. \end_inset
  16955. thus represents a cost-effective and low-effort way to squeeze more data
  16956. and statistical power out of the same blood samples and the same amount
  16957. of sequencing.
  16958. In conclusion,
  16959. \begin_inset Flex Glossary Term
  16960. status open
  16961. \begin_layout Plain Layout
  16962. GB
  16963. \end_layout
  16964. \end_inset
  16965. greatly increases the yield of useful
  16966. \begin_inset Flex Glossary Term
  16967. status open
  16968. \begin_layout Plain Layout
  16969. RNA-seq
  16970. \end_layout
  16971. \end_inset
  16972. reads mapping to the rest of the genome, with minimal perturbations in
  16973. the relative levels of non-globin genes.
  16974. Based on these results, globin transcript reduction using sequence-specific,
  16975. complementary blocking
  16976. \begin_inset Flex Glossary Term (pl)
  16977. status open
  16978. \begin_layout Plain Layout
  16979. oligo
  16980. \end_layout
  16981. \end_inset
  16982. is recommended for all deep
  16983. \begin_inset Flex Glossary Term
  16984. status open
  16985. \begin_layout Plain Layout
  16986. RNA-seq
  16987. \end_layout
  16988. \end_inset
  16989. of cynomolgus and other nonhuman primate blood samples.
  16990. \end_layout
  16991. \begin_layout Section
  16992. Future Directions
  16993. \end_layout
  16994. \begin_layout Standard
  16995. One drawback of the
  16996. \begin_inset Flex Glossary Term
  16997. status open
  16998. \begin_layout Plain Layout
  16999. GB
  17000. \end_layout
  17001. \end_inset
  17002. method presented in this analysis is a poor yield of genic reads, only
  17003. around 50%.
  17004. In a separate experiment, the reagent mixture was modified so as to address
  17005. this drawback, resulting in a method that produces an even better reduction
  17006. in globin reads without reducing the overall fraction of genic reads.
  17007. However, the data showing this improvement consists of only a few test
  17008. samples, so the larger data set analyzed above was chosen in order to demonstra
  17009. te the effectiveness of the method in reducing globin reads while preserving
  17010. the biological signal.
  17011. \end_layout
  17012. \begin_layout Standard
  17013. The motivation for developing a fast practical way to enrich for non-globin
  17014. reads in cyno blood samples was to enable a large-scale
  17015. \begin_inset Flex Glossary Term
  17016. status open
  17017. \begin_layout Plain Layout
  17018. RNA-seq
  17019. \end_layout
  17020. \end_inset
  17021. experiment investigating the effects of mesenchymal stem cell infusion
  17022. on blood gene expression in cynomologus transplant recipients in a time
  17023. course after transplantation.
  17024. With the
  17025. \begin_inset Flex Glossary Term
  17026. status open
  17027. \begin_layout Plain Layout
  17028. GB
  17029. \end_layout
  17030. \end_inset
  17031. method in place, the way is now clear for this experiment to proceed.
  17032. \end_layout
  17033. \begin_layout Standard
  17034. \begin_inset Note Note
  17035. status open
  17036. \begin_layout Chapter*
  17037. Future Directions
  17038. \end_layout
  17039. \begin_layout Plain Layout
  17040. \begin_inset ERT
  17041. status collapsed
  17042. \begin_layout Plain Layout
  17043. \backslash
  17044. glsresetall
  17045. \end_layout
  17046. \end_inset
  17047. \begin_inset Note Note
  17048. status collapsed
  17049. \begin_layout Plain Layout
  17050. Reintroduce all abbreviations
  17051. \end_layout
  17052. \end_inset
  17053. \end_layout
  17054. \begin_layout Plain Layout
  17055. \begin_inset Flex TODO Note (inline)
  17056. status open
  17057. \begin_layout Plain Layout
  17058. If there are any chapter-independent future directions, put them here.
  17059. Otherwise, delete this section.
  17060. \end_layout
  17061. \end_inset
  17062. \end_layout
  17063. \end_inset
  17064. \end_layout
  17065. \begin_layout Chapter
  17066. Closing remarks
  17067. \end_layout
  17068. \begin_layout Standard
  17069. \align center
  17070. \begin_inset ERT
  17071. status open
  17072. \begin_layout Plain Layout
  17073. \backslash
  17074. addcontentsline{toc}{chapter}{Test}
  17075. \end_layout
  17076. \end_inset
  17077. \end_layout
  17078. \begin_layout Standard
  17079. \begin_inset ERT
  17080. status collapsed
  17081. \begin_layout Plain Layout
  17082. \backslash
  17083. glsresetall
  17084. \end_layout
  17085. \end_inset
  17086. \begin_inset Note Note
  17087. status collapsed
  17088. \begin_layout Plain Layout
  17089. Reintroduce all abbreviations
  17090. \end_layout
  17091. \end_inset
  17092. \end_layout
  17093. \begin_layout Standard
  17094. \align center
  17095. \begin_inset ERT
  17096. status collapsed
  17097. \begin_layout Plain Layout
  17098. % Use "References" as the title of the Bibliography
  17099. \end_layout
  17100. \begin_layout Plain Layout
  17101. \backslash
  17102. renewcommand{
  17103. \backslash
  17104. bibname}{References}
  17105. \end_layout
  17106. \end_inset
  17107. \end_layout
  17108. \begin_layout Standard
  17109. \begin_inset CommandInset bibtex
  17110. LatexCommand bibtex
  17111. btprint "btPrintCited"
  17112. bibfiles "code-refs,refs-PROCESSED"
  17113. options "bibtotoc"
  17114. \end_inset
  17115. \end_layout
  17116. \begin_layout Standard
  17117. \begin_inset Flex TODO Note (inline)
  17118. status open
  17119. \begin_layout Plain Layout
  17120. Reference URLs that span pages have clickable links that include the page
  17121. numbers and watermark.
  17122. Try to fix that.
  17123. \end_layout
  17124. \end_inset
  17125. \end_layout
  17126. \end_body
  17127. \end_document