thesis.lyx 396 KB

1234567891011121314151617181920212223242526272829303132333435363738394041424344454647484950515253545556575859606162636465666768697071727374757677787980818283848586878889909192939495969798991001011021031041051061071081091101111121131141151161171181191201211221231241251261271281291301311321331341351361371381391401411421431441451461471481491501511521531541551561571581591601611621631641651661671681691701711721731741751761771781791801811821831841851861871881891901911921931941951961971981992002012022032042052062072082092102112122132142152162172182192202212222232242252262272282292302312322332342352362372382392402412422432442452462472482492502512522532542552562572582592602612622632642652662672682692702712722732742752762772782792802812822832842852862872882892902912922932942952962972982993003013023033043053063073083093103113123133143153163173183193203213223233243253263273283293303313323333343353363373383393403413423433443453463473483493503513523533543553563573583593603613623633643653663673683693703713723733743753763773783793803813823833843853863873883893903913923933943953963973983994004014024034044054064074084094104114124134144154164174184194204214224234244254264274284294304314324334344354364374384394404414424434444454464474484494504514524534544554564574584594604614624634644654664674684694704714724734744754764774784794804814824834844854864874884894904914924934944954964974984995005015025035045055065075085095105115125135145155165175185195205215225235245255265275285295305315325335345355365375385395405415425435445455465475485495505515525535545555565575585595605615625635645655665675685695705715725735745755765775785795805815825835845855865875885895905915925935945955965975985996006016026036046056066076086096106116126136146156166176186196206216226236246256266276286296306316326336346356366376386396406416426436446456466476486496506516526536546556566576586596606616626636646656666676686696706716726736746756766776786796806816826836846856866876886896906916926936946956966976986997007017027037047057067077087097107117127137147157167177187197207217227237247257267277287297307317327337347357367377387397407417427437447457467477487497507517527537547557567577587597607617627637647657667677687697707717727737747757767777787797807817827837847857867877887897907917927937947957967977987998008018028038048058068078088098108118128138148158168178188198208218228238248258268278288298308318328338348358368378388398408418428438448458468478488498508518528538548558568578588598608618628638648658668678688698708718728738748758768778788798808818828838848858868878888898908918928938948958968978988999009019029039049059069079089099109119129139149159169179189199209219229239249259269279289299309319329339349359369379389399409419429439449459469479489499509519529539549559569579589599609619629639649659669679689699709719729739749759769779789799809819829839849859869879889899909919929939949959969979989991000100110021003100410051006100710081009101010111012101310141015101610171018101910201021102210231024102510261027102810291030103110321033103410351036103710381039104010411042104310441045104610471048104910501051105210531054105510561057105810591060106110621063106410651066106710681069107010711072107310741075107610771078107910801081108210831084108510861087108810891090109110921093109410951096109710981099110011011102110311041105110611071108110911101111111211131114111511161117111811191120112111221123112411251126112711281129113011311132113311341135113611371138113911401141114211431144114511461147114811491150115111521153115411551156115711581159116011611162116311641165116611671168116911701171117211731174117511761177117811791180118111821183118411851186118711881189119011911192119311941195119611971198119912001201120212031204120512061207120812091210121112121213121412151216121712181219122012211222122312241225122612271228122912301231123212331234123512361237123812391240124112421243124412451246124712481249125012511252125312541255125612571258125912601261126212631264126512661267126812691270127112721273127412751276127712781279128012811282128312841285128612871288128912901291129212931294129512961297129812991300130113021303130413051306130713081309131013111312131313141315131613171318131913201321132213231324132513261327132813291330133113321333133413351336133713381339134013411342134313441345134613471348134913501351135213531354135513561357135813591360136113621363136413651366136713681369137013711372137313741375137613771378137913801381138213831384138513861387138813891390139113921393139413951396139713981399140014011402140314041405140614071408140914101411141214131414141514161417141814191420142114221423142414251426142714281429143014311432143314341435143614371438143914401441144214431444144514461447144814491450145114521453145414551456145714581459146014611462146314641465146614671468146914701471147214731474147514761477147814791480148114821483148414851486148714881489149014911492149314941495149614971498149915001501150215031504150515061507150815091510151115121513151415151516151715181519152015211522152315241525152615271528152915301531153215331534153515361537153815391540154115421543154415451546154715481549155015511552155315541555155615571558155915601561156215631564156515661567156815691570157115721573157415751576157715781579158015811582158315841585158615871588158915901591159215931594159515961597159815991600160116021603160416051606160716081609161016111612161316141615161616171618161916201621162216231624162516261627162816291630163116321633163416351636163716381639164016411642164316441645164616471648164916501651165216531654165516561657165816591660166116621663166416651666166716681669167016711672167316741675167616771678167916801681168216831684168516861687168816891690169116921693169416951696169716981699170017011702170317041705170617071708170917101711171217131714171517161717171817191720172117221723172417251726172717281729173017311732173317341735173617371738173917401741174217431744174517461747174817491750175117521753175417551756175717581759176017611762176317641765176617671768176917701771177217731774177517761777177817791780178117821783178417851786178717881789179017911792179317941795179617971798179918001801180218031804180518061807180818091810181118121813181418151816181718181819182018211822182318241825182618271828182918301831183218331834183518361837183818391840184118421843184418451846184718481849185018511852185318541855185618571858185918601861186218631864186518661867186818691870187118721873187418751876187718781879188018811882188318841885188618871888188918901891189218931894189518961897189818991900190119021903190419051906190719081909191019111912191319141915191619171918191919201921192219231924192519261927192819291930193119321933193419351936193719381939194019411942194319441945194619471948194919501951195219531954195519561957195819591960196119621963196419651966196719681969197019711972197319741975197619771978197919801981198219831984198519861987198819891990199119921993199419951996199719981999200020012002200320042005200620072008200920102011201220132014201520162017201820192020202120222023202420252026202720282029203020312032203320342035203620372038203920402041204220432044204520462047204820492050205120522053205420552056205720582059206020612062206320642065206620672068206920702071207220732074207520762077207820792080208120822083208420852086208720882089209020912092209320942095209620972098209921002101210221032104210521062107210821092110211121122113211421152116211721182119212021212122212321242125212621272128212921302131213221332134213521362137213821392140214121422143214421452146214721482149215021512152215321542155215621572158215921602161216221632164216521662167216821692170217121722173217421752176217721782179218021812182218321842185218621872188218921902191219221932194219521962197219821992200220122022203220422052206220722082209221022112212221322142215221622172218221922202221222222232224222522262227222822292230223122322233223422352236223722382239224022412242224322442245224622472248224922502251225222532254225522562257225822592260226122622263226422652266226722682269227022712272227322742275227622772278227922802281228222832284228522862287228822892290229122922293229422952296229722982299230023012302230323042305230623072308230923102311231223132314231523162317231823192320232123222323232423252326232723282329233023312332233323342335233623372338233923402341234223432344234523462347234823492350235123522353235423552356235723582359236023612362236323642365236623672368236923702371237223732374237523762377237823792380238123822383238423852386238723882389239023912392239323942395239623972398239924002401240224032404240524062407240824092410241124122413241424152416241724182419242024212422242324242425242624272428242924302431243224332434243524362437243824392440244124422443244424452446244724482449245024512452245324542455245624572458245924602461246224632464246524662467246824692470247124722473247424752476247724782479248024812482248324842485248624872488248924902491249224932494249524962497249824992500250125022503250425052506250725082509251025112512251325142515251625172518251925202521252225232524252525262527252825292530253125322533253425352536253725382539254025412542254325442545254625472548254925502551255225532554255525562557255825592560256125622563256425652566256725682569257025712572257325742575257625772578257925802581258225832584258525862587258825892590259125922593259425952596259725982599260026012602260326042605260626072608260926102611261226132614261526162617261826192620262126222623262426252626262726282629263026312632263326342635263626372638263926402641264226432644264526462647264826492650265126522653265426552656265726582659266026612662266326642665266626672668266926702671267226732674267526762677267826792680268126822683268426852686268726882689269026912692269326942695269626972698269927002701270227032704270527062707270827092710271127122713271427152716271727182719272027212722272327242725272627272728272927302731273227332734273527362737273827392740274127422743274427452746274727482749275027512752275327542755275627572758275927602761276227632764276527662767276827692770277127722773277427752776277727782779278027812782278327842785278627872788278927902791279227932794279527962797279827992800280128022803280428052806280728082809281028112812281328142815281628172818281928202821282228232824282528262827282828292830283128322833283428352836283728382839284028412842284328442845284628472848284928502851285228532854285528562857285828592860286128622863286428652866286728682869287028712872287328742875287628772878287928802881288228832884288528862887288828892890289128922893289428952896289728982899290029012902290329042905290629072908290929102911291229132914291529162917291829192920292129222923292429252926292729282929293029312932293329342935293629372938293929402941294229432944294529462947294829492950295129522953295429552956295729582959296029612962296329642965296629672968296929702971297229732974297529762977297829792980298129822983298429852986298729882989299029912992299329942995299629972998299930003001300230033004300530063007300830093010301130123013301430153016301730183019302030213022302330243025302630273028302930303031303230333034303530363037303830393040304130423043304430453046304730483049305030513052305330543055305630573058305930603061306230633064306530663067306830693070307130723073307430753076307730783079308030813082308330843085308630873088308930903091309230933094309530963097309830993100310131023103310431053106310731083109311031113112311331143115311631173118311931203121312231233124312531263127312831293130313131323133313431353136313731383139314031413142314331443145314631473148314931503151315231533154315531563157315831593160316131623163316431653166316731683169317031713172317331743175317631773178317931803181318231833184318531863187318831893190319131923193319431953196319731983199320032013202320332043205320632073208320932103211321232133214321532163217321832193220322132223223322432253226322732283229323032313232323332343235323632373238323932403241324232433244324532463247324832493250325132523253325432553256325732583259326032613262326332643265326632673268326932703271327232733274327532763277327832793280328132823283328432853286328732883289329032913292329332943295329632973298329933003301330233033304330533063307330833093310331133123313331433153316331733183319332033213322332333243325332633273328332933303331333233333334333533363337333833393340334133423343334433453346334733483349335033513352335333543355335633573358335933603361336233633364336533663367336833693370337133723373337433753376337733783379338033813382338333843385338633873388338933903391339233933394339533963397339833993400340134023403340434053406340734083409341034113412341334143415341634173418341934203421342234233424342534263427342834293430343134323433343434353436343734383439344034413442344334443445344634473448344934503451345234533454345534563457345834593460346134623463346434653466346734683469347034713472347334743475347634773478347934803481348234833484348534863487348834893490349134923493349434953496349734983499350035013502350335043505350635073508350935103511351235133514351535163517351835193520352135223523352435253526352735283529353035313532353335343535353635373538353935403541354235433544354535463547354835493550355135523553355435553556355735583559356035613562356335643565356635673568356935703571357235733574357535763577357835793580358135823583358435853586358735883589359035913592359335943595359635973598359936003601360236033604360536063607360836093610361136123613361436153616361736183619362036213622362336243625362636273628362936303631363236333634363536363637363836393640364136423643364436453646364736483649365036513652365336543655365636573658365936603661366236633664366536663667366836693670367136723673367436753676367736783679368036813682368336843685368636873688368936903691369236933694369536963697369836993700370137023703370437053706370737083709371037113712371337143715371637173718371937203721372237233724372537263727372837293730373137323733373437353736373737383739374037413742374337443745374637473748374937503751375237533754375537563757375837593760376137623763376437653766376737683769377037713772377337743775377637773778377937803781378237833784378537863787378837893790379137923793379437953796379737983799380038013802380338043805380638073808380938103811381238133814381538163817381838193820382138223823382438253826382738283829383038313832383338343835383638373838383938403841384238433844384538463847384838493850385138523853385438553856385738583859386038613862386338643865386638673868386938703871387238733874387538763877387838793880388138823883388438853886388738883889389038913892389338943895389638973898389939003901390239033904390539063907390839093910391139123913391439153916391739183919392039213922392339243925392639273928392939303931393239333934393539363937393839393940394139423943394439453946394739483949395039513952395339543955395639573958395939603961396239633964396539663967396839693970397139723973397439753976397739783979398039813982398339843985398639873988398939903991399239933994399539963997399839994000400140024003400440054006400740084009401040114012401340144015401640174018401940204021402240234024402540264027402840294030403140324033403440354036403740384039404040414042404340444045404640474048404940504051405240534054405540564057405840594060406140624063406440654066406740684069407040714072407340744075407640774078407940804081408240834084408540864087408840894090409140924093409440954096409740984099410041014102410341044105410641074108410941104111411241134114411541164117411841194120412141224123412441254126412741284129413041314132413341344135413641374138413941404141414241434144414541464147414841494150415141524153415441554156415741584159416041614162416341644165416641674168416941704171417241734174417541764177417841794180418141824183418441854186418741884189419041914192419341944195419641974198419942004201420242034204420542064207420842094210421142124213421442154216421742184219422042214222422342244225422642274228422942304231423242334234423542364237423842394240424142424243424442454246424742484249425042514252425342544255425642574258425942604261426242634264426542664267426842694270427142724273427442754276427742784279428042814282428342844285428642874288428942904291429242934294429542964297429842994300430143024303430443054306430743084309431043114312431343144315431643174318431943204321432243234324432543264327432843294330433143324333433443354336433743384339434043414342434343444345434643474348434943504351435243534354435543564357435843594360436143624363436443654366436743684369437043714372437343744375437643774378437943804381438243834384438543864387438843894390439143924393439443954396439743984399440044014402440344044405440644074408440944104411441244134414441544164417441844194420442144224423442444254426442744284429443044314432443344344435443644374438443944404441444244434444444544464447444844494450445144524453445444554456445744584459446044614462446344644465446644674468446944704471447244734474447544764477447844794480448144824483448444854486448744884489449044914492449344944495449644974498449945004501450245034504450545064507450845094510451145124513451445154516451745184519452045214522452345244525452645274528452945304531453245334534453545364537453845394540454145424543454445454546454745484549455045514552455345544555455645574558455945604561456245634564456545664567456845694570457145724573457445754576457745784579458045814582458345844585458645874588458945904591459245934594459545964597459845994600460146024603460446054606460746084609461046114612461346144615461646174618461946204621462246234624462546264627462846294630463146324633463446354636463746384639464046414642464346444645464646474648464946504651465246534654465546564657465846594660466146624663466446654666466746684669467046714672467346744675467646774678467946804681468246834684468546864687468846894690469146924693469446954696469746984699470047014702470347044705470647074708470947104711471247134714471547164717471847194720472147224723472447254726472747284729473047314732473347344735473647374738473947404741474247434744474547464747474847494750475147524753475447554756475747584759476047614762476347644765476647674768476947704771477247734774477547764777477847794780478147824783478447854786478747884789479047914792479347944795479647974798479948004801480248034804480548064807480848094810481148124813481448154816481748184819482048214822482348244825482648274828482948304831483248334834483548364837483848394840484148424843484448454846484748484849485048514852485348544855485648574858485948604861486248634864486548664867486848694870487148724873487448754876487748784879488048814882488348844885488648874888488948904891489248934894489548964897489848994900490149024903490449054906490749084909491049114912491349144915491649174918491949204921492249234924492549264927492849294930493149324933493449354936493749384939494049414942494349444945494649474948494949504951495249534954495549564957495849594960496149624963496449654966496749684969497049714972497349744975497649774978497949804981498249834984498549864987498849894990499149924993499449954996499749984999500050015002500350045005500650075008500950105011501250135014501550165017501850195020502150225023502450255026502750285029503050315032503350345035503650375038503950405041504250435044504550465047504850495050505150525053505450555056505750585059506050615062506350645065506650675068506950705071507250735074507550765077507850795080508150825083508450855086508750885089509050915092509350945095509650975098509951005101510251035104510551065107510851095110511151125113511451155116511751185119512051215122512351245125512651275128512951305131513251335134513551365137513851395140514151425143514451455146514751485149515051515152515351545155515651575158515951605161516251635164516551665167516851695170517151725173517451755176517751785179518051815182518351845185518651875188518951905191519251935194519551965197519851995200520152025203520452055206520752085209521052115212521352145215521652175218521952205221522252235224522552265227522852295230523152325233523452355236523752385239524052415242524352445245524652475248524952505251525252535254525552565257525852595260526152625263526452655266526752685269527052715272527352745275527652775278527952805281528252835284528552865287528852895290529152925293529452955296529752985299530053015302530353045305530653075308530953105311531253135314531553165317531853195320532153225323532453255326532753285329533053315332533353345335533653375338533953405341534253435344534553465347534853495350535153525353535453555356535753585359536053615362536353645365536653675368536953705371537253735374537553765377537853795380538153825383538453855386538753885389539053915392539353945395539653975398539954005401540254035404540554065407540854095410541154125413541454155416541754185419542054215422542354245425542654275428542954305431543254335434543554365437543854395440544154425443544454455446544754485449545054515452545354545455545654575458545954605461546254635464546554665467546854695470547154725473547454755476547754785479548054815482548354845485548654875488548954905491549254935494549554965497549854995500550155025503550455055506550755085509551055115512551355145515551655175518551955205521552255235524552555265527552855295530553155325533553455355536553755385539554055415542554355445545554655475548554955505551555255535554555555565557555855595560556155625563556455655566556755685569557055715572557355745575557655775578557955805581558255835584558555865587558855895590559155925593559455955596559755985599560056015602560356045605560656075608560956105611561256135614561556165617561856195620562156225623562456255626562756285629563056315632563356345635563656375638563956405641564256435644564556465647564856495650565156525653565456555656565756585659566056615662566356645665566656675668566956705671567256735674567556765677567856795680568156825683568456855686568756885689569056915692569356945695569656975698569957005701570257035704570557065707570857095710571157125713571457155716571757185719572057215722572357245725572657275728572957305731573257335734573557365737573857395740574157425743574457455746574757485749575057515752575357545755575657575758575957605761576257635764576557665767576857695770577157725773577457755776577757785779578057815782578357845785578657875788578957905791579257935794579557965797579857995800580158025803580458055806580758085809581058115812581358145815581658175818581958205821582258235824582558265827582858295830583158325833583458355836583758385839584058415842584358445845584658475848584958505851585258535854585558565857585858595860586158625863586458655866586758685869587058715872587358745875587658775878587958805881588258835884588558865887588858895890589158925893589458955896589758985899590059015902590359045905590659075908590959105911591259135914591559165917591859195920592159225923592459255926592759285929593059315932593359345935593659375938593959405941594259435944594559465947594859495950595159525953595459555956595759585959596059615962596359645965596659675968596959705971597259735974597559765977597859795980598159825983598459855986598759885989599059915992599359945995599659975998599960006001600260036004600560066007600860096010601160126013601460156016601760186019602060216022602360246025602660276028602960306031603260336034603560366037603860396040604160426043604460456046604760486049605060516052605360546055605660576058605960606061606260636064606560666067606860696070607160726073607460756076607760786079608060816082608360846085608660876088608960906091609260936094609560966097609860996100610161026103610461056106610761086109611061116112611361146115611661176118611961206121612261236124612561266127612861296130613161326133613461356136613761386139614061416142614361446145614661476148614961506151615261536154615561566157615861596160616161626163616461656166616761686169617061716172617361746175617661776178617961806181618261836184618561866187618861896190619161926193619461956196619761986199620062016202620362046205620662076208620962106211621262136214621562166217621862196220622162226223622462256226622762286229623062316232623362346235623662376238623962406241624262436244624562466247624862496250625162526253625462556256625762586259626062616262626362646265626662676268626962706271627262736274627562766277627862796280628162826283628462856286628762886289629062916292629362946295629662976298629963006301630263036304630563066307630863096310631163126313631463156316631763186319632063216322632363246325632663276328632963306331633263336334633563366337633863396340634163426343634463456346634763486349635063516352635363546355635663576358635963606361636263636364636563666367636863696370637163726373637463756376637763786379638063816382638363846385638663876388638963906391639263936394639563966397639863996400640164026403640464056406640764086409641064116412641364146415641664176418641964206421642264236424642564266427642864296430643164326433643464356436643764386439644064416442644364446445644664476448644964506451645264536454645564566457645864596460646164626463646464656466646764686469647064716472647364746475647664776478647964806481648264836484648564866487648864896490649164926493649464956496649764986499650065016502650365046505650665076508650965106511651265136514651565166517651865196520652165226523652465256526652765286529653065316532653365346535653665376538653965406541654265436544654565466547654865496550655165526553655465556556655765586559656065616562656365646565656665676568656965706571657265736574657565766577657865796580658165826583658465856586658765886589659065916592659365946595659665976598659966006601660266036604660566066607660866096610661166126613661466156616661766186619662066216622662366246625662666276628662966306631663266336634663566366637663866396640664166426643664466456646664766486649665066516652665366546655665666576658665966606661666266636664666566666667666866696670667166726673667466756676667766786679668066816682668366846685668666876688668966906691669266936694669566966697669866996700670167026703670467056706670767086709671067116712671367146715671667176718671967206721672267236724672567266727672867296730673167326733673467356736673767386739674067416742674367446745674667476748674967506751675267536754675567566757675867596760676167626763676467656766676767686769677067716772677367746775677667776778677967806781678267836784678567866787678867896790679167926793679467956796679767986799680068016802680368046805680668076808680968106811681268136814681568166817681868196820682168226823682468256826682768286829683068316832683368346835683668376838683968406841684268436844684568466847684868496850685168526853685468556856685768586859686068616862686368646865686668676868686968706871687268736874687568766877687868796880688168826883688468856886688768886889689068916892689368946895689668976898689969006901690269036904690569066907690869096910691169126913691469156916691769186919692069216922692369246925692669276928692969306931693269336934693569366937693869396940694169426943694469456946694769486949695069516952695369546955695669576958695969606961696269636964696569666967696869696970697169726973697469756976697769786979698069816982698369846985698669876988698969906991699269936994699569966997699869997000700170027003700470057006700770087009701070117012701370147015701670177018701970207021702270237024702570267027702870297030703170327033703470357036703770387039704070417042704370447045704670477048704970507051705270537054705570567057705870597060706170627063706470657066706770687069707070717072707370747075707670777078707970807081708270837084708570867087708870897090709170927093709470957096709770987099710071017102710371047105710671077108710971107111711271137114711571167117711871197120712171227123712471257126712771287129713071317132713371347135713671377138713971407141714271437144714571467147714871497150715171527153715471557156715771587159716071617162716371647165716671677168716971707171717271737174717571767177717871797180718171827183718471857186718771887189719071917192719371947195719671977198719972007201720272037204720572067207720872097210721172127213721472157216721772187219722072217222722372247225722672277228722972307231723272337234723572367237723872397240724172427243724472457246724772487249725072517252725372547255725672577258725972607261726272637264726572667267726872697270727172727273727472757276727772787279728072817282728372847285728672877288728972907291729272937294729572967297729872997300730173027303730473057306730773087309731073117312731373147315731673177318731973207321732273237324732573267327732873297330733173327333733473357336733773387339734073417342734373447345734673477348734973507351735273537354735573567357735873597360736173627363736473657366736773687369737073717372737373747375737673777378737973807381738273837384738573867387738873897390739173927393739473957396739773987399740074017402740374047405740674077408740974107411741274137414741574167417741874197420742174227423742474257426742774287429743074317432743374347435743674377438743974407441744274437444744574467447744874497450745174527453745474557456745774587459746074617462746374647465746674677468746974707471747274737474747574767477747874797480748174827483748474857486748774887489749074917492749374947495749674977498749975007501750275037504750575067507750875097510751175127513751475157516751775187519752075217522752375247525752675277528752975307531753275337534753575367537753875397540754175427543754475457546754775487549755075517552755375547555755675577558755975607561756275637564756575667567756875697570757175727573757475757576757775787579758075817582758375847585758675877588758975907591759275937594759575967597759875997600760176027603760476057606760776087609761076117612761376147615761676177618761976207621762276237624762576267627762876297630763176327633763476357636763776387639764076417642764376447645764676477648764976507651765276537654765576567657765876597660766176627663766476657666766776687669767076717672767376747675767676777678767976807681768276837684768576867687768876897690769176927693769476957696769776987699770077017702770377047705770677077708770977107711771277137714771577167717771877197720772177227723772477257726772777287729773077317732773377347735773677377738773977407741774277437744774577467747774877497750775177527753775477557756775777587759776077617762776377647765776677677768776977707771777277737774777577767777777877797780778177827783778477857786778777887789779077917792779377947795779677977798779978007801780278037804780578067807780878097810781178127813781478157816781778187819782078217822782378247825782678277828782978307831783278337834783578367837783878397840784178427843784478457846784778487849785078517852785378547855785678577858785978607861786278637864786578667867786878697870787178727873787478757876787778787879788078817882788378847885788678877888788978907891789278937894789578967897789878997900790179027903790479057906790779087909791079117912791379147915791679177918791979207921792279237924792579267927792879297930793179327933793479357936793779387939794079417942794379447945794679477948794979507951795279537954795579567957795879597960796179627963796479657966796779687969797079717972797379747975797679777978797979807981798279837984798579867987798879897990799179927993799479957996799779987999800080018002800380048005800680078008800980108011801280138014801580168017801880198020802180228023802480258026802780288029803080318032803380348035803680378038803980408041804280438044804580468047804880498050805180528053805480558056805780588059806080618062806380648065806680678068806980708071807280738074807580768077807880798080808180828083808480858086808780888089809080918092809380948095809680978098809981008101810281038104810581068107810881098110811181128113811481158116811781188119812081218122812381248125812681278128812981308131813281338134813581368137813881398140814181428143814481458146814781488149815081518152815381548155815681578158815981608161816281638164816581668167816881698170817181728173817481758176817781788179818081818182818381848185818681878188818981908191819281938194819581968197819881998200820182028203820482058206820782088209821082118212821382148215821682178218821982208221822282238224822582268227822882298230823182328233823482358236823782388239824082418242824382448245824682478248824982508251825282538254825582568257825882598260826182628263826482658266826782688269827082718272827382748275827682778278827982808281828282838284828582868287828882898290829182928293829482958296829782988299830083018302830383048305830683078308830983108311831283138314831583168317831883198320832183228323832483258326832783288329833083318332833383348335833683378338833983408341834283438344834583468347834883498350835183528353835483558356835783588359836083618362836383648365836683678368836983708371837283738374837583768377837883798380838183828383838483858386838783888389839083918392839383948395839683978398839984008401840284038404840584068407840884098410841184128413841484158416841784188419842084218422842384248425842684278428842984308431843284338434843584368437843884398440844184428443844484458446844784488449845084518452845384548455845684578458845984608461846284638464846584668467846884698470847184728473847484758476847784788479848084818482848384848485848684878488848984908491849284938494849584968497849884998500850185028503850485058506850785088509851085118512851385148515851685178518851985208521852285238524852585268527852885298530853185328533853485358536853785388539854085418542854385448545854685478548854985508551855285538554855585568557855885598560856185628563856485658566856785688569857085718572857385748575857685778578857985808581858285838584858585868587858885898590859185928593859485958596859785988599860086018602860386048605860686078608860986108611861286138614861586168617861886198620862186228623862486258626862786288629863086318632863386348635863686378638863986408641864286438644864586468647864886498650865186528653865486558656865786588659866086618662866386648665866686678668866986708671867286738674867586768677867886798680868186828683868486858686868786888689869086918692869386948695869686978698869987008701870287038704870587068707870887098710871187128713871487158716871787188719872087218722872387248725872687278728872987308731873287338734873587368737873887398740874187428743874487458746874787488749875087518752875387548755875687578758875987608761876287638764876587668767876887698770877187728773877487758776877787788779878087818782878387848785878687878788878987908791879287938794879587968797879887998800880188028803880488058806880788088809881088118812881388148815881688178818881988208821882288238824882588268827882888298830883188328833883488358836883788388839884088418842884388448845884688478848884988508851885288538854885588568857885888598860886188628863886488658866886788688869887088718872887388748875887688778878887988808881888288838884888588868887888888898890889188928893889488958896889788988899890089018902890389048905890689078908890989108911891289138914891589168917891889198920892189228923892489258926892789288929893089318932893389348935893689378938893989408941894289438944894589468947894889498950895189528953895489558956895789588959896089618962896389648965896689678968896989708971897289738974897589768977897889798980898189828983898489858986898789888989899089918992899389948995899689978998899990009001900290039004900590069007900890099010901190129013901490159016901790189019902090219022902390249025902690279028902990309031903290339034903590369037903890399040904190429043904490459046904790489049905090519052905390549055905690579058905990609061906290639064906590669067906890699070907190729073907490759076907790789079908090819082908390849085908690879088908990909091909290939094909590969097909890999100910191029103910491059106910791089109911091119112911391149115911691179118911991209121912291239124912591269127912891299130913191329133913491359136913791389139914091419142914391449145914691479148914991509151915291539154915591569157915891599160916191629163916491659166916791689169917091719172917391749175917691779178917991809181918291839184918591869187918891899190919191929193919491959196919791989199920092019202920392049205920692079208920992109211921292139214921592169217921892199220922192229223922492259226922792289229923092319232923392349235923692379238923992409241924292439244924592469247924892499250925192529253925492559256925792589259926092619262926392649265926692679268926992709271927292739274927592769277927892799280928192829283928492859286928792889289929092919292929392949295929692979298929993009301930293039304930593069307930893099310931193129313931493159316931793189319932093219322932393249325932693279328932993309331933293339334933593369337933893399340934193429343934493459346934793489349935093519352935393549355935693579358935993609361936293639364936593669367936893699370937193729373937493759376937793789379938093819382938393849385938693879388938993909391939293939394939593969397939893999400940194029403940494059406940794089409941094119412941394149415941694179418941994209421942294239424942594269427942894299430943194329433943494359436943794389439944094419442944394449445944694479448944994509451945294539454945594569457945894599460946194629463946494659466946794689469947094719472947394749475947694779478947994809481948294839484948594869487948894899490949194929493949494959496949794989499950095019502950395049505950695079508950995109511951295139514951595169517951895199520952195229523952495259526952795289529953095319532953395349535953695379538953995409541954295439544954595469547954895499550955195529553955495559556955795589559956095619562956395649565956695679568956995709571957295739574957595769577957895799580958195829583958495859586958795889589959095919592959395949595959695979598959996009601960296039604960596069607960896099610961196129613961496159616961796189619962096219622962396249625962696279628962996309631963296339634963596369637963896399640964196429643964496459646964796489649965096519652965396549655965696579658965996609661966296639664966596669667966896699670967196729673967496759676967796789679968096819682968396849685968696879688968996909691969296939694969596969697969896999700970197029703970497059706970797089709971097119712971397149715971697179718971997209721972297239724972597269727972897299730973197329733973497359736973797389739974097419742974397449745974697479748974997509751975297539754975597569757975897599760976197629763976497659766976797689769977097719772977397749775977697779778977997809781978297839784978597869787978897899790979197929793979497959796979797989799980098019802980398049805980698079808980998109811981298139814981598169817981898199820982198229823982498259826982798289829983098319832983398349835983698379838983998409841984298439844984598469847984898499850985198529853985498559856985798589859986098619862986398649865986698679868986998709871987298739874987598769877987898799880988198829883988498859886988798889889989098919892989398949895989698979898989999009901990299039904990599069907990899099910991199129913991499159916991799189919992099219922992399249925992699279928992999309931993299339934993599369937993899399940994199429943994499459946994799489949995099519952995399549955995699579958995999609961996299639964996599669967996899699970997199729973997499759976997799789979998099819982998399849985998699879988998999909991999299939994999599969997999899991000010001100021000310004100051000610007100081000910010100111001210013100141001510016100171001810019100201002110022100231002410025100261002710028100291003010031100321003310034100351003610037100381003910040100411004210043100441004510046100471004810049100501005110052100531005410055100561005710058100591006010061100621006310064100651006610067100681006910070100711007210073100741007510076100771007810079100801008110082100831008410085100861008710088100891009010091100921009310094100951009610097100981009910100101011010210103101041010510106101071010810109101101011110112101131011410115101161011710118101191012010121101221012310124101251012610127101281012910130101311013210133101341013510136101371013810139101401014110142101431014410145101461014710148101491015010151101521015310154101551015610157101581015910160101611016210163101641016510166101671016810169101701017110172101731017410175101761017710178101791018010181101821018310184101851018610187101881018910190101911019210193101941019510196101971019810199102001020110202102031020410205102061020710208102091021010211102121021310214102151021610217102181021910220102211022210223102241022510226102271022810229102301023110232102331023410235102361023710238102391024010241102421024310244102451024610247102481024910250102511025210253102541025510256102571025810259102601026110262102631026410265102661026710268102691027010271102721027310274102751027610277102781027910280102811028210283102841028510286102871028810289102901029110292102931029410295102961029710298102991030010301103021030310304103051030610307103081030910310103111031210313103141031510316103171031810319103201032110322103231032410325103261032710328103291033010331103321033310334103351033610337103381033910340103411034210343103441034510346103471034810349103501035110352103531035410355103561035710358103591036010361103621036310364103651036610367103681036910370103711037210373103741037510376103771037810379103801038110382103831038410385103861038710388103891039010391103921039310394103951039610397103981039910400104011040210403104041040510406104071040810409104101041110412104131041410415104161041710418104191042010421104221042310424104251042610427104281042910430104311043210433104341043510436104371043810439104401044110442104431044410445104461044710448104491045010451104521045310454104551045610457104581045910460104611046210463104641046510466104671046810469104701047110472104731047410475104761047710478104791048010481104821048310484104851048610487104881048910490104911049210493104941049510496104971049810499105001050110502105031050410505105061050710508105091051010511105121051310514105151051610517105181051910520105211052210523105241052510526105271052810529105301053110532105331053410535105361053710538105391054010541105421054310544105451054610547105481054910550105511055210553105541055510556105571055810559105601056110562105631056410565105661056710568105691057010571105721057310574105751057610577105781057910580105811058210583105841058510586105871058810589105901059110592105931059410595105961059710598105991060010601106021060310604106051060610607106081060910610106111061210613106141061510616106171061810619106201062110622106231062410625106261062710628106291063010631106321063310634106351063610637106381063910640106411064210643106441064510646106471064810649106501065110652106531065410655106561065710658106591066010661106621066310664106651066610667106681066910670106711067210673106741067510676106771067810679106801068110682106831068410685106861068710688106891069010691106921069310694106951069610697106981069910700107011070210703107041070510706107071070810709107101071110712107131071410715107161071710718107191072010721107221072310724107251072610727107281072910730107311073210733107341073510736107371073810739107401074110742107431074410745107461074710748107491075010751107521075310754107551075610757107581075910760107611076210763107641076510766107671076810769107701077110772107731077410775107761077710778107791078010781107821078310784107851078610787107881078910790107911079210793107941079510796107971079810799108001080110802108031080410805108061080710808108091081010811108121081310814108151081610817108181081910820108211082210823108241082510826108271082810829108301083110832108331083410835108361083710838108391084010841108421084310844108451084610847108481084910850108511085210853108541085510856108571085810859108601086110862108631086410865108661086710868108691087010871108721087310874108751087610877108781087910880108811088210883108841088510886108871088810889108901089110892108931089410895108961089710898108991090010901109021090310904109051090610907109081090910910109111091210913109141091510916109171091810919109201092110922109231092410925109261092710928109291093010931109321093310934109351093610937109381093910940109411094210943109441094510946109471094810949109501095110952109531095410955109561095710958109591096010961109621096310964109651096610967109681096910970109711097210973109741097510976109771097810979109801098110982109831098410985109861098710988109891099010991109921099310994109951099610997109981099911000110011100211003110041100511006110071100811009110101101111012110131101411015110161101711018110191102011021110221102311024110251102611027110281102911030110311103211033110341103511036110371103811039110401104111042110431104411045110461104711048110491105011051110521105311054110551105611057110581105911060110611106211063110641106511066110671106811069110701107111072110731107411075110761107711078110791108011081110821108311084110851108611087110881108911090110911109211093110941109511096110971109811099111001110111102111031110411105111061110711108111091111011111111121111311114111151111611117111181111911120111211112211123111241112511126111271112811129111301113111132111331113411135111361113711138111391114011141111421114311144111451114611147111481114911150111511115211153111541115511156111571115811159111601116111162111631116411165111661116711168111691117011171111721117311174111751117611177111781117911180111811118211183111841118511186111871118811189111901119111192111931119411195111961119711198111991120011201112021120311204112051120611207112081120911210112111121211213112141121511216112171121811219112201122111222112231122411225112261122711228112291123011231112321123311234112351123611237112381123911240112411124211243112441124511246112471124811249112501125111252112531125411255112561125711258112591126011261112621126311264112651126611267112681126911270112711127211273112741127511276112771127811279112801128111282112831128411285112861128711288112891129011291112921129311294112951129611297112981129911300113011130211303113041130511306113071130811309113101131111312113131131411315113161131711318113191132011321113221132311324113251132611327113281132911330113311133211333113341133511336113371133811339113401134111342113431134411345113461134711348113491135011351113521135311354113551135611357113581135911360113611136211363113641136511366113671136811369113701137111372113731137411375113761137711378113791138011381113821138311384113851138611387113881138911390113911139211393113941139511396113971139811399114001140111402114031140411405114061140711408114091141011411114121141311414114151141611417114181141911420114211142211423114241142511426114271142811429114301143111432114331143411435114361143711438114391144011441114421144311444114451144611447114481144911450114511145211453114541145511456114571145811459114601146111462114631146411465114661146711468114691147011471114721147311474114751147611477114781147911480114811148211483114841148511486114871148811489114901149111492114931149411495114961149711498114991150011501115021150311504115051150611507115081150911510115111151211513115141151511516115171151811519115201152111522115231152411525115261152711528115291153011531115321153311534115351153611537115381153911540115411154211543115441154511546115471154811549115501155111552115531155411555115561155711558115591156011561115621156311564115651156611567115681156911570115711157211573115741157511576115771157811579115801158111582115831158411585115861158711588115891159011591115921159311594115951159611597115981159911600116011160211603116041160511606116071160811609116101161111612116131161411615116161161711618116191162011621116221162311624116251162611627116281162911630116311163211633116341163511636116371163811639116401164111642116431164411645116461164711648116491165011651116521165311654116551165611657116581165911660116611166211663116641166511666116671166811669116701167111672116731167411675116761167711678116791168011681116821168311684116851168611687116881168911690116911169211693116941169511696116971169811699117001170111702117031170411705117061170711708117091171011711117121171311714117151171611717117181171911720117211172211723117241172511726117271172811729117301173111732117331173411735117361173711738117391174011741117421174311744117451174611747117481174911750117511175211753117541175511756117571175811759117601176111762117631176411765117661176711768117691177011771117721177311774117751177611777117781177911780117811178211783117841178511786117871178811789117901179111792117931179411795117961179711798117991180011801118021180311804118051180611807118081180911810118111181211813118141181511816118171181811819118201182111822118231182411825118261182711828118291183011831118321183311834118351183611837118381183911840118411184211843118441184511846118471184811849118501185111852118531185411855118561185711858118591186011861118621186311864118651186611867118681186911870118711187211873118741187511876118771187811879118801188111882118831188411885118861188711888118891189011891118921189311894118951189611897118981189911900119011190211903119041190511906119071190811909119101191111912119131191411915119161191711918119191192011921119221192311924119251192611927119281192911930119311193211933119341193511936119371193811939119401194111942119431194411945119461194711948119491195011951119521195311954119551195611957119581195911960119611196211963119641196511966119671196811969119701197111972119731197411975119761197711978119791198011981119821198311984119851198611987119881198911990119911199211993119941199511996119971199811999120001200112002120031200412005120061200712008120091201012011120121201312014120151201612017120181201912020120211202212023120241202512026120271202812029120301203112032120331203412035120361203712038120391204012041120421204312044120451204612047120481204912050120511205212053120541205512056120571205812059120601206112062120631206412065120661206712068120691207012071120721207312074120751207612077120781207912080120811208212083120841208512086120871208812089120901209112092120931209412095120961209712098120991210012101121021210312104121051210612107121081210912110121111211212113121141211512116121171211812119121201212112122121231212412125121261212712128121291213012131121321213312134121351213612137121381213912140121411214212143121441214512146121471214812149121501215112152121531215412155121561215712158121591216012161121621216312164121651216612167121681216912170121711217212173121741217512176121771217812179121801218112182121831218412185121861218712188121891219012191121921219312194121951219612197121981219912200122011220212203122041220512206122071220812209122101221112212122131221412215122161221712218122191222012221122221222312224122251222612227122281222912230122311223212233122341223512236122371223812239122401224112242122431224412245122461224712248122491225012251122521225312254122551225612257122581225912260122611226212263122641226512266122671226812269122701227112272122731227412275122761227712278122791228012281122821228312284122851228612287122881228912290122911229212293122941229512296122971229812299123001230112302123031230412305123061230712308123091231012311123121231312314123151231612317123181231912320123211232212323123241232512326123271232812329123301233112332123331233412335123361233712338123391234012341123421234312344123451234612347123481234912350123511235212353123541235512356123571235812359123601236112362123631236412365123661236712368123691237012371123721237312374123751237612377123781237912380123811238212383123841238512386123871238812389123901239112392123931239412395123961239712398123991240012401124021240312404124051240612407124081240912410124111241212413124141241512416124171241812419124201242112422124231242412425124261242712428124291243012431124321243312434124351243612437124381243912440124411244212443124441244512446124471244812449124501245112452124531245412455124561245712458124591246012461124621246312464124651246612467124681246912470124711247212473124741247512476124771247812479124801248112482124831248412485124861248712488124891249012491124921249312494124951249612497124981249912500125011250212503125041250512506125071250812509125101251112512125131251412515125161251712518125191252012521125221252312524125251252612527125281252912530125311253212533125341253512536125371253812539125401254112542125431254412545125461254712548125491255012551125521255312554125551255612557125581255912560125611256212563125641256512566125671256812569125701257112572125731257412575125761257712578125791258012581125821258312584125851258612587125881258912590125911259212593125941259512596125971259812599126001260112602126031260412605126061260712608126091261012611126121261312614126151261612617126181261912620126211262212623126241262512626126271262812629126301263112632126331263412635126361263712638126391264012641126421264312644126451264612647126481264912650126511265212653126541265512656126571265812659126601266112662126631266412665126661266712668126691267012671126721267312674126751267612677126781267912680126811268212683126841268512686126871268812689126901269112692126931269412695126961269712698126991270012701127021270312704127051270612707127081270912710127111271212713127141271512716127171271812719127201272112722127231272412725127261272712728127291273012731127321273312734127351273612737127381273912740127411274212743127441274512746127471274812749127501275112752127531275412755127561275712758127591276012761127621276312764127651276612767127681276912770127711277212773127741277512776127771277812779127801278112782127831278412785127861278712788127891279012791127921279312794127951279612797127981279912800128011280212803128041280512806128071280812809128101281112812128131281412815128161281712818128191282012821128221282312824128251282612827128281282912830128311283212833128341283512836128371283812839128401284112842128431284412845128461284712848128491285012851128521285312854128551285612857128581285912860128611286212863128641286512866128671286812869128701287112872128731287412875128761287712878128791288012881128821288312884128851288612887128881288912890128911289212893128941289512896128971289812899129001290112902129031290412905129061290712908129091291012911129121291312914129151291612917129181291912920129211292212923129241292512926129271292812929129301293112932129331293412935129361293712938129391294012941129421294312944129451294612947129481294912950129511295212953129541295512956129571295812959129601296112962129631296412965129661296712968129691297012971129721297312974129751297612977129781297912980129811298212983129841298512986129871298812989129901299112992129931299412995129961299712998129991300013001130021300313004130051300613007130081300913010130111301213013130141301513016130171301813019130201302113022130231302413025130261302713028130291303013031130321303313034130351303613037130381303913040130411304213043130441304513046130471304813049130501305113052130531305413055130561305713058130591306013061130621306313064130651306613067130681306913070130711307213073130741307513076130771307813079130801308113082130831308413085130861308713088130891309013091130921309313094130951309613097130981309913100131011310213103131041310513106131071310813109131101311113112131131311413115131161311713118131191312013121131221312313124131251312613127131281312913130131311313213133131341313513136131371313813139131401314113142131431314413145131461314713148131491315013151131521315313154131551315613157131581315913160131611316213163131641316513166131671316813169131701317113172131731317413175131761317713178131791318013181131821318313184131851318613187131881318913190131911319213193131941319513196131971319813199132001320113202132031320413205132061320713208132091321013211132121321313214132151321613217132181321913220132211322213223132241322513226132271322813229132301323113232132331323413235132361323713238132391324013241132421324313244132451324613247132481324913250132511325213253132541325513256132571325813259132601326113262132631326413265132661326713268132691327013271132721327313274132751327613277132781327913280132811328213283132841328513286132871328813289132901329113292132931329413295132961329713298132991330013301133021330313304133051330613307133081330913310133111331213313133141331513316133171331813319133201332113322133231332413325133261332713328133291333013331133321333313334133351333613337133381333913340133411334213343133441334513346133471334813349133501335113352133531335413355133561335713358133591336013361133621336313364133651336613367133681336913370133711337213373133741337513376133771337813379133801338113382133831338413385133861338713388133891339013391133921339313394133951339613397133981339913400134011340213403134041340513406134071340813409134101341113412134131341413415134161341713418134191342013421134221342313424134251342613427134281342913430134311343213433134341343513436134371343813439134401344113442134431344413445134461344713448134491345013451134521345313454134551345613457134581345913460134611346213463134641346513466134671346813469134701347113472134731347413475134761347713478134791348013481134821348313484134851348613487134881348913490134911349213493134941349513496134971349813499135001350113502135031350413505135061350713508135091351013511135121351313514135151351613517135181351913520135211352213523135241352513526135271352813529135301353113532135331353413535135361353713538135391354013541135421354313544135451354613547135481354913550135511355213553135541355513556135571355813559135601356113562135631356413565135661356713568135691357013571135721357313574135751357613577135781357913580135811358213583135841358513586135871358813589135901359113592135931359413595135961359713598135991360013601136021360313604136051360613607136081360913610136111361213613136141361513616136171361813619136201362113622136231362413625136261362713628136291363013631136321363313634136351363613637136381363913640136411364213643136441364513646136471364813649136501365113652136531365413655136561365713658136591366013661136621366313664136651366613667136681366913670136711367213673136741367513676136771367813679136801368113682136831368413685136861368713688136891369013691136921369313694136951369613697136981369913700137011370213703137041370513706137071370813709137101371113712137131371413715137161371713718137191372013721137221372313724137251372613727137281372913730137311373213733137341373513736137371373813739137401374113742137431374413745137461374713748137491375013751137521375313754137551375613757137581375913760137611376213763137641376513766137671376813769137701377113772137731377413775137761377713778137791378013781137821378313784137851378613787137881378913790137911379213793137941379513796137971379813799138001380113802138031380413805138061380713808138091381013811138121381313814138151381613817138181381913820138211382213823138241382513826138271382813829138301383113832138331383413835138361383713838138391384013841138421384313844138451384613847138481384913850138511385213853138541385513856138571385813859138601386113862138631386413865138661386713868138691387013871138721387313874138751387613877138781387913880138811388213883138841388513886138871388813889138901389113892138931389413895138961389713898138991390013901139021390313904139051390613907139081390913910139111391213913139141391513916139171391813919139201392113922139231392413925139261392713928139291393013931139321393313934139351393613937139381393913940139411394213943139441394513946139471394813949139501395113952139531395413955139561395713958139591396013961139621396313964139651396613967139681396913970139711397213973139741397513976139771397813979139801398113982139831398413985139861398713988139891399013991139921399313994139951399613997139981399914000140011400214003140041400514006140071400814009140101401114012140131401414015140161401714018140191402014021140221402314024140251402614027140281402914030140311403214033140341403514036140371403814039140401404114042140431404414045140461404714048140491405014051140521405314054140551405614057140581405914060140611406214063140641406514066140671406814069140701407114072140731407414075140761407714078140791408014081140821408314084140851408614087140881408914090140911409214093140941409514096140971409814099141001410114102141031410414105141061410714108141091411014111141121411314114141151411614117141181411914120141211412214123141241412514126141271412814129141301413114132141331413414135141361413714138141391414014141141421414314144141451414614147141481414914150141511415214153141541415514156141571415814159141601416114162141631416414165141661416714168141691417014171141721417314174141751417614177141781417914180141811418214183141841418514186141871418814189141901419114192141931419414195141961419714198141991420014201142021420314204142051420614207142081420914210142111421214213142141421514216142171421814219142201422114222142231422414225142261422714228142291423014231142321423314234142351423614237142381423914240142411424214243142441424514246142471424814249142501425114252142531425414255142561425714258142591426014261142621426314264142651426614267142681426914270142711427214273142741427514276142771427814279142801428114282142831428414285142861428714288142891429014291142921429314294142951429614297142981429914300143011430214303143041430514306143071430814309143101431114312143131431414315143161431714318143191432014321143221432314324143251432614327143281432914330143311433214333143341433514336143371433814339143401434114342143431434414345143461434714348143491435014351143521435314354143551435614357143581435914360143611436214363143641436514366143671436814369143701437114372143731437414375143761437714378143791438014381143821438314384143851438614387143881438914390143911439214393143941439514396143971439814399144001440114402144031440414405144061440714408144091441014411144121441314414144151441614417144181441914420144211442214423144241442514426144271442814429144301443114432144331443414435144361443714438144391444014441144421444314444144451444614447144481444914450144511445214453144541445514456144571445814459144601446114462144631446414465144661446714468144691447014471144721447314474144751447614477144781447914480144811448214483144841448514486144871448814489144901449114492144931449414495144961449714498144991450014501145021450314504145051450614507145081450914510145111451214513145141451514516145171451814519145201452114522145231452414525145261452714528145291453014531145321453314534145351453614537145381453914540145411454214543145441454514546145471454814549145501455114552145531455414555145561455714558145591456014561145621456314564145651456614567145681456914570145711457214573145741457514576145771457814579145801458114582145831458414585145861458714588145891459014591145921459314594145951459614597145981459914600146011460214603146041460514606146071460814609146101461114612146131461414615146161461714618146191462014621146221462314624146251462614627146281462914630146311463214633146341463514636146371463814639146401464114642146431464414645146461464714648146491465014651146521465314654146551465614657146581465914660146611466214663146641466514666146671466814669146701467114672146731467414675146761467714678146791468014681146821468314684146851468614687146881468914690146911469214693146941469514696146971469814699147001470114702147031470414705147061470714708147091471014711147121471314714147151471614717147181471914720147211472214723147241472514726147271472814729147301473114732147331473414735147361473714738147391474014741147421474314744147451474614747147481474914750147511475214753147541475514756147571475814759147601476114762147631476414765147661476714768147691477014771147721477314774147751477614777147781477914780147811478214783147841478514786147871478814789147901479114792147931479414795147961479714798147991480014801148021480314804148051480614807148081480914810148111481214813148141481514816148171481814819148201482114822148231482414825148261482714828148291483014831148321483314834148351483614837148381483914840148411484214843148441484514846148471484814849148501485114852148531485414855148561485714858148591486014861148621486314864148651486614867148681486914870148711487214873148741487514876148771487814879148801488114882148831488414885148861488714888148891489014891148921489314894148951489614897148981489914900149011490214903149041490514906149071490814909149101491114912149131491414915149161491714918149191492014921149221492314924149251492614927149281492914930149311493214933149341493514936149371493814939149401494114942149431494414945149461494714948149491495014951149521495314954149551495614957149581495914960149611496214963149641496514966149671496814969149701497114972149731497414975149761497714978149791498014981149821498314984149851498614987149881498914990149911499214993149941499514996149971499814999150001500115002150031500415005150061500715008150091501015011150121501315014150151501615017150181501915020150211502215023150241502515026150271502815029150301503115032150331503415035150361503715038150391504015041150421504315044150451504615047150481504915050150511505215053150541505515056150571505815059150601506115062150631506415065150661506715068150691507015071150721507315074150751507615077150781507915080150811508215083150841508515086150871508815089150901509115092150931509415095150961509715098150991510015101151021510315104151051510615107151081510915110151111511215113151141511515116151171511815119151201512115122151231512415125151261512715128151291513015131151321513315134151351513615137151381513915140151411514215143151441514515146151471514815149151501515115152151531515415155151561515715158151591516015161151621516315164151651516615167151681516915170151711517215173151741517515176151771517815179151801518115182151831518415185151861518715188151891519015191151921519315194151951519615197151981519915200152011520215203152041520515206152071520815209152101521115212152131521415215152161521715218152191522015221152221522315224152251522615227152281522915230152311523215233152341523515236152371523815239152401524115242152431524415245152461524715248152491525015251152521525315254152551525615257152581525915260152611526215263152641526515266152671526815269152701527115272152731527415275152761527715278152791528015281152821528315284152851528615287152881528915290152911529215293152941529515296152971529815299153001530115302153031530415305153061530715308153091531015311153121531315314153151531615317153181531915320153211532215323153241532515326153271532815329153301533115332153331533415335153361533715338153391534015341153421534315344153451534615347153481534915350153511535215353153541535515356153571535815359153601536115362153631536415365153661536715368153691537015371153721537315374153751537615377153781537915380153811538215383153841538515386153871538815389153901539115392153931539415395153961539715398153991540015401154021540315404154051540615407154081540915410154111541215413154141541515416154171541815419154201542115422154231542415425154261542715428154291543015431154321543315434154351543615437154381543915440154411544215443154441544515446154471544815449154501545115452154531545415455154561545715458154591546015461154621546315464154651546615467154681546915470154711547215473154741547515476154771547815479154801548115482154831548415485154861548715488154891549015491154921549315494154951549615497154981549915500155011550215503155041550515506155071550815509155101551115512155131551415515155161551715518155191552015521155221552315524155251552615527155281552915530155311553215533155341553515536155371553815539155401554115542155431554415545155461554715548155491555015551155521555315554155551555615557155581555915560155611556215563155641556515566155671556815569155701557115572155731557415575155761557715578155791558015581155821558315584155851558615587155881558915590155911559215593155941559515596155971559815599156001560115602156031560415605156061560715608156091561015611156121561315614156151561615617156181561915620156211562215623156241562515626156271562815629156301563115632156331563415635156361563715638156391564015641156421564315644156451564615647156481564915650156511565215653156541565515656156571565815659156601566115662156631566415665156661566715668156691567015671156721567315674156751567615677156781567915680156811568215683156841568515686156871568815689156901569115692156931569415695156961569715698156991570015701157021570315704157051570615707157081570915710157111571215713157141571515716157171571815719157201572115722157231572415725157261572715728157291573015731157321573315734157351573615737157381573915740157411574215743157441574515746157471574815749157501575115752157531575415755157561575715758157591576015761157621576315764157651576615767157681576915770157711577215773157741577515776157771577815779157801578115782157831578415785157861578715788157891579015791157921579315794157951579615797157981579915800158011580215803158041580515806158071580815809158101581115812158131581415815158161581715818158191582015821158221582315824158251582615827158281582915830158311583215833158341583515836158371583815839158401584115842158431584415845158461584715848158491585015851158521585315854158551585615857158581585915860158611586215863158641586515866158671586815869158701587115872158731587415875158761587715878158791588015881158821588315884158851588615887158881588915890158911589215893158941589515896158971589815899159001590115902159031590415905159061590715908159091591015911159121591315914159151591615917159181591915920159211592215923159241592515926159271592815929159301593115932159331593415935159361593715938159391594015941159421594315944159451594615947159481594915950159511595215953159541595515956159571595815959159601596115962159631596415965159661596715968159691597015971159721597315974159751597615977159781597915980159811598215983159841598515986159871598815989159901599115992159931599415995159961599715998159991600016001160021600316004160051600616007160081600916010160111601216013160141601516016160171601816019160201602116022160231602416025160261602716028160291603016031160321603316034160351603616037160381603916040160411604216043160441604516046160471604816049160501605116052160531605416055160561605716058160591606016061160621606316064160651606616067160681606916070160711607216073160741607516076160771607816079160801608116082160831608416085160861608716088160891609016091160921609316094160951609616097160981609916100161011610216103161041610516106161071610816109161101611116112161131611416115161161611716118161191612016121161221612316124161251612616127161281612916130161311613216133161341613516136161371613816139161401614116142161431614416145161461614716148161491615016151161521615316154161551615616157161581615916160161611616216163161641616516166161671616816169161701617116172161731617416175161761617716178161791618016181161821618316184161851618616187161881618916190161911619216193161941619516196161971619816199162001620116202162031620416205162061620716208162091621016211162121621316214162151621616217162181621916220162211622216223162241622516226162271622816229162301623116232162331623416235162361623716238162391624016241162421624316244162451624616247162481624916250162511625216253162541625516256162571625816259162601626116262162631626416265162661626716268162691627016271162721627316274162751627616277162781627916280162811628216283162841628516286162871628816289162901629116292162931629416295162961629716298162991630016301163021630316304163051630616307163081630916310163111631216313163141631516316163171631816319163201632116322163231632416325163261632716328163291633016331163321633316334163351633616337163381633916340163411634216343163441634516346163471634816349163501635116352163531635416355163561635716358163591636016361163621636316364163651636616367163681636916370163711637216373163741637516376163771637816379163801638116382163831638416385163861638716388163891639016391163921639316394163951639616397163981639916400164011640216403164041640516406164071640816409164101641116412164131641416415164161641716418164191642016421164221642316424164251642616427164281642916430164311643216433164341643516436164371643816439164401644116442164431644416445164461644716448164491645016451164521645316454164551645616457164581645916460164611646216463164641646516466164671646816469164701647116472164731647416475164761647716478164791648016481164821648316484164851648616487164881648916490164911649216493164941649516496164971649816499165001650116502165031650416505165061650716508165091651016511165121651316514165151651616517165181651916520165211652216523165241652516526165271652816529165301653116532165331653416535165361653716538165391654016541165421654316544165451654616547165481654916550165511655216553165541655516556165571655816559165601656116562165631656416565165661656716568165691657016571165721657316574165751657616577165781657916580165811658216583165841658516586165871658816589165901659116592165931659416595165961659716598165991660016601166021660316604166051660616607166081660916610166111661216613166141661516616166171661816619166201662116622166231662416625166261662716628166291663016631166321663316634166351663616637166381663916640166411664216643166441664516646166471664816649166501665116652166531665416655166561665716658166591666016661166621666316664166651666616667166681666916670166711667216673166741667516676166771667816679166801668116682166831668416685166861668716688166891669016691166921669316694166951669616697166981669916700167011670216703167041670516706167071670816709167101671116712167131671416715167161671716718167191672016721167221672316724167251672616727167281672916730167311673216733167341673516736167371673816739167401674116742167431674416745167461674716748167491675016751167521675316754167551675616757167581675916760167611676216763167641676516766167671676816769167701677116772167731677416775167761677716778167791678016781167821678316784167851678616787167881678916790167911679216793167941679516796167971679816799168001680116802168031680416805168061680716808168091681016811168121681316814168151681616817168181681916820168211682216823168241682516826168271682816829168301683116832168331683416835168361683716838168391684016841168421684316844168451684616847168481684916850168511685216853168541685516856168571685816859168601686116862168631686416865168661686716868168691687016871168721687316874168751687616877168781687916880168811688216883168841688516886168871688816889168901689116892168931689416895168961689716898168991690016901169021690316904169051690616907169081690916910169111691216913169141691516916169171691816919169201692116922169231692416925169261692716928169291693016931169321693316934169351693616937169381693916940169411694216943169441694516946169471694816949169501695116952169531695416955169561695716958169591696016961169621696316964169651696616967169681696916970169711697216973169741697516976169771697816979169801698116982169831698416985169861698716988169891699016991169921699316994169951699616997169981699917000170011700217003170041700517006170071700817009170101701117012170131701417015170161701717018170191702017021170221702317024170251702617027170281702917030170311703217033170341703517036170371703817039170401704117042170431704417045170461704717048170491705017051170521705317054170551705617057170581705917060170611706217063170641706517066170671706817069170701707117072170731707417075170761707717078170791708017081170821708317084170851708617087170881708917090170911709217093170941709517096170971709817099171001710117102171031710417105171061710717108171091711017111171121711317114171151711617117171181711917120171211712217123171241712517126171271712817129171301713117132171331713417135171361713717138171391714017141171421714317144171451714617147171481714917150171511715217153171541715517156171571715817159171601716117162171631716417165171661716717168171691717017171171721717317174171751717617177171781717917180171811718217183171841718517186171871718817189171901719117192171931719417195171961719717198171991720017201172021720317204172051720617207172081720917210172111721217213172141721517216172171721817219172201722117222172231722417225172261722717228172291723017231172321723317234172351723617237172381723917240172411724217243172441724517246172471724817249172501725117252172531725417255172561725717258172591726017261172621726317264172651726617267172681726917270172711727217273172741727517276172771727817279172801728117282172831728417285172861728717288172891729017291172921729317294172951729617297172981729917300173011730217303173041730517306173071730817309173101731117312173131731417315173161731717318173191732017321173221732317324173251732617327173281732917330173311733217333173341733517336173371733817339173401734117342173431734417345173461734717348173491735017351173521735317354173551735617357173581735917360173611736217363173641736517366173671736817369173701737117372173731737417375173761737717378173791738017381173821738317384173851738617387173881738917390173911739217393173941739517396173971739817399174001740117402174031740417405174061740717408174091741017411174121741317414174151741617417174181741917420174211742217423174241742517426174271742817429174301743117432174331743417435174361743717438174391744017441174421744317444174451744617447174481744917450174511745217453174541745517456174571745817459174601746117462174631746417465174661746717468174691747017471174721747317474174751747617477174781747917480174811748217483174841748517486174871748817489174901749117492174931749417495174961749717498174991750017501175021750317504175051750617507175081750917510175111751217513175141751517516175171751817519175201752117522175231752417525175261752717528175291753017531175321753317534175351753617537175381753917540175411754217543175441754517546175471754817549175501755117552175531755417555175561755717558175591756017561175621756317564175651756617567175681756917570175711757217573175741757517576175771757817579175801758117582175831758417585175861758717588175891759017591175921759317594175951759617597175981759917600176011760217603176041760517606176071760817609176101761117612176131761417615176161761717618176191762017621176221762317624176251762617627176281762917630176311763217633176341763517636176371763817639176401764117642176431764417645176461764717648176491765017651176521765317654176551765617657176581765917660176611766217663176641766517666176671766817669176701767117672176731767417675176761767717678176791768017681176821768317684176851768617687176881768917690176911769217693176941769517696176971769817699177001770117702177031770417705177061770717708177091771017711177121771317714177151771617717177181771917720177211772217723177241772517726177271772817729177301773117732177331773417735177361773717738177391774017741177421774317744177451774617747177481774917750177511775217753177541775517756177571775817759177601776117762177631776417765177661776717768177691777017771177721777317774177751777617777177781777917780177811778217783177841778517786177871778817789177901779117792177931779417795177961779717798177991780017801178021780317804178051780617807178081780917810178111781217813178141781517816178171781817819178201782117822178231782417825178261782717828178291783017831178321783317834178351783617837178381783917840178411784217843178441784517846178471784817849178501785117852178531785417855178561785717858178591786017861178621786317864178651786617867178681786917870178711787217873178741787517876178771787817879178801788117882178831788417885178861788717888178891789017891178921789317894178951789617897178981789917900179011790217903179041790517906179071790817909179101791117912179131791417915179161791717918179191792017921179221792317924179251792617927179281792917930179311793217933179341793517936179371793817939179401794117942179431794417945179461794717948179491795017951179521795317954179551795617957179581795917960179611796217963179641796517966179671796817969179701797117972179731797417975179761797717978179791798017981179821798317984179851798617987179881798917990179911799217993179941799517996179971799817999180001800118002180031800418005180061800718008180091801018011180121801318014180151801618017180181801918020180211802218023180241802518026180271802818029180301803118032180331803418035180361803718038180391804018041180421804318044180451804618047180481804918050180511805218053180541805518056180571805818059180601806118062180631806418065180661806718068180691807018071180721807318074180751807618077180781807918080180811808218083180841808518086180871808818089180901809118092180931809418095180961809718098180991810018101181021810318104181051810618107181081810918110181111811218113181141811518116181171811818119181201812118122181231812418125181261812718128181291813018131181321813318134181351813618137181381813918140181411814218143181441814518146181471814818149181501815118152181531815418155181561815718158181591816018161181621816318164181651816618167181681816918170181711817218173181741817518176181771817818179181801818118182181831818418185181861818718188181891819018191181921819318194181951819618197181981819918200182011820218203182041820518206182071820818209182101821118212182131821418215182161821718218182191822018221182221822318224182251822618227182281822918230182311823218233182341823518236182371823818239182401824118242182431824418245182461824718248182491825018251182521825318254182551825618257182581825918260182611826218263182641826518266182671826818269182701827118272182731827418275182761827718278182791828018281182821828318284182851828618287182881828918290182911829218293182941829518296182971829818299183001830118302183031830418305183061830718308183091831018311183121831318314183151831618317183181831918320183211832218323183241832518326183271832818329183301833118332183331833418335183361833718338183391834018341183421834318344183451834618347183481834918350183511835218353183541835518356183571835818359183601836118362183631836418365183661836718368183691837018371183721837318374183751837618377183781837918380183811838218383183841838518386183871838818389183901839118392183931839418395183961839718398183991840018401184021840318404184051840618407184081840918410184111841218413184141841518416184171841818419184201842118422184231842418425184261842718428184291843018431184321843318434184351843618437184381843918440184411844218443184441844518446184471844818449184501845118452184531845418455184561845718458184591846018461184621846318464184651846618467184681846918470184711847218473184741847518476184771847818479184801848118482184831848418485184861848718488184891849018491184921849318494184951849618497184981849918500185011850218503185041850518506185071850818509185101851118512185131851418515185161851718518185191852018521185221852318524185251852618527185281852918530185311853218533185341853518536185371853818539185401854118542185431854418545185461854718548185491855018551185521855318554185551855618557185581855918560185611856218563185641856518566185671856818569185701857118572185731857418575185761857718578185791858018581185821858318584185851858618587185881858918590185911859218593185941859518596185971859818599186001860118602186031860418605186061860718608186091861018611186121861318614186151861618617186181861918620186211862218623186241862518626186271862818629186301863118632186331863418635186361863718638186391864018641186421864318644186451864618647186481864918650186511865218653186541865518656186571865818659186601866118662186631866418665186661866718668186691867018671186721867318674186751867618677186781867918680186811868218683186841868518686186871868818689186901869118692186931869418695186961869718698186991870018701187021870318704187051870618707187081870918710187111871218713187141871518716187171871818719187201872118722187231872418725187261872718728187291873018731187321873318734187351873618737187381873918740187411874218743187441874518746187471874818749187501875118752187531875418755187561875718758187591876018761187621876318764187651876618767187681876918770187711877218773187741877518776187771877818779187801878118782187831878418785187861878718788187891879018791187921879318794187951879618797187981879918800188011880218803188041880518806188071880818809188101881118812188131881418815188161881718818188191882018821188221882318824188251882618827188281882918830188311883218833188341883518836188371883818839188401884118842188431884418845188461884718848188491885018851188521885318854188551885618857188581885918860188611886218863188641886518866188671886818869188701887118872188731887418875188761887718878188791888018881188821888318884188851888618887188881888918890188911889218893188941889518896188971889818899189001890118902189031890418905189061890718908189091891018911189121891318914189151891618917189181891918920189211892218923189241892518926189271892818929189301893118932189331893418935189361893718938189391894018941189421894318944189451894618947189481894918950189511895218953189541895518956189571895818959189601896118962189631896418965189661896718968189691897018971189721897318974189751897618977189781897918980189811898218983189841898518986189871898818989189901899118992189931899418995189961899718998189991900019001190021900319004190051900619007190081900919010190111901219013190141901519016190171901819019190201902119022190231902419025190261902719028190291903019031190321903319034190351903619037190381903919040190411904219043190441904519046190471904819049190501905119052190531905419055190561905719058190591906019061190621906319064190651906619067190681906919070190711907219073190741907519076190771907819079190801908119082190831908419085190861908719088190891909019091190921909319094190951909619097190981909919100191011910219103191041910519106191071910819109191101911119112191131911419115191161911719118191191912019121191221912319124191251912619127191281912919130191311913219133191341913519136191371913819139191401914119142191431914419145191461914719148191491915019151191521915319154191551915619157191581915919160191611916219163191641916519166191671916819169191701917119172191731917419175191761917719178191791918019181191821918319184191851918619187191881918919190191911919219193191941919519196191971919819199192001920119202192031920419205192061920719208192091921019211192121921319214192151921619217192181921919220192211922219223192241922519226192271922819229192301923119232192331923419235192361923719238192391924019241192421924319244192451924619247192481924919250192511925219253192541925519256192571925819259192601926119262192631926419265192661926719268192691927019271192721927319274192751927619277192781927919280192811928219283192841928519286192871928819289192901929119292192931929419295192961929719298192991930019301193021930319304193051930619307193081930919310193111931219313193141931519316193171931819319193201932119322193231932419325193261932719328193291933019331193321933319334193351933619337193381933919340193411934219343193441934519346193471934819349193501935119352193531935419355193561935719358193591936019361193621936319364193651936619367193681936919370193711937219373193741937519376193771937819379193801938119382193831938419385193861938719388193891939019391193921939319394193951939619397193981939919400194011940219403194041940519406194071940819409194101941119412194131941419415194161941719418194191942019421194221942319424194251942619427194281942919430194311943219433194341943519436194371943819439194401944119442194431944419445194461944719448194491945019451194521945319454194551945619457194581945919460194611946219463194641946519466194671946819469194701947119472194731947419475194761947719478194791948019481194821948319484194851948619487194881948919490194911949219493194941949519496194971949819499195001950119502195031950419505195061950719508195091951019511195121951319514195151951619517195181951919520195211952219523195241952519526195271952819529195301953119532195331953419535195361953719538195391954019541195421954319544195451954619547195481954919550195511955219553195541955519556195571955819559195601956119562195631956419565195661956719568195691957019571195721957319574195751957619577195781957919580195811958219583195841958519586195871958819589195901959119592195931959419595195961959719598195991960019601196021960319604196051960619607196081960919610196111961219613196141961519616196171961819619196201962119622196231962419625196261962719628196291963019631196321963319634196351963619637196381963919640196411964219643196441964519646196471964819649196501965119652196531965419655196561965719658196591966019661196621966319664196651966619667196681966919670196711967219673196741967519676196771967819679196801968119682196831968419685196861968719688196891969019691196921969319694196951969619697196981969919700197011970219703197041970519706197071970819709197101971119712197131971419715197161971719718197191972019721197221972319724197251972619727197281972919730197311973219733197341973519736197371973819739197401974119742197431974419745197461974719748197491975019751197521975319754197551975619757197581975919760197611976219763197641976519766197671976819769197701977119772197731977419775197761977719778197791978019781197821978319784197851978619787197881978919790197911979219793197941979519796197971979819799198001980119802198031980419805198061980719808198091981019811198121981319814198151981619817198181981919820198211982219823198241982519826198271982819829198301983119832198331983419835198361983719838198391984019841198421984319844198451984619847198481984919850198511985219853198541985519856198571985819859198601986119862198631986419865198661986719868198691987019871198721987319874198751987619877198781987919880198811988219883198841988519886198871988819889198901989119892198931989419895198961989719898198991990019901199021990319904199051990619907199081990919910199111991219913199141991519916199171991819919199201992119922199231992419925199261992719928199291993019931199321993319934199351993619937199381993919940199411994219943199441994519946199471994819949199501995119952199531995419955199561995719958199591996019961199621996319964199651996619967199681996919970199711997219973199741997519976199771997819979199801998119982199831998419985199861998719988199891999019991199921999319994199951999619997199981999920000200012000220003200042000520006200072000820009200102001120012200132001420015200162001720018200192002020021200222002320024200252002620027200282002920030200312003220033200342003520036200372003820039200402004120042200432004420045200462004720048200492005020051200522005320054200552005620057200582005920060200612006220063200642006520066200672006820069200702007120072200732007420075200762007720078200792008020081200822008320084200852008620087200882008920090200912009220093200942009520096200972009820099201002010120102201032010420105201062010720108201092011020111201122011320114201152011620117201182011920120201212012220123201242012520126201272012820129201302013120132201332013420135201362013720138201392014020141201422014320144201452014620147201482014920150201512015220153201542015520156201572015820159201602016120162201632016420165201662016720168201692017020171201722017320174201752017620177201782017920180201812018220183201842018520186201872018820189201902019120192201932019420195201962019720198201992020020201202022020320204202052020620207202082020920210202112021220213202142021520216202172021820219202202022120222202232022420225202262022720228202292023020231202322023320234202352023620237202382023920240202412024220243202442024520246202472024820249202502025120252202532025420255202562025720258202592026020261202622026320264202652026620267202682026920270202712027220273202742027520276202772027820279202802028120282202832028420285202862028720288202892029020291202922029320294202952029620297202982029920300203012030220303203042030520306203072030820309203102031120312203132031420315203162031720318203192032020321203222032320324203252032620327203282032920330203312033220333203342033520336203372033820339203402034120342203432034420345203462034720348203492035020351203522035320354203552035620357203582035920360203612036220363203642036520366203672036820369203702037120372203732037420375203762037720378203792038020381203822038320384203852038620387203882038920390203912039220393203942039520396203972039820399204002040120402204032040420405204062040720408204092041020411204122041320414204152041620417204182041920420204212042220423204242042520426204272042820429204302043120432204332043420435204362043720438204392044020441204422044320444204452044620447204482044920450204512045220453204542045520456204572045820459204602046120462204632046420465204662046720468204692047020471204722047320474204752047620477204782047920480204812048220483204842048520486204872048820489204902049120492204932049420495204962049720498204992050020501205022050320504205052050620507205082050920510205112051220513205142051520516205172051820519205202052120522205232052420525205262052720528205292053020531205322053320534205352053620537205382053920540205412054220543205442054520546205472054820549205502055120552205532055420555205562055720558205592056020561205622056320564205652056620567205682056920570205712057220573205742057520576205772057820579205802058120582205832058420585205862058720588205892059020591205922059320594205952059620597205982059920600206012060220603206042060520606206072060820609206102061120612206132061420615206162061720618206192062020621206222062320624206252062620627206282062920630206312063220633206342063520636206372063820639206402064120642206432064420645206462064720648206492065020651206522065320654206552065620657206582065920660206612066220663206642066520666206672066820669206702067120672206732067420675206762067720678206792068020681206822068320684206852068620687206882068920690206912069220693206942069520696206972069820699207002070120702207032070420705207062070720708207092071020711207122071320714207152071620717207182071920720207212072220723207242072520726207272072820729207302073120732207332073420735207362073720738207392074020741207422074320744
  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 a DRAFT watermark
  12. \usepackage{draftwatermark}
  13. \usepackage{accsupp}
  14. \SetWatermarkLightness{0.97}
  15. \SetWatermarkScale{1}
  16. % Make watermark not copyable (in Adobe Reader)
  17. \SetWatermarkText{\BeginAccSupp{method=escape,ActualText={}}DRAFT\EndAccSupp{}}
  18. % Set up required header format
  19. \usepackage{fancyhdr}
  20. \pagestyle{fancy}
  21. \renewcommand{\headrulewidth}{0pt}
  22. \rhead{}
  23. \lhead{}
  24. \chead{}
  25. \rfoot{}
  26. \lfoot{}
  27. % Make page number not copyable (in Adobe Reader)
  28. \cfoot{\BeginAccSupp{method=escape,ActualText={}}\thepage\EndAccSupp{}} % Page number bottom center
  29. % Allow FloatBarrier command
  30. \usepackage{placeins}
  31. % Allow landscape pages
  32. \usepackage{pdflscape}
  33. % Allow doing things after the end of the current page
  34. % (to avoid landscape figures breaking up text)
  35. \usepackage{afterpage}
  36. % This one breaks subfigs so it's disabled
  37. % https://tex.stackexchange.com/questions/65680/automatically-bold-first-sentence-of-a-floats-caption
  38. % Bold all nomenclature entries
  39. \renewcommand{\nomlabel}[1]{\textsf{\textbf{#1}}}
  40. % https://tex.stackexchange.com/a/31083/5654
  41. %\let\nomenclOrig\nomenclature
  42. %\renewcommand*{\nomenclature}[3][]{#2\nomenclOrig[#1]{#2}{#3}}
  43. \usepackage[nohypertypes={abbreviation}]{glossaries-extra}
  44. \setabbreviationstyle{long-short}
  45. \input{abbrevs.tex}
  46. \makeglossaries
  47. \end_preamble
  48. \use_default_options true
  49. \begin_modules
  50. todonotes
  51. logicalmkup
  52. \end_modules
  53. \maintain_unincluded_children false
  54. \begin_local_layout
  55. Format 66
  56. InsetLayout "Flex:Glossary Term"
  57. LyxType custom
  58. LabelString gls
  59. LatexType command
  60. LatexName gls*
  61. InToc true
  62. CustomPars false
  63. End
  64. InsetLayout "Flex:Glossary Term (Capital)"
  65. LyxType custom
  66. LabelString Gls
  67. LatexType command
  68. LatexName Gls*
  69. InToc true
  70. CustomPars false
  71. End
  72. InsetLayout "Flex:Glossary Term (glstext)"
  73. LyxType custom
  74. LabelString glstext
  75. LatexType command
  76. LatexName glstext*
  77. InToc true
  78. CustomPars false
  79. End
  80. InsetLayout "Flex:Glossary Term (Glstext)"
  81. LyxType custom
  82. LabelString Glstext
  83. LatexType command
  84. LatexName Glstext*
  85. InToc true
  86. CustomPars false
  87. End
  88. InsetLayout "Flex:Glossary Term (glsfirst)"
  89. LyxType custom
  90. LabelString glsfirst
  91. LatexType command
  92. LatexName glsfirst*
  93. InToc true
  94. CustomPars false
  95. End
  96. InsetLayout "Flex:Glossary Term (Glsfirst)"
  97. LyxType custom
  98. LabelString Glsfirst
  99. LatexType command
  100. LatexName Glsfirst*
  101. InToc true
  102. CustomPars false
  103. End
  104. InsetLayout "Flex:Glossary Term (glsdesc)"
  105. LyxType custom
  106. LabelString glsdesc
  107. LatexType command
  108. LatexName glsdesc*
  109. InToc true
  110. CustomPars false
  111. End
  112. InsetLayout "Flex:Glossary Term (Glsdesc)"
  113. LyxType custom
  114. LabelString Glsdesc
  115. LatexType command
  116. LatexName Glsdesc*
  117. InToc true
  118. CustomPars false
  119. End
  120. \end_local_layout
  121. \language english
  122. \language_package default
  123. \inputencoding utf8
  124. \fontencoding default
  125. \font_roman "default" "default"
  126. \font_sans "default" "default"
  127. \font_typewriter "default" "default"
  128. \font_math "auto" "auto"
  129. \font_default_family default
  130. \use_non_tex_fonts false
  131. \font_sc false
  132. \font_osf false
  133. \font_sf_scale 100 100
  134. \font_tt_scale 100 100
  135. \use_microtype false
  136. \use_dash_ligatures true
  137. \graphics default
  138. \default_output_format pdf4
  139. \output_sync 0
  140. \bibtex_command biber
  141. \index_command default
  142. \paperfontsize 12
  143. \spacing double
  144. \use_hyperref true
  145. \pdf_bookmarks true
  146. \pdf_bookmarksnumbered false
  147. \pdf_bookmarksopen false
  148. \pdf_bookmarksopenlevel 1
  149. \pdf_breaklinks false
  150. \pdf_pdfborder false
  151. \pdf_colorlinks false
  152. \pdf_backref false
  153. \pdf_pdfusetitle true
  154. \papersize letterpaper
  155. \use_geometry true
  156. \use_package amsmath 1
  157. \use_package amssymb 1
  158. \use_package cancel 1
  159. \use_package esint 1
  160. \use_package mathdots 1
  161. \use_package mathtools 1
  162. \use_package mhchem 1
  163. \use_package stackrel 1
  164. \use_package stmaryrd 1
  165. \use_package undertilde 1
  166. \cite_engine biblatex
  167. \cite_engine_type authoryear
  168. \biblio_style plain
  169. \biblatex_bibstyle authoryear
  170. \biblatex_citestyle numeric
  171. \use_bibtopic false
  172. \use_indices false
  173. \paperorientation portrait
  174. \suppress_date false
  175. \justification true
  176. \use_refstyle 1
  177. \use_minted 0
  178. \index Index
  179. \shortcut idx
  180. \color #008000
  181. \end_index
  182. \leftmargin 1.5in
  183. \topmargin 1in
  184. \rightmargin 1in
  185. \bottommargin 1in
  186. \secnumdepth 3
  187. \tocdepth 3
  188. \paragraph_separation indent
  189. \paragraph_indentation default
  190. \is_math_indent 0
  191. \math_numbering_side default
  192. \quotes_style english
  193. \dynamic_quotes 0
  194. \papercolumns 1
  195. \papersides 1
  196. \paperpagestyle default
  197. \tracking_changes false
  198. \output_changes false
  199. \html_math_output 0
  200. \html_css_as_file 0
  201. \html_be_strict false
  202. \end_header
  203. \begin_body
  204. \begin_layout Title
  205. Bioinformatic analysis of complex, high-throughput genomic and epigenomic
  206. data in the context of immunology and transplant rejection
  207. \end_layout
  208. \begin_layout Author
  209. A thesis presented
  210. \begin_inset Newline newline
  211. \end_inset
  212. by
  213. \begin_inset Newline newline
  214. \end_inset
  215. Ryan C.
  216. Thompson
  217. \begin_inset Newline newline
  218. \end_inset
  219. to
  220. \begin_inset Newline newline
  221. \end_inset
  222. The Scripps Research Institute Graduate Program
  223. \begin_inset Newline newline
  224. \end_inset
  225. in partial fulfillment of the requirements for the degree of
  226. \begin_inset Newline newline
  227. \end_inset
  228. Doctor of Philosophy in the subject of Biology
  229. \begin_inset Newline newline
  230. \end_inset
  231. for
  232. \begin_inset Newline newline
  233. \end_inset
  234. The Scripps Research Institute
  235. \begin_inset Newline newline
  236. \end_inset
  237. La Jolla, California
  238. \end_layout
  239. \begin_layout Date
  240. October 2019
  241. \end_layout
  242. \begin_layout Standard
  243. [Copyright notice]
  244. \end_layout
  245. \begin_layout Standard
  246. [Thesis acceptance form]
  247. \end_layout
  248. \begin_layout Standard
  249. [Dedication]
  250. \end_layout
  251. \begin_layout Standard
  252. [Acknowledgements]
  253. \end_layout
  254. \begin_layout Standard
  255. \begin_inset CommandInset toc
  256. LatexCommand tableofcontents
  257. \end_inset
  258. \end_layout
  259. \begin_layout Standard
  260. \begin_inset FloatList table
  261. \end_inset
  262. \end_layout
  263. \begin_layout Standard
  264. \begin_inset FloatList figure
  265. \end_inset
  266. \end_layout
  267. \begin_layout Standard
  268. \begin_inset Note Note
  269. status open
  270. \begin_layout Plain Layout
  271. To create a new nomenclature entry:
  272. \end_layout
  273. \begin_layout Enumerate
  274. Add an entry to abbrevs.tex
  275. \end_layout
  276. \begin_layout Enumerate
  277. Find the first instance of the term, and wrap it in Insert -> Custom Insets
  278. -> Glossary Term (use Capital if starting a sentence)
  279. \end_layout
  280. \begin_layout Enumerate
  281. Add a nomenclature entry after the first instance
  282. \end_layout
  283. \begin_layout Enumerate
  284. Replace every relevant instance throughout the document with the Glossary
  285. Term wrapped version, using Edit -> Find & Replace (Advanced).
  286. Skip section headers and floats.
  287. \end_layout
  288. \begin_layout Plain Layout
  289. \begin_inset CommandInset href
  290. LatexCommand href
  291. target "https://ctan.org/pkg/glossaries?lang=en"
  292. literal "false"
  293. \end_inset
  294. \end_layout
  295. \begin_layout Plain Layout
  296. \begin_inset CommandInset href
  297. LatexCommand href
  298. target "https://wiki.lyx.org/Tips/Nomenclature"
  299. literal "false"
  300. \end_inset
  301. \end_layout
  302. \end_inset
  303. \end_layout
  304. \begin_layout Standard
  305. \begin_inset CommandInset nomencl_print
  306. LatexCommand printnomenclature
  307. set_width "auto"
  308. \end_inset
  309. \end_layout
  310. \begin_layout List of TODOs
  311. \end_layout
  312. \begin_layout Standard
  313. \begin_inset Flex TODO Note (inline)
  314. status open
  315. \begin_layout Plain Layout
  316. Check all figures to make sure they fit on the page with their legends.
  317. \end_layout
  318. \end_inset
  319. \end_layout
  320. \begin_layout Standard
  321. \begin_inset Flex TODO Note (inline)
  322. status open
  323. \begin_layout Plain Layout
  324. Look into auto-generated nomenclature list:
  325. \begin_inset CommandInset href
  326. LatexCommand href
  327. target "https://wiki.lyx.org/Tips/Nomenclature"
  328. \end_inset
  329. .
  330. Otherwise, do a manual pass for all abbreviations at the end.
  331. Do nomenclature/abbreviations independently for each chapter.
  332. \end_layout
  333. \end_inset
  334. \end_layout
  335. \begin_layout Standard
  336. \begin_inset Flex TODO Note (inline)
  337. status open
  338. \begin_layout Plain Layout
  339. Make all descriptions consistent in terms of
  340. \begin_inset Quotes eld
  341. \end_inset
  342. we did X
  343. \begin_inset Quotes erd
  344. \end_inset
  345. vs
  346. \begin_inset Quotes eld
  347. \end_inset
  348. I did X
  349. \begin_inset Quotes erd
  350. \end_inset
  351. vs
  352. \begin_inset Quotes eld
  353. \end_inset
  354. X was done
  355. \begin_inset Quotes erd
  356. \end_inset
  357. .
  358. \end_layout
  359. \end_inset
  360. \end_layout
  361. \begin_layout Chapter*
  362. Abstract
  363. \end_layout
  364. \begin_layout Standard
  365. \begin_inset Note Note
  366. status open
  367. \begin_layout Plain Layout
  368. It is included as an integral part of the thesis and should immediately
  369. precede the introduction.
  370. \end_layout
  371. \begin_layout Plain Layout
  372. Preparing your Abstract.
  373. Your abstract (a succinct description of your work) is limited to 350 words.
  374. UMI will shorten it if they must; please do not exceed the limit.
  375. \end_layout
  376. \begin_layout Itemize
  377. Include pertinent place names, names of persons (in full), and other proper
  378. nouns.
  379. These are useful in automated retrieval.
  380. \end_layout
  381. \begin_layout Itemize
  382. Display symbols, as well as foreign words and phrases, clearly and accurately.
  383. Include transliterations for characters other than Roman and Greek letters
  384. and Arabic numerals.
  385. Include accents and diacritical marks.
  386. \end_layout
  387. \begin_layout Itemize
  388. Do not include graphs, charts, tables, or illustrations in your abstract.
  389. \end_layout
  390. \end_inset
  391. \end_layout
  392. \begin_layout Standard
  393. \begin_inset Flex TODO Note (inline)
  394. status open
  395. \begin_layout Plain Layout
  396. Obviously the abstract gets written last.
  397. \end_layout
  398. \end_inset
  399. \end_layout
  400. \begin_layout Chapter*
  401. Notes to draft readers
  402. \end_layout
  403. \begin_layout Standard
  404. Thank you so much for agreeing to read my thesis and give me feedback on
  405. it.
  406. What you are currently reading is a rough draft, in need of many revisions.
  407. You can always find the latest version at
  408. \begin_inset CommandInset href
  409. LatexCommand href
  410. target "https://mneme.dedyn.io/~ryan/Thesis/thesis.pdf"
  411. literal "false"
  412. \end_inset
  413. .
  414. the PDF at this link is updated periodically with my latest revisions,
  415. but you can just download the current version and give me feedback on that.
  416. Don't worry about keeping up with the updates.
  417. \end_layout
  418. \begin_layout Standard
  419. As for what feedback I'm looking for, first of all, don't waste your time
  420. marking spelling mistakes and such.
  421. I haven't run a spell checker on it yet, so let me worry about that.
  422. Also, I'm aware that many abbreviations are not properly introduced the
  423. first time they are used, so don't worry about that either.
  424. However, if you see any glaring formatting issues, such as a figure being
  425. too large and getting cut off at the edge of the page, please note them.
  426. In addition, if any of the text in the figures is too small, please note
  427. that as well.
  428. \end_layout
  429. \begin_layout Standard
  430. Beyond that, what I'm mainly interested in is feedback on the content.
  431. For example: does the introduction flow logically, and does it provide
  432. enough background to understand the other chapters? Does each chapter make
  433. it clear what work and analyses I have done? Do the figures clearly communicate
  434. the results I'm trying to show? Do you feel that the claims in the results
  435. and discussion sections are well-supported? There's no need to suggest
  436. improvements; just note areas that you feel need improvement.
  437. Additionally, while I am well aware that Chapter 1 (the introduction) contains
  438. many un-cited claims, all the other chapters (2,3, and 4)
  439. \emph on
  440. should
  441. \emph default
  442. be fully cited.
  443. So if you notice any un-cited claims in those chapters, please flag them
  444. for my attention.
  445. Similarly, if you discover any factual errors, please note them as well.
  446. \end_layout
  447. \begin_layout Standard
  448. You can provide your feedback in whatever way is most convenient to you.
  449. You could mark up this PDF with highlights and notes, then send it back
  450. to me.
  451. Or you could collect your comments in a separate text file and send that
  452. to me, or whatever else you like.
  453. However, if you send me your feedback in a separate document, please note
  454. a section/figure/table number for each comment, and
  455. \emph on
  456. also
  457. \emph default
  458. send me the exact PDF that you read so I can reference it while reading
  459. your comments, since as mentioned above, the current version I'm working
  460. on will have changed by that point (which might include shuffling sections
  461. and figures around, changing their numbers).
  462. One last thing: you'll see a bunch of text in orange boxes throughout the
  463. PDF.
  464. These are notes to myself about things that need to be fixed later, so
  465. if you see a problem noted in an orange box, that means I'm already aware
  466. of it, and there's no need to comment on it.
  467. \end_layout
  468. \begin_layout Standard
  469. My thesis is due Thursday, October 10th, so in order to be useful to me,
  470. I'll need your feedback at least a few days before that, ideally by Monday,
  471. October 7th.
  472. If you have limited time and are unable to get through the whole thesis,
  473. please focus your efforts on Chapters 1 and 2, since those are the roughest
  474. and most in need of revision.
  475. Chapter 3 is fairly short and straightforward, and Chapter 4 is an adaptation
  476. of a paper that's already been through a few rounds of revision, so they
  477. should be a lot tighter.
  478. If you can't spare any time between now and then, or if something unexpected
  479. comes up, I understand.
  480. Just let me know.
  481. \end_layout
  482. \begin_layout Standard
  483. Thanks again for your help, and happy reading!
  484. \end_layout
  485. \begin_layout Chapter
  486. Introduction
  487. \end_layout
  488. \begin_layout Section
  489. Background & Significance
  490. \end_layout
  491. \begin_layout Subsection
  492. Biological motivation
  493. \end_layout
  494. \begin_layout Standard
  495. \begin_inset Flex TODO Note (inline)
  496. status open
  497. \begin_layout Plain Layout
  498. Rethink the subsection organization after the intro is written.
  499. \end_layout
  500. \end_inset
  501. \end_layout
  502. \begin_layout Standard
  503. \begin_inset Flex TODO Note (inline)
  504. status open
  505. \begin_layout Plain Layout
  506. Citations are needed all over the place.
  507. A lot of this is knowledge I've just absorbed from years of conversation
  508. in the Salomon lab, without ever having seen a citation for it.
  509. \end_layout
  510. \end_inset
  511. \end_layout
  512. \begin_layout Subsubsection
  513. Rejection is the major long-term threat to organ and tissue allografts
  514. \end_layout
  515. \begin_layout Standard
  516. Organ and tissue transplants are a life-saving treatment for people who
  517. have lost the function of an important organ [CITE?].
  518. In some cases, it is possible to transplant a patient's own tissue from
  519. one area of their body to another, referred to as an autograft.
  520. This is common for tissues that are distributed throughout many areas of
  521. the body, such as skin and bone.
  522. However, in cases of organ failure, there is no functional self tissue
  523. remaining, and a transplant from another person – a donor – is required.
  524. This is referred to as an allograft.
  525. \end_layout
  526. \begin_layout Standard
  527. \begin_inset Flex TODO Note (inline)
  528. status open
  529. \begin_layout Plain Layout
  530. Possible citation for degree of generic variability:
  531. \begin_inset CommandInset href
  532. LatexCommand href
  533. target "https://www.ncbi.nlm.nih.gov/pubmed/22424236?dopt=Abstract"
  534. \end_inset
  535. \end_layout
  536. \end_inset
  537. \end_layout
  538. \begin_layout Standard
  539. \begin_inset Flex TODO Note (inline)
  540. status open
  541. \begin_layout Plain Layout
  542. How much mechanistic detail is needed here? My work doesn't really go into
  543. specific rejection mechanisms, so I think it's best to keep it basic.
  544. \end_layout
  545. \end_inset
  546. \end_layout
  547. \begin_layout Standard
  548. Because an allograft comes from a different person, it is genetically distinct
  549. from the rest of the recipient's body.
  550. Some genetic variants occur in protein coding regions and affect the polypeptid
  551. e sequences encoded by the affected genes, resulting in protein products
  552. that differ from the equivalent proteins produced by the graft recipient's
  553. own tissue.
  554. As a result, without intervention, the recipient's immune system will eventuall
  555. y identify the graft as foreign tissue and begin attacking it, eventually
  556. resulting in failure and death of the graft, a process referred to as transplan
  557. t rejection.
  558. Rejection is the most significant challenge to the long-term health and
  559. survival of an allograft [CITE?].
  560. Like any adaptive immune response, graft rejection generally occurs via
  561. two broad mechanisms: cellular immunity, in which CD8+ T-cells recognizing
  562. graft-specific antigens induce apoptosis in the graft cells; and humoral
  563. immunity, in which B-cells produce antibodies that bind to graft proteins
  564. and direct an immune response against the graft [CITE?].
  565. In either case, rejection shows most of the typical hallmarks of an adaptive
  566. immune response, in particular mediation by CD4+ T-cells and formation
  567. of immune memory.
  568. \end_layout
  569. \begin_layout Subsubsection
  570. Diagnosis and treatment of allograft rejection is a major challenge
  571. \end_layout
  572. \begin_layout Standard
  573. To prevent rejection, allograft recipients are treated with immune suppressive
  574. drugs [CITE?].
  575. The goal is to achieve sufficient suppression of the immune system to prevent
  576. rejection of the graft without compromising the ability of the immune system
  577. to raise a normal response against infection.
  578. As such, a delicate balance must be struck: insufficient immune suppression
  579. may lead to rejection and ultimately loss of the graft; excessive suppression
  580. leaves the patient vulnerable to life-threatening opportunistic infections.
  581. Because every patient is different, immune suppression must be tailored
  582. for each patient.
  583. Furthermore, immune suppression must be tuned over time, as the immune
  584. system's activity is not static, nor is it held in a steady state [CITE?].
  585. In order to properly adjust the dosage of immune suppression drugs, it
  586. is necessary to monitor the health of the transplant and increase the dosage
  587. if evidence of rejection is observed.
  588. \end_layout
  589. \begin_layout Standard
  590. However, diagnosis of rejection is a significant challenge.
  591. Early diagnosis is essential in order to step up immune suppression before
  592. the immune system damages the graft beyond recovery [CITE?].
  593. The current gold standard test for graft rejection is a tissue biopsy,
  594. examined for visible signs of rejection by a trained histologist [CITE?].
  595. When a patient shows symptoms of possible rejection, a
  596. \begin_inset Quotes eld
  597. \end_inset
  598. for cause
  599. \begin_inset Quotes erd
  600. \end_inset
  601. biopsy is performed to confirm the diagnosis, and immune suppression is
  602. adjusted as necessary.
  603. However, in many cases, the early stages of rejection are asymptomatic,
  604. known as
  605. \begin_inset Quotes eld
  606. \end_inset
  607. sub-clinical
  608. \begin_inset Quotes erd
  609. \end_inset
  610. rejection [CITE?].
  611. In light of this, is is now common to perform
  612. \begin_inset Quotes eld
  613. \end_inset
  614. protocol biopsies
  615. \begin_inset Quotes erd
  616. \end_inset
  617. at specific times after transplantation of a graft, even if no symptoms
  618. of rejection are apparent, in addition to
  619. \begin_inset Quotes eld
  620. \end_inset
  621. for cause
  622. \begin_inset Quotes erd
  623. \end_inset
  624. biopsies
  625. \begin_inset CommandInset citation
  626. LatexCommand cite
  627. key "Wilkinson2006"
  628. literal "false"
  629. \end_inset
  630. .
  631. \end_layout
  632. \begin_layout Standard
  633. However, biopsies have a number of downsides that limit their effectiveness
  634. as a diagnostic tool.
  635. First, the need for manual inspection by a histologist means that diagnosis
  636. is subject to the biases of the particular histologist examining the biopsy
  637. [CITE?].
  638. In marginal cases, two different histologists may give two different diagnoses
  639. to the same biopsy.
  640. Second, a biopsy can only evaluate if rejection is occurring in the section
  641. of the graft from which the tissue was extracted.
  642. If rejection is localized to one section of the graft and the tissue is
  643. extracted from a different section, a false negative diagnosis may result.
  644. Most importantly, extraction of tissue from a graft is invasive and is
  645. treated as an injury by the body, which results in inflammation that in
  646. turn promotes increased immune system activity [CITE?].
  647. Hence, the invasiveness of biopsies severely limits the frequency with
  648. which they can safely be performed.
  649. Typically, protocol biopsies are not scheduled more than about once per
  650. month
  651. \begin_inset CommandInset citation
  652. LatexCommand cite
  653. key "Wilkinson2006"
  654. literal "false"
  655. \end_inset
  656. .
  657. A less invasive diagnostic test for rejection would bring manifold benefits.
  658. Such a test would enable more frequent testing and therefore earlier detection
  659. of rejection events.
  660. In addition, having a larger pool of historical data for a given patient
  661. would make it easier to evaluate when a given test is outside the normal
  662. parameters for that specific patient, rather than relying on normal ranges
  663. for the population as a whole.
  664. Lastly, the accumulated data from more frequent tests would be a boon to
  665. the transplant research community.
  666. Beyond simply providing more data overall, the better time granularity
  667. of the tests will enable studying the progression of a rejection event
  668. on the scale of days to weeks, rather than months.
  669. \end_layout
  670. \begin_layout Subsubsection
  671. Memory cells are resistant to immune suppression
  672. \end_layout
  673. \begin_layout Standard
  674. One of the defining features of the adaptive immune system is immune memory:
  675. the ability of the immune system to recognize a previously encountered
  676. foreign antigen and respond more quickly and more strongly to that antigen
  677. in subsequent encounters.
  678. When the immune system first encounters a new antigen, the lymphocytes
  679. that respond are known as naïve cells – T-cells and B-cells that have never
  680. detected their target antigens before.
  681. Once activated by their specific antigen presented by an antigen-presenting
  682. cell in the proper co-stimulatory context, naïve cells differentiate into
  683. effector cells that carry out their respective functions in targeting and
  684. destroying the source of the foreign antigen.
  685. The requirement for co-stimulation is an important feature of naïve cells
  686. that limits
  687. \begin_inset Quotes eld
  688. \end_inset
  689. false positive
  690. \begin_inset Quotes erd
  691. \end_inset
  692. immune responses, because antigen-presenting cells usually only express
  693. the proper co-stimulation after detecting evidence of an infection, such
  694. as the presence of common bacterial cell components or inflamed tissue.
  695. Most effector cells die after the foreign antigen is cleared, since they
  696. are no longer needed, but some remain and differentiate into memory cells.
  697. Like naïve cells, memory cells respond to detection of their specific antigen
  698. by differentiating into effector cells, ready to fight an infection.
  699. However, unlike naïve cells, memory cells do not require the same degree
  700. of co-stimulatory signaling for activation, and once activated, they proliferat
  701. e and differentiate into effector cells more quickly than naïve cells do.
  702. \end_layout
  703. \begin_layout Standard
  704. In the context of a pathogenic infection, immune memory is a major advantage,
  705. allowing an organism to rapidly fight off a previously encountered pathogen
  706. much more quickly and effectively than the first time it was encountered.
  707. However, if effector cells that recognize an antigen from an allograft
  708. are allowed to differentiate into memory cells, preventing rejection of
  709. the graft becomes much more difficult.
  710. Many immune suppression drugs work by interfering with the co-stimulation
  711. that naïve cells require in order to mount an immune response [CITE?].
  712. Since memory cells do not require this co-stimulation, these drugs are
  713. not effective at suppressing an immune response that is mediated by memory
  714. cells.
  715. Secondly, because memory cells are able to mount a stronger and faster
  716. response to an antigen, all else being equal they require stronger immune
  717. suppression than naïve cells to prevent an immune response.
  718. However, immune suppression affects the entire immune system, not just
  719. cells recognizing a specific antigen, so increasing the dosage of immune
  720. suppression drugs also increases the risk of complications from a compromised
  721. immune system, such as opportunistic infections.
  722. While the differences in cell surface markers between naïve and memory
  723. cells have been fairly well characterized, the internal regulatory mechanisms
  724. that allow memory cells to respond more quickly and without co-stimulation
  725. are still poorly understood.
  726. In order to develop methods of immune suppression that either prevent the
  727. formation of memory cells or work more effectively against memory cells,
  728. we require a more complete understanding of the mechanisms of immune memory
  729. formation and regulation.
  730. \end_layout
  731. \begin_layout Standard
  732. \begin_inset Flex TODO Note (inline)
  733. status open
  734. \begin_layout Plain Layout
  735. Some kind of transition into bioinformatics would be good here
  736. \end_layout
  737. \end_inset
  738. \end_layout
  739. \begin_layout Subsection
  740. Overview of bioinformatic analysis methods
  741. \end_layout
  742. \begin_layout Standard
  743. \begin_inset Flex TODO Note (inline)
  744. status open
  745. \begin_layout Plain Layout
  746. Also cite somewhere: R, Bioconductor
  747. \end_layout
  748. \end_inset
  749. \end_layout
  750. \begin_layout Standard
  751. The studies presented in this work all involve the analysis of high-throughput
  752. genomic and epigenomic data.
  753. These data present many unique analysis challenges, and a wide array of
  754. software tools are available to analyze them.
  755. This section presents an overview of the methods used, including what problems
  756. they solve, what assumptions they make, and a basic description of how
  757. they work.
  758. \end_layout
  759. \begin_layout Subsubsection
  760. \begin_inset Flex Code
  761. status open
  762. \begin_layout Plain Layout
  763. Limma
  764. \end_layout
  765. \end_inset
  766. : The standard linear modeling framework for genomics
  767. \end_layout
  768. \begin_layout Standard
  769. Linear models are a generalization of the
  770. \begin_inset Formula $t$
  771. \end_inset
  772. -test and ANOVA to arbitrarily complex experimental designs
  773. \begin_inset CommandInset citation
  774. LatexCommand cite
  775. key "chambers:1992"
  776. literal "false"
  777. \end_inset
  778. .
  779. In a typical linear model, there is one dependent variable observation
  780. per sample and a large number of samples.
  781. For example, in a linear model of height as a function of age and sex,
  782. there is one height measurement per person.
  783. However, when analyzing genomic data, each sample consists of observations
  784. of thousands of dependent variables.
  785. For example, in a
  786. \begin_inset Flex Glossary Term
  787. status open
  788. \begin_layout Plain Layout
  789. RNA-seq
  790. \end_layout
  791. \end_inset
  792. \begin_inset CommandInset nomenclature
  793. LatexCommand nomenclature
  794. symbol "RNA-seq"
  795. description "High-throughput RNA sequencing"
  796. literal "false"
  797. \end_inset
  798. experiment, the dependent variables may be the count of
  799. \begin_inset Flex Glossary Term
  800. status open
  801. \begin_layout Plain Layout
  802. RNA-seq
  803. \end_layout
  804. \end_inset
  805. reads for each annotated gene.
  806. In abstract terms, each dependent variable being measured is referred to
  807. as a feature.
  808. The simplest approach to analyzing such data would be to fit the same model
  809. independently to each feature.
  810. However, this is undesirable for most genomics data sets.
  811. Genomics assays like high-throughput sequencing are expensive, and often
  812. the process of generating the samples is also quite expensive and time-consumin
  813. g.
  814. This expense limits the sample sizes typically employed in genomics experiments
  815. , and as a result the statistical power of the linear model for each individual
  816. feature is likewise limited.
  817. However, because thousands of features from the same samples are analyzed
  818. together, there is an opportunity to improve the statistical power of the
  819. analysis by exploiting shared patterns of variation across features.
  820. This is the core feature of
  821. \begin_inset Flex Code
  822. status open
  823. \begin_layout Plain Layout
  824. limma
  825. \end_layout
  826. \end_inset
  827. , a linear modeling framework designed for genomic data.
  828. \begin_inset Flex Code
  829. status open
  830. \begin_layout Plain Layout
  831. Limma
  832. \end_layout
  833. \end_inset
  834. is typically used to analyze expression microarray data, and more recently
  835. \begin_inset Flex Glossary Term
  836. status open
  837. \begin_layout Plain Layout
  838. RNA-seq
  839. \end_layout
  840. \end_inset
  841. data, but it can also be used to analyze any other data for which linear
  842. modeling is appropriate.
  843. \end_layout
  844. \begin_layout Standard
  845. The central challenge when fitting a linear model is to estimate the variance
  846. of the data accurately.
  847. Out of all parameters required to evaluate statistical significance of
  848. an effect, the variance is the most difficult to estimate when sample sizes
  849. are small.
  850. A single shared variance could be estimated for all of the features together,
  851. and this estimate would be very stable, in contrast to the individual feature
  852. variance estimates.
  853. However, this would require the assumption that every feature is equally
  854. variable, which is known to be false for most genomic data sets.
  855. \begin_inset Flex Code
  856. status open
  857. \begin_layout Plain Layout
  858. limma
  859. \end_layout
  860. \end_inset
  861. offers a compromise between these two extremes by using a method called
  862. empirical Bayes moderation to
  863. \begin_inset Quotes eld
  864. \end_inset
  865. squeeze
  866. \begin_inset Quotes erd
  867. \end_inset
  868. the distribution of estimated variances toward a single common value that
  869. represents the variance of an average feature in the data
  870. \begin_inset CommandInset citation
  871. LatexCommand cite
  872. key "Smyth2004"
  873. literal "false"
  874. \end_inset
  875. .
  876. While the individual feature variance estimates are not stable, the common
  877. variance estimate for the entire data set is quite stable, so using a combinati
  878. on of the two yields a variance estimate for each feature with greater precision
  879. than the individual feature variances.
  880. The trade-off for this improvement is that squeezing each estimated variance
  881. toward the common value introduces some bias – the variance will be underestima
  882. ted for features with high variance and overestimated for features with
  883. low variance.
  884. Essentially,
  885. \begin_inset Flex Code
  886. status open
  887. \begin_layout Plain Layout
  888. limma
  889. \end_layout
  890. \end_inset
  891. assumes that extreme variances are less common than variances close to
  892. the common value.
  893. The variance estimates from this empirical Bayes procedure are shown empiricall
  894. y to yield greater statistical power than either the individual feature
  895. variances or the single common value.
  896. \end_layout
  897. \begin_layout Standard
  898. On top of this core framework,
  899. \begin_inset Flex Code
  900. status open
  901. \begin_layout Plain Layout
  902. limma
  903. \end_layout
  904. \end_inset
  905. also implements many other enhancements that, further relax the assumptions
  906. of the model and extend the scope of what kinds of data it can analyze.
  907. Instead of squeezing toward a single common variance value,
  908. \begin_inset Flex Code
  909. status open
  910. \begin_layout Plain Layout
  911. limma
  912. \end_layout
  913. \end_inset
  914. can model the common variance as a function of a covariate, such as average
  915. expression
  916. \begin_inset CommandInset citation
  917. LatexCommand cite
  918. key "Law2013"
  919. literal "false"
  920. \end_inset
  921. .
  922. This is essential for
  923. \begin_inset Flex Glossary Term
  924. status open
  925. \begin_layout Plain Layout
  926. RNA-seq
  927. \end_layout
  928. \end_inset
  929. data, where higher gene counts yield more precise expression measurements
  930. and therefore smaller variances than low-count genes.
  931. While linear models typically assume that all samples have equal variance,
  932. \begin_inset Flex Code
  933. status open
  934. \begin_layout Plain Layout
  935. limma
  936. \end_layout
  937. \end_inset
  938. is able to relax this assumption by identifying and down-weighting samples
  939. that diverge more strongly from the linear model across many features
  940. \begin_inset CommandInset citation
  941. LatexCommand cite
  942. key "Ritchie2006,Liu2015"
  943. literal "false"
  944. \end_inset
  945. .
  946. In addition,
  947. \begin_inset Flex Code
  948. status open
  949. \begin_layout Plain Layout
  950. limma
  951. \end_layout
  952. \end_inset
  953. is also able to fit simple mixed models incorporating one random effect
  954. in addition to the fixed effects represented by an ordinary linear model
  955. \begin_inset CommandInset citation
  956. LatexCommand cite
  957. key "Smyth2005a"
  958. literal "false"
  959. \end_inset
  960. .
  961. Once again,
  962. \begin_inset Flex Code
  963. status open
  964. \begin_layout Plain Layout
  965. limma
  966. \end_layout
  967. \end_inset
  968. shares information between features to obtain a robust estimate for the
  969. random effect correlation.
  970. \end_layout
  971. \begin_layout Subsubsection
  972. \begin_inset Flex Code
  973. status open
  974. \begin_layout Plain Layout
  975. edgeR
  976. \end_layout
  977. \end_inset
  978. provides
  979. \begin_inset Flex Code
  980. status open
  981. \begin_layout Plain Layout
  982. limma
  983. \end_layout
  984. \end_inset
  985. -like analysis features for count data
  986. \end_layout
  987. \begin_layout Standard
  988. Although
  989. \begin_inset Flex Code
  990. status open
  991. \begin_layout Plain Layout
  992. limma
  993. \end_layout
  994. \end_inset
  995. can be applied to read counts from
  996. \begin_inset Flex Glossary Term
  997. status open
  998. \begin_layout Plain Layout
  999. RNA-seq
  1000. \end_layout
  1001. \end_inset
  1002. data, it is less suitable for counts from
  1003. \begin_inset Flex Glossary Term
  1004. status open
  1005. \begin_layout Plain Layout
  1006. ChIP-seq
  1007. \end_layout
  1008. \end_inset
  1009. \begin_inset CommandInset nomenclature
  1010. LatexCommand nomenclature
  1011. symbol "ChIP-seq"
  1012. description "Chromatin immunoprecipitation followed by high-throughput DNA sequencing"
  1013. literal "false"
  1014. \end_inset
  1015. , which tend to be much smaller and therefore violate the assumption of
  1016. a normal distribution more severely.
  1017. For all count-based data, the
  1018. \begin_inset Flex Code
  1019. status open
  1020. \begin_layout Plain Layout
  1021. edgeR
  1022. \end_layout
  1023. \end_inset
  1024. package works similarly to
  1025. \begin_inset Flex Code
  1026. status open
  1027. \begin_layout Plain Layout
  1028. limma
  1029. \end_layout
  1030. \end_inset
  1031. , but uses a
  1032. \begin_inset Flex Glossary Term
  1033. status open
  1034. \begin_layout Plain Layout
  1035. GLM
  1036. \end_layout
  1037. \end_inset
  1038. \begin_inset CommandInset nomenclature
  1039. LatexCommand nomenclature
  1040. symbol "GLM"
  1041. description "generalized linear model"
  1042. literal "false"
  1043. \end_inset
  1044. instead of a linear model.
  1045. Relative to a linear model, a
  1046. \begin_inset Flex Glossary Term
  1047. status open
  1048. \begin_layout Plain Layout
  1049. GLM
  1050. \end_layout
  1051. \end_inset
  1052. gains flexibility by relaxing several assumptions, the most important of
  1053. which is the assumption of normally distributed errors.
  1054. This allows the
  1055. \begin_inset Flex Glossary Term
  1056. status open
  1057. \begin_layout Plain Layout
  1058. GLM
  1059. \end_layout
  1060. \end_inset
  1061. in
  1062. \begin_inset Flex Code
  1063. status open
  1064. \begin_layout Plain Layout
  1065. edgeR
  1066. \end_layout
  1067. \end_inset
  1068. to model the counts directly using a
  1069. \begin_inset Flex Glossary Term
  1070. status open
  1071. \begin_layout Plain Layout
  1072. NB
  1073. \end_layout
  1074. \end_inset
  1075. \begin_inset CommandInset nomenclature
  1076. LatexCommand nomenclature
  1077. symbol "NB"
  1078. description "negative binomial"
  1079. literal "false"
  1080. \end_inset
  1081. distribution rather than modeling the normalized log counts using a normal
  1082. distribution
  1083. \begin_inset CommandInset citation
  1084. LatexCommand cite
  1085. key "Chen2014,McCarthy2012,Robinson2010a"
  1086. literal "false"
  1087. \end_inset
  1088. .
  1089. The
  1090. \begin_inset Flex Glossary Term
  1091. status open
  1092. \begin_layout Plain Layout
  1093. NB
  1094. \end_layout
  1095. \end_inset
  1096. is a good fit for count data because it can be derived as a gamma-distributed
  1097. mixture of Poisson distributions.
  1098. The Poisson distribution accurately represents the distribution of counts
  1099. expected for a given gene abundance, and the gamma distribution is then
  1100. used to represent the variation in gene abundance between biological replicates.
  1101. For this reason, the square root of the dispersion parameter of the
  1102. \begin_inset Flex Glossary Term
  1103. status open
  1104. \begin_layout Plain Layout
  1105. NB
  1106. \end_layout
  1107. \end_inset
  1108. is sometimes referred to as the
  1109. \begin_inset Flex Glossary Term
  1110. status open
  1111. \begin_layout Plain Layout
  1112. BCV
  1113. \end_layout
  1114. \end_inset
  1115. , since it represents the variability that was present in the samples prior
  1116. to the Poisson
  1117. \begin_inset Quotes eld
  1118. \end_inset
  1119. noise
  1120. \begin_inset Quotes erd
  1121. \end_inset
  1122. that was generated by the random sampling of reads in proportion to feature
  1123. abundances.
  1124. The choice of a gamma distribution is arbitrary and motivated by mathematical
  1125. convenience, since a gamma-Poisson mixture yields the numerically tractable
  1126. \begin_inset Flex Glossary Term
  1127. status open
  1128. \begin_layout Plain Layout
  1129. NB
  1130. \end_layout
  1131. \end_inset
  1132. distribution.
  1133. Thus,
  1134. \begin_inset Flex Code
  1135. status open
  1136. \begin_layout Plain Layout
  1137. edgeR
  1138. \end_layout
  1139. \end_inset
  1140. assumes
  1141. \emph on
  1142. a prioi
  1143. \emph default
  1144. that the variation in abundances between replicates follows a gamma distribution.
  1145. For differential abundance testing,
  1146. \begin_inset Flex Code
  1147. status open
  1148. \begin_layout Plain Layout
  1149. edgeR
  1150. \end_layout
  1151. \end_inset
  1152. offers a likelihood ratio test, but more recently recommends a quasi-likelihood
  1153. test that properly factors the uncertainty in variance estimation into
  1154. the statistical significance for each feature
  1155. \begin_inset CommandInset citation
  1156. LatexCommand cite
  1157. key "Lund2012"
  1158. literal "false"
  1159. \end_inset
  1160. .
  1161. \end_layout
  1162. \begin_layout Subsubsection
  1163. ChIP-seq Peak calling
  1164. \end_layout
  1165. \begin_layout Standard
  1166. Unlike
  1167. \begin_inset Flex Glossary Term
  1168. status open
  1169. \begin_layout Plain Layout
  1170. RNA-seq
  1171. \end_layout
  1172. \end_inset
  1173. data, in which gene annotations provide a well-defined set of discrete
  1174. genomic regions in which to count reads,
  1175. \begin_inset Flex Glossary Term
  1176. status open
  1177. \begin_layout Plain Layout
  1178. ChIP-seq
  1179. \end_layout
  1180. \end_inset
  1181. reads can potentially occur anywhere in the genome.
  1182. However, most genome regions will not contain significant
  1183. \begin_inset Flex Glossary Term
  1184. status open
  1185. \begin_layout Plain Layout
  1186. ChIP-seq
  1187. \end_layout
  1188. \end_inset
  1189. read coverage, and analyzing every position in the entire genome is statistical
  1190. ly and computationally infeasible, so it is necessary to identify regions
  1191. of interest inside which
  1192. \begin_inset Flex Glossary Term
  1193. status open
  1194. \begin_layout Plain Layout
  1195. ChIP-seq
  1196. \end_layout
  1197. \end_inset
  1198. reads will be counted and analyzed.
  1199. One option is to define a set of interesting regions
  1200. \emph on
  1201. a priori
  1202. \emph default
  1203. , for example by defining a promoter region for each annotated gene.
  1204. However, it is also possible to use the
  1205. \begin_inset Flex Glossary Term
  1206. status open
  1207. \begin_layout Plain Layout
  1208. ChIP-seq
  1209. \end_layout
  1210. \end_inset
  1211. data itself to identify regions with
  1212. \begin_inset Flex Glossary Term
  1213. status open
  1214. \begin_layout Plain Layout
  1215. ChIP-seq
  1216. \end_layout
  1217. \end_inset
  1218. read coverage significantly above the background level, known as peaks.
  1219. \end_layout
  1220. \begin_layout Standard
  1221. There are generally two kinds of peaks that can be identified: narrow peaks
  1222. and broadly enriched regions.
  1223. Proteins like transcription factors that bind specific sites in the genome
  1224. typically show most of their
  1225. \begin_inset Flex Glossary Term
  1226. status open
  1227. \begin_layout Plain Layout
  1228. ChIP-seq
  1229. \end_layout
  1230. \end_inset
  1231. read coverage at these specific sites and very little coverage anywhere
  1232. else.
  1233. Because the footprint of the protein is consistent wherever it binds, each
  1234. peak has a consistent width, typically tens to hundreds of base pairs,
  1235. representing the length of DNA that it binds to.
  1236. Algorithms like
  1237. \begin_inset Flex Glossary Term
  1238. status open
  1239. \begin_layout Plain Layout
  1240. MACS
  1241. \end_layout
  1242. \end_inset
  1243. \begin_inset CommandInset nomenclature
  1244. LatexCommand nomenclature
  1245. symbol "MACS"
  1246. description "Model-based Analysis of ChIP-seq"
  1247. literal "false"
  1248. \end_inset
  1249. exploit this pattern to identify specific loci at which such
  1250. \begin_inset Quotes eld
  1251. \end_inset
  1252. narrow peaks
  1253. \begin_inset Quotes erd
  1254. \end_inset
  1255. occur by looking for the characteristic peak shape in the
  1256. \begin_inset Flex Glossary Term
  1257. status open
  1258. \begin_layout Plain Layout
  1259. ChIP-seq
  1260. \end_layout
  1261. \end_inset
  1262. coverage rising above the surrounding background coverage
  1263. \begin_inset CommandInset citation
  1264. LatexCommand cite
  1265. key "Zhang2008"
  1266. literal "false"
  1267. \end_inset
  1268. .
  1269. In contrast, some proteins, chief among them histones, do not bind only
  1270. at a small number of specific sites, but rather bind potentially almost
  1271. everywhere in the entire genome.
  1272. When looking at histone marks, adjacent histones tend to be similarly marked,
  1273. and a given mark may be present on an arbitrary number of consecutive histones
  1274. along the genome.
  1275. Hence, there is no consistent
  1276. \begin_inset Quotes eld
  1277. \end_inset
  1278. footprint size
  1279. \begin_inset Quotes erd
  1280. \end_inset
  1281. for
  1282. \begin_inset Flex Glossary Term
  1283. status open
  1284. \begin_layout Plain Layout
  1285. ChIP-seq
  1286. \end_layout
  1287. \end_inset
  1288. peaks based on histone marks, and peaks typically span many histones.
  1289. Hence, typical peaks span many hundreds or even thousands of base pairs.
  1290. Instead of identifying specific loci of strong enrichment, algorithms like
  1291. \begin_inset Flex Glossary Term
  1292. status open
  1293. \begin_layout Plain Layout
  1294. SICER
  1295. \end_layout
  1296. \end_inset
  1297. \begin_inset CommandInset nomenclature
  1298. LatexCommand nomenclature
  1299. symbol "SICER"
  1300. description "Spatial Clustering for Identification of ChIP-Enriched Regions"
  1301. literal "false"
  1302. \end_inset
  1303. assume that peaks are represented in the
  1304. \begin_inset Flex Glossary Term
  1305. status open
  1306. \begin_layout Plain Layout
  1307. ChIP-seq
  1308. \end_layout
  1309. \end_inset
  1310. data by modest enrichment above background occurring across broad regions,
  1311. and they attempt to identify the extent of those regions
  1312. \begin_inset CommandInset citation
  1313. LatexCommand cite
  1314. key "Zang2009"
  1315. literal "false"
  1316. \end_inset
  1317. .
  1318. In all cases, better results are obtained if the local background coverage
  1319. level can be estimated from
  1320. \begin_inset Flex Glossary Term
  1321. status open
  1322. \begin_layout Plain Layout
  1323. ChIP-seq
  1324. \end_layout
  1325. \end_inset
  1326. input samples, since various biases can result in uneven background coverage.
  1327. \end_layout
  1328. \begin_layout Standard
  1329. Regardless of the type of peak identified, it is important to identify peaks
  1330. that occur consistently across biological replicates.
  1331. The
  1332. \begin_inset Flex Glossary Term
  1333. status open
  1334. \begin_layout Plain Layout
  1335. ENCODE
  1336. \end_layout
  1337. \end_inset
  1338. \begin_inset CommandInset nomenclature
  1339. LatexCommand nomenclature
  1340. symbol "ENCODE"
  1341. description "Encyclopedia Of DNA Elements"
  1342. literal "false"
  1343. \end_inset
  1344. project has developed a method called
  1345. \begin_inset Flex Glossary Term
  1346. status open
  1347. \begin_layout Plain Layout
  1348. IDR
  1349. \end_layout
  1350. \end_inset
  1351. \begin_inset CommandInset nomenclature
  1352. LatexCommand nomenclature
  1353. symbol "IDR"
  1354. description "irreproducible discovery rate"
  1355. literal "false"
  1356. \end_inset
  1357. for this purpose
  1358. \begin_inset CommandInset citation
  1359. LatexCommand cite
  1360. key "Li2006"
  1361. literal "false"
  1362. \end_inset
  1363. .
  1364. The
  1365. \begin_inset Flex Glossary Term
  1366. status open
  1367. \begin_layout Plain Layout
  1368. IDR
  1369. \end_layout
  1370. \end_inset
  1371. is defined as the probability that a peak identified in one biological
  1372. replicate will
  1373. \emph on
  1374. not
  1375. \emph default
  1376. also be identified in a second replicate.
  1377. Where the more familiar false discovery rate measures the degree of corresponde
  1378. nce between a data-derived ranked list and the true list of significant
  1379. features,
  1380. \begin_inset Flex Glossary Term
  1381. status open
  1382. \begin_layout Plain Layout
  1383. IDR
  1384. \end_layout
  1385. \end_inset
  1386. instead measures the degree of correspondence between two ranked lists
  1387. derived from different data.
  1388. \begin_inset Flex Glossary Term
  1389. status open
  1390. \begin_layout Plain Layout
  1391. IDR
  1392. \end_layout
  1393. \end_inset
  1394. assumes that the highest-ranked features are
  1395. \begin_inset Quotes eld
  1396. \end_inset
  1397. signal
  1398. \begin_inset Quotes erd
  1399. \end_inset
  1400. peaks that tend to be listed in the same order in both lists, while the
  1401. lowest-ranked features are essentially noise peaks, listed in random order
  1402. with no correspondence between the lists.
  1403. \begin_inset Flex Glossary Term (Capital)
  1404. status open
  1405. \begin_layout Plain Layout
  1406. IDR
  1407. \end_layout
  1408. \end_inset
  1409. attempts to locate the
  1410. \begin_inset Quotes eld
  1411. \end_inset
  1412. crossover point
  1413. \begin_inset Quotes erd
  1414. \end_inset
  1415. between the signal and the noise by determining how far down the list the
  1416. correspondence between feature ranks breaks down.
  1417. \end_layout
  1418. \begin_layout Standard
  1419. In addition to other considerations, if called peaks are to be used as regions
  1420. of interest for differential abundance analysis, then care must be taken
  1421. to call peaks in a way that is blind to differential abundance between
  1422. experimental conditions, or else the statistical significance calculations
  1423. for differential abundance will overstate their confidence in the results.
  1424. The
  1425. \begin_inset Flex Code
  1426. status open
  1427. \begin_layout Plain Layout
  1428. csaw
  1429. \end_layout
  1430. \end_inset
  1431. package provides guidelines for calling peaks in this way: peaks are called
  1432. based on a combination of all
  1433. \begin_inset Flex Glossary Term
  1434. status open
  1435. \begin_layout Plain Layout
  1436. ChIP-seq
  1437. \end_layout
  1438. \end_inset
  1439. reads from all experimental conditions, so that the identified peaks are
  1440. based on the average abundance across all conditions, which is independent
  1441. of any differential abundance between conditions
  1442. \begin_inset CommandInset citation
  1443. LatexCommand cite
  1444. key "Lun2015a"
  1445. literal "false"
  1446. \end_inset
  1447. .
  1448. \end_layout
  1449. \begin_layout Subsubsection
  1450. Normalization of high-throughput data is non-trivial and application-dependent
  1451. \end_layout
  1452. \begin_layout Standard
  1453. High-throughput data sets invariably require some kind of normalization
  1454. before further analysis can be conducted.
  1455. In general, the goal of normalization is to remove effects in the data
  1456. that are caused by technical factors that have nothing to do with the biology
  1457. being studied.
  1458. \end_layout
  1459. \begin_layout Standard
  1460. For Affymetrix expression arrays, the standard normalization algorithm used
  1461. in most analyses is
  1462. \begin_inset Flex Glossary Term
  1463. status open
  1464. \begin_layout Plain Layout
  1465. RMA
  1466. \end_layout
  1467. \end_inset
  1468. \begin_inset CommandInset nomenclature
  1469. LatexCommand nomenclature
  1470. symbol "RMA"
  1471. description "robust multichip average"
  1472. literal "false"
  1473. \end_inset
  1474. \begin_inset CommandInset citation
  1475. LatexCommand cite
  1476. key "Irizarry2003a"
  1477. literal "false"
  1478. \end_inset
  1479. .
  1480. \begin_inset Flex Glossary Term
  1481. status open
  1482. \begin_layout Plain Layout
  1483. RMA
  1484. \end_layout
  1485. \end_inset
  1486. is designed with the assumption that some fraction of probes on each array
  1487. will be artifactual and takes advantage of the fact that each gene is represent
  1488. ed by multiple probes by implementing normalization and summarization steps
  1489. that are robust against outlier probes.
  1490. However,
  1491. \begin_inset Flex Glossary Term
  1492. status open
  1493. \begin_layout Plain Layout
  1494. RMA
  1495. \end_layout
  1496. \end_inset
  1497. uses the probe intensities of all arrays in the data set in the normalization
  1498. of each individual array, meaning that the normalized expression values
  1499. in each array depend on every array in the data set, and will necessarily
  1500. change each time an array is added or removed from the data set.
  1501. If this is undesirable,
  1502. \begin_inset Flex Glossary Term
  1503. status open
  1504. \begin_layout Plain Layout
  1505. fRMA
  1506. \end_layout
  1507. \end_inset
  1508. implements a variant of
  1509. \begin_inset Flex Glossary Term
  1510. status open
  1511. \begin_layout Plain Layout
  1512. RMA
  1513. \end_layout
  1514. \end_inset
  1515. where the relevant distributional parameters are learned from a large reference
  1516. set of diverse public array data sets and then
  1517. \begin_inset Quotes eld
  1518. \end_inset
  1519. frozen
  1520. \begin_inset Quotes erd
  1521. \end_inset
  1522. , so that each array is effectively normalized against this frozen reference
  1523. set rather than the other arrays in the data set under study [CITE].
  1524. Other array normalization methods considered include dChip,
  1525. \begin_inset Flex Glossary Term
  1526. status open
  1527. \begin_layout Plain Layout
  1528. GRSN
  1529. \end_layout
  1530. \end_inset
  1531. \begin_inset CommandInset nomenclature
  1532. LatexCommand nomenclature
  1533. symbol "GRSN"
  1534. description "global rank-invariant set normalization"
  1535. literal "false"
  1536. \end_inset
  1537. , and
  1538. \begin_inset Flex Glossary Term
  1539. status open
  1540. \begin_layout Plain Layout
  1541. SCAN
  1542. \end_layout
  1543. \end_inset
  1544. \begin_inset CommandInset nomenclature
  1545. LatexCommand nomenclature
  1546. symbol "SCAN"
  1547. description "Single-Channel Array Normalization"
  1548. literal "false"
  1549. \end_inset
  1550. \begin_inset CommandInset citation
  1551. LatexCommand cite
  1552. key "Li2001,Pelz2008,Piccolo2012"
  1553. literal "false"
  1554. \end_inset
  1555. .
  1556. \end_layout
  1557. \begin_layout Standard
  1558. In contrast, high-throughput sequencing data present very different normalizatio
  1559. n challenges.
  1560. The simplest case is
  1561. \begin_inset Flex Glossary Term
  1562. status open
  1563. \begin_layout Plain Layout
  1564. RNA-seq
  1565. \end_layout
  1566. \end_inset
  1567. in which read counts are obtained for a set of gene annotations, yielding
  1568. a matrix of counts with rows representing genes and columns representing
  1569. samples.
  1570. Because
  1571. \begin_inset Flex Glossary Term
  1572. status open
  1573. \begin_layout Plain Layout
  1574. RNA-seq
  1575. \end_layout
  1576. \end_inset
  1577. approximates a process of sampling from a population with replacement,
  1578. each gene's count is only interpretable as a fraction of the total reads
  1579. for that sample.
  1580. For that reason,
  1581. \begin_inset Flex Glossary Term
  1582. status open
  1583. \begin_layout Plain Layout
  1584. RNA-seq
  1585. \end_layout
  1586. \end_inset
  1587. abundances are often reported as
  1588. \begin_inset Flex Glossary Term
  1589. status open
  1590. \begin_layout Plain Layout
  1591. CPM
  1592. \end_layout
  1593. \end_inset
  1594. \begin_inset CommandInset nomenclature
  1595. LatexCommand nomenclature
  1596. symbol "CPM"
  1597. description "counts per million"
  1598. literal "false"
  1599. \end_inset
  1600. .
  1601. Furthermore, if the abundance of a single gene increases, then in order
  1602. for its fraction of the total reads to increase, all other genes' fractions
  1603. must decrease to accommodate it.
  1604. This effect is known as composition bias, and it is an artifact of the
  1605. read sampling process that has nothing to do with the biology of the samples
  1606. and must therefore be normalized out.
  1607. The most commonly used methods to normalize for composition bias in
  1608. \begin_inset Flex Glossary Term
  1609. status open
  1610. \begin_layout Plain Layout
  1611. RNA-seq
  1612. \end_layout
  1613. \end_inset
  1614. data seek to equalize the average gene abundance across samples, under
  1615. the assumption that the average gene is likely not changing
  1616. \begin_inset CommandInset citation
  1617. LatexCommand cite
  1618. key "Robinson2010,Anders2010"
  1619. literal "false"
  1620. \end_inset
  1621. .
  1622. \end_layout
  1623. \begin_layout Standard
  1624. In
  1625. \begin_inset Flex Glossary Term
  1626. status open
  1627. \begin_layout Plain Layout
  1628. ChIP-seq
  1629. \end_layout
  1630. \end_inset
  1631. data, normalization is not as straightforward.
  1632. The
  1633. \begin_inset Flex Code
  1634. status open
  1635. \begin_layout Plain Layout
  1636. csaw
  1637. \end_layout
  1638. \end_inset
  1639. package implements several different normalization strategies and provides
  1640. guidance on when to use each one
  1641. \begin_inset CommandInset citation
  1642. LatexCommand cite
  1643. key "Lun2015a"
  1644. literal "false"
  1645. \end_inset
  1646. .
  1647. Briefly, a typical
  1648. \begin_inset Flex Glossary Term
  1649. status open
  1650. \begin_layout Plain Layout
  1651. ChIP-seq
  1652. \end_layout
  1653. \end_inset
  1654. sample has a bimodal distribution of read counts: a low-abundance mode
  1655. representing background regions and a high-abundance mode representing
  1656. signal regions.
  1657. This offers two potential normalization targets: equalizing background
  1658. coverage or equalizing signal coverage.
  1659. If the experiment is well controlled and ChIP efficiency is known to be
  1660. consistent across all samples, then normalizing the background coverage
  1661. to be equal across all samples is a reasonable strategy.
  1662. If this is not a safe assumption, then the preferred strategy is to normalize
  1663. the signal regions in a way similar to
  1664. \begin_inset Flex Glossary Term
  1665. status open
  1666. \begin_layout Plain Layout
  1667. RNA-seq
  1668. \end_layout
  1669. \end_inset
  1670. data by assuming that the average signal region is not changing abundance
  1671. between samples.
  1672. Beyond this, if a
  1673. \begin_inset Flex Glossary Term
  1674. status open
  1675. \begin_layout Plain Layout
  1676. ChIP-seq
  1677. \end_layout
  1678. \end_inset
  1679. experiment has a more complicated structure that doesn't show the typical
  1680. bimodal count distribution, it may be necessary to implement a normalization
  1681. as a smooth function of abundance.
  1682. However, this strategy makes a much stronger assumption about the data:
  1683. that the average
  1684. \begin_inset Flex Glossary Term
  1685. status open
  1686. \begin_layout Plain Layout
  1687. logFC
  1688. \end_layout
  1689. \end_inset
  1690. is zero across all abundance levels.
  1691. Hence, the simpler scaling normalization based on background or signal
  1692. regions are generally preferred whenever possible.
  1693. \end_layout
  1694. \begin_layout Subsubsection
  1695. ComBat and SVA for correction of known and unknown batch effects
  1696. \end_layout
  1697. \begin_layout Standard
  1698. In addition to well-understood effects that can be easily normalized out,
  1699. a data set often contains confounding biological effects that must be accounted
  1700. for in the modeling step.
  1701. For instance, in an experiment with pre-treatment and post-treatment samples
  1702. of cells from several different donors, donor variability represents a
  1703. known batch effect.
  1704. The most straightforward correction for known batches is to estimate the
  1705. mean for each batch independently and subtract out the differences, so
  1706. that all batches have identical means for each feature.
  1707. However, as with variance estimation, estimating the differences in batch
  1708. means is not necessarily robust at the feature level, so the ComBat method
  1709. adds empirical Bayes squeezing of the batch mean differences toward a common
  1710. value, analogous to
  1711. \begin_inset Flex Code
  1712. status open
  1713. \begin_layout Plain Layout
  1714. limma
  1715. \end_layout
  1716. \end_inset
  1717. 's empirical Bayes squeezing of feature variance estimates
  1718. \begin_inset CommandInset citation
  1719. LatexCommand cite
  1720. key "Johnson2007"
  1721. literal "false"
  1722. \end_inset
  1723. .
  1724. Effectively, ComBat assumes that modest differences between batch means
  1725. are real batch effects, but extreme differences between batch means are
  1726. more likely to be the result of outlier observations that happen to line
  1727. up with the batches rather than a genuine batch effect.
  1728. The result is a batch correction that is more robust against outliers than
  1729. simple subtraction of mean differences subtraction.
  1730. \end_layout
  1731. \begin_layout Standard
  1732. In some data sets, unknown batch effects may be present due to inherent
  1733. variability in in the data, either caused by technical or biological effects.
  1734. Examples of unknown batch effects include variations in enrichment efficiency
  1735. between
  1736. \begin_inset Flex Glossary Term
  1737. status open
  1738. \begin_layout Plain Layout
  1739. ChIP-seq
  1740. \end_layout
  1741. \end_inset
  1742. samples, variations in populations of different cell types, and the effects
  1743. of uncontrolled environmental factors on gene expression in humans or live
  1744. animals.
  1745. In an ordinary linear model context, unknown batch effects cannot be inferred
  1746. and must be treated as random noise.
  1747. However, in high-throughput experiments, once again information can be
  1748. shared across features to identify patterns of un-modeled variation that
  1749. are repeated in many features.
  1750. One attractive strategy would be to perform
  1751. \begin_inset Flex Glossary Term
  1752. status open
  1753. \begin_layout Plain Layout
  1754. SVD
  1755. \end_layout
  1756. \end_inset
  1757. \begin_inset CommandInset nomenclature
  1758. LatexCommand nomenclature
  1759. symbol "SVD"
  1760. description "singular value decomposition"
  1761. literal "false"
  1762. \end_inset
  1763. on the matrix of linear model residuals (which contain all the un-modeled
  1764. variation in the data) and take the first few singular vectors as batch
  1765. effects.
  1766. While this can be effective, it makes the unreasonable assumption that
  1767. all batch effects are uncorrelated with any of the effects being modeled.
  1768. \begin_inset Flex Glossary Term
  1769. status open
  1770. \begin_layout Plain Layout
  1771. SVA
  1772. \end_layout
  1773. \end_inset
  1774. \begin_inset CommandInset nomenclature
  1775. LatexCommand nomenclature
  1776. symbol "SVA"
  1777. description "surrogate variable analysis"
  1778. literal "false"
  1779. \end_inset
  1780. starts with this approach, but takes some additional steps to identify
  1781. batch effects in the full data that are both highly correlated with the
  1782. singular vectors in the residuals and least correlated with the effects
  1783. of interest
  1784. \begin_inset CommandInset citation
  1785. LatexCommand cite
  1786. key "Leek2007"
  1787. literal "false"
  1788. \end_inset
  1789. .
  1790. Since the final batch effects are estimated from the full data, moderate
  1791. correlations between the batch effects and effects of interest are allowed,
  1792. which gives
  1793. \begin_inset Flex Glossary Term
  1794. status open
  1795. \begin_layout Plain Layout
  1796. SVA
  1797. \end_layout
  1798. \end_inset
  1799. much more freedom to estimate the true extent of the batch effects compared
  1800. to simple residual
  1801. \begin_inset Flex Glossary Term
  1802. status open
  1803. \begin_layout Plain Layout
  1804. SVD
  1805. \end_layout
  1806. \end_inset
  1807. .
  1808. Once the surrogate variables are estimated, they can be included as coefficient
  1809. s in the linear model in a similar fashion to known batch effects in order
  1810. to subtract out their effects on each feature's abundance.
  1811. \end_layout
  1812. \begin_layout Subsubsection
  1813. Factor analysis: PCA, MDS, MOFA
  1814. \end_layout
  1815. \begin_layout Standard
  1816. \begin_inset Flex TODO Note (inline)
  1817. status open
  1818. \begin_layout Plain Layout
  1819. Not sure if this merits a subsection here.
  1820. \end_layout
  1821. \end_inset
  1822. \end_layout
  1823. \begin_layout Itemize
  1824. Batch-corrected
  1825. \begin_inset Flex Glossary Term
  1826. status open
  1827. \begin_layout Plain Layout
  1828. PCA
  1829. \end_layout
  1830. \end_inset
  1831. is informative, but careful application is required to avoid bias
  1832. \end_layout
  1833. \begin_layout Section
  1834. Innovation
  1835. \end_layout
  1836. \begin_layout Standard
  1837. \begin_inset Flex TODO Note (inline)
  1838. status open
  1839. \begin_layout Plain Layout
  1840. Is this entire section redundant with the Approach sections of each chapter?
  1841. I'm not really sure what to write here.
  1842. \end_layout
  1843. \end_inset
  1844. \end_layout
  1845. \begin_layout Subsection
  1846. MSC infusion to improve transplant outcomes (prevent/delay rejection)
  1847. \end_layout
  1848. \begin_layout Standard
  1849. \begin_inset Flex TODO Note (inline)
  1850. status open
  1851. \begin_layout Plain Layout
  1852. Do I still talk about this? It's the motivation for chapter 4, but I don't
  1853. actually present any work related to MSCs.
  1854. \end_layout
  1855. \end_inset
  1856. \end_layout
  1857. \begin_layout Itemize
  1858. Demonstrated in mice, but not yet in primates
  1859. \end_layout
  1860. \begin_layout Itemize
  1861. Mechanism currently unknown, but MSC are known to be immune modulatory
  1862. \end_layout
  1863. \begin_layout Itemize
  1864. Characterize MSC response to interferon gamma
  1865. \end_layout
  1866. \begin_layout Itemize
  1867. IFN-g is thought to stimulate their function
  1868. \end_layout
  1869. \begin_layout Itemize
  1870. Test IFN-g treated MSC infusion as a therapy to delay graft rejection in
  1871. cynomolgus monkeys
  1872. \end_layout
  1873. \begin_layout Itemize
  1874. Monitor animals post-transplant using blood
  1875. \begin_inset Flex Glossary Term
  1876. status open
  1877. \begin_layout Plain Layout
  1878. RNA-seq
  1879. \end_layout
  1880. \end_inset
  1881. at serial time points
  1882. \end_layout
  1883. \begin_layout Subsection
  1884. Investigate dynamics of histone marks in CD4 T-cell activation and memory
  1885. \end_layout
  1886. \begin_layout Itemize
  1887. Previous studies have looked at single snapshots of histone marks
  1888. \end_layout
  1889. \begin_layout Itemize
  1890. Instead, look at changes in histone marks across activation and memory
  1891. \end_layout
  1892. \begin_layout Subsection
  1893. High-throughput sequencing and microarray technologies
  1894. \end_layout
  1895. \begin_layout Itemize
  1896. Powerful methods for assaying gene expression and epigenetics across entire
  1897. genomes
  1898. \end_layout
  1899. \begin_layout Itemize
  1900. Proper analysis requires finding and exploiting systematic genome-wide trends
  1901. \end_layout
  1902. \begin_layout Chapter
  1903. Reproducible genome-wide epigenetic analysis of H3K4 and H3K27 methylation
  1904. in naïve and memory CD4 T-cell activation
  1905. \end_layout
  1906. \begin_layout Standard
  1907. \begin_inset Flex TODO Note (inline)
  1908. status open
  1909. \begin_layout Plain Layout
  1910. Chapter author list: Me, Sarah, Dan
  1911. \end_layout
  1912. \end_inset
  1913. \end_layout
  1914. \begin_layout Standard
  1915. \begin_inset ERT
  1916. status collapsed
  1917. \begin_layout Plain Layout
  1918. \backslash
  1919. glsresetall
  1920. \end_layout
  1921. \end_inset
  1922. \end_layout
  1923. \begin_layout Standard
  1924. \begin_inset Flex TODO Note (inline)
  1925. status open
  1926. \begin_layout Plain Layout
  1927. Need better section titles throughout the entire chapter
  1928. \end_layout
  1929. \end_inset
  1930. \end_layout
  1931. \begin_layout Section
  1932. Approach
  1933. \end_layout
  1934. \begin_layout Standard
  1935. \begin_inset Flex TODO Note (inline)
  1936. status open
  1937. \begin_layout Plain Layout
  1938. Check on the exact correct way to write
  1939. \begin_inset Quotes eld
  1940. \end_inset
  1941. CD4 T-cell
  1942. \begin_inset Quotes erd
  1943. \end_inset
  1944. .
  1945. I think there might be a plus sign somewhere in there now? Also, maybe
  1946. figure out a reasonable way to abbreviate
  1947. \begin_inset Quotes eld
  1948. \end_inset
  1949. naïve CD4 T-cells
  1950. \begin_inset Quotes erd
  1951. \end_inset
  1952. and
  1953. \begin_inset Quotes eld
  1954. \end_inset
  1955. memory CD4 T-cells
  1956. \begin_inset Quotes erd
  1957. \end_inset
  1958. .
  1959. \end_layout
  1960. \end_inset
  1961. \end_layout
  1962. \begin_layout Standard
  1963. \begin_inset Flex TODO Note (inline)
  1964. status open
  1965. \begin_layout Plain Layout
  1966. Is it ok to just copy a bunch of citations from the intros to Sarah's papers?
  1967. That feels like cheating somehow.
  1968. \end_layout
  1969. \end_inset
  1970. \end_layout
  1971. \begin_layout Standard
  1972. CD4 T-cells are central to all adaptive immune responses, as well as immune
  1973. memory [CITE?].
  1974. After an infection is cleared, a subset of the naïve CD4 T-cells that responded
  1975. to that infection differentiate into memory CD4 T-cells, which are responsible
  1976. for responding to the same pathogen in the future.
  1977. Memory CD4 T-cells are functionally distinct, able to respond to an infection
  1978. more quickly and without the co-stimulation required by naïve CD4 T-cells.
  1979. However, the molecular mechanisms underlying this functional distinction
  1980. are not well-understood.
  1981. Epigenetic regulation via histone modification is thought to play an important
  1982. role, but while many studies have looked at static snapshots of histone
  1983. methylation in T-cells, few studies have looked at the dynamics of histone
  1984. regulation after T-cell activation, nor the differences in histone methylation
  1985. between naïve and memory T-cells.
  1986. H3K4me2, H3K4me3 and H3K27me3 are three histone marks thought to be major
  1987. epigenetic regulators of gene expression.
  1988. The goal of the present study is to investigate the role of these histone
  1989. marks in CD4 T-cell activation kinetics and memory differentiation.
  1990. In static snapshots, H3K4me2 and H3K4me3 are often observed in the promoters
  1991. of highly transcribed genes, while H3K27me3 is more often observed in promoters
  1992. of inactive genes with little to no transcription occurring.
  1993. As a result, the two H3K4 marks have been characterized as
  1994. \begin_inset Quotes eld
  1995. \end_inset
  1996. activating
  1997. \begin_inset Quotes erd
  1998. \end_inset
  1999. marks, while H3K27me3 has been characterized as
  2000. \begin_inset Quotes eld
  2001. \end_inset
  2002. deactivating
  2003. \begin_inset Quotes erd
  2004. \end_inset
  2005. .
  2006. Despite these characterizations, the actual causal relationship between
  2007. these histone modifications and gene transcription is complex and likely
  2008. involves positive and negative feedback loops between the two.
  2009. \end_layout
  2010. \begin_layout Standard
  2011. In order to investigate the relationship between gene expression and these
  2012. histone modifications in the context of naïve and memory CD4 T-cell activation,
  2013. a previously published data set of
  2014. \begin_inset Flex Glossary Term
  2015. status open
  2016. \begin_layout Plain Layout
  2017. RNA-seq
  2018. \end_layout
  2019. \end_inset
  2020. data and
  2021. \begin_inset Flex Glossary Term
  2022. status open
  2023. \begin_layout Plain Layout
  2024. ChIP-seq
  2025. \end_layout
  2026. \end_inset
  2027. data was re-analyzed using up-to-date methods designed to address the specific
  2028. analysis challenges posed by this data set.
  2029. The data set contains naïve and memory CD4 T-cell samples in a time course
  2030. before and after activation.
  2031. Like the original analysis, this analysis looks at the dynamics of these
  2032. marks histone marks and compare them to gene expression dynamics at the
  2033. same time points during activation, as well as compare them between naïve
  2034. and memory cells, in hope of discovering evidence of new mechanistic details
  2035. in the interplay between them.
  2036. The original analysis of this data treated each gene promoter as a monolithic
  2037. unit and mostly assumed that
  2038. \begin_inset Flex Glossary Term
  2039. status open
  2040. \begin_layout Plain Layout
  2041. ChIP-seq
  2042. \end_layout
  2043. \end_inset
  2044. reads or peaks occurring anywhere within a promoter were equivalent, regardless
  2045. of where they occurred relative to the gene structure.
  2046. For an initial analysis of the data, this was a necessary simplifying assumptio
  2047. n.
  2048. The current analysis aims to relax this assumption, first by directly analyzing
  2049. \begin_inset Flex Glossary Term
  2050. status open
  2051. \begin_layout Plain Layout
  2052. ChIP-seq
  2053. \end_layout
  2054. \end_inset
  2055. peaks for differential modification, and second by taking a more granular
  2056. look at the
  2057. \begin_inset Flex Glossary Term
  2058. status open
  2059. \begin_layout Plain Layout
  2060. ChIP-seq
  2061. \end_layout
  2062. \end_inset
  2063. read coverage within promoter regions to ask whether the location of histone
  2064. modifications relative to the gene's
  2065. \begin_inset Flex Glossary Term
  2066. status open
  2067. \begin_layout Plain Layout
  2068. TSS
  2069. \end_layout
  2070. \end_inset
  2071. \begin_inset CommandInset nomenclature
  2072. LatexCommand nomenclature
  2073. symbol "TSS"
  2074. description "transcription start site"
  2075. literal "false"
  2076. \end_inset
  2077. is an important factor, as opposed to simple proximity.
  2078. \end_layout
  2079. \begin_layout Section
  2080. Methods
  2081. \end_layout
  2082. \begin_layout Standard
  2083. \begin_inset Flex TODO Note (inline)
  2084. status open
  2085. \begin_layout Plain Layout
  2086. Look up some more details from the papers (e.g.
  2087. activation method).
  2088. \end_layout
  2089. \end_inset
  2090. \end_layout
  2091. \begin_layout Standard
  2092. A reproducible workflow was written to analyze the raw
  2093. \begin_inset Flex Glossary Term
  2094. status open
  2095. \begin_layout Plain Layout
  2096. ChIP-seq
  2097. \end_layout
  2098. \end_inset
  2099. and
  2100. \begin_inset Flex Glossary Term
  2101. status open
  2102. \begin_layout Plain Layout
  2103. RNA-seq
  2104. \end_layout
  2105. \end_inset
  2106. data from previous studies
  2107. \begin_inset CommandInset citation
  2108. LatexCommand cite
  2109. key "gh-cd4-csaw,LaMere2016,LaMere2017"
  2110. literal "true"
  2111. \end_inset
  2112. .
  2113. Briefly, this data consists of
  2114. \begin_inset Flex Glossary Term
  2115. status open
  2116. \begin_layout Plain Layout
  2117. RNA-seq
  2118. \end_layout
  2119. \end_inset
  2120. and
  2121. \begin_inset Flex Glossary Term
  2122. status open
  2123. \begin_layout Plain Layout
  2124. ChIP-seq
  2125. \end_layout
  2126. \end_inset
  2127. from CD4 T-cells cultured from 4 donors.
  2128. From each donor, naïve and memory CD4 T-cells were isolated separately.
  2129. Then cultures of both cells were activated [how?], and samples were taken
  2130. at 4 time points: Day 0 (pre-activation), Day 1 (early activation), Day
  2131. 5 (peak activation), and Day 14 (post-activation).
  2132. For each combination of cell type and time point, RNA was isolated and
  2133. sequenced, and
  2134. \begin_inset Flex Glossary Term
  2135. status open
  2136. \begin_layout Plain Layout
  2137. ChIP-seq
  2138. \end_layout
  2139. \end_inset
  2140. was performed for each of 3 histone marks: H3K4me2, H3K4me3, and H3K27me3.
  2141. The
  2142. \begin_inset Flex Glossary Term
  2143. status open
  2144. \begin_layout Plain Layout
  2145. ChIP-seq
  2146. \end_layout
  2147. \end_inset
  2148. input DNA was also sequenced for each sample.
  2149. The result was 32 samples for each assay.
  2150. \end_layout
  2151. \begin_layout Subsection
  2152. RNA-seq differential expression analysis
  2153. \end_layout
  2154. \begin_layout Standard
  2155. \begin_inset Note Note
  2156. status collapsed
  2157. \begin_layout Plain Layout
  2158. \begin_inset Float figure
  2159. wide false
  2160. sideways false
  2161. status open
  2162. \begin_layout Plain Layout
  2163. \align center
  2164. \begin_inset Float figure
  2165. wide false
  2166. sideways false
  2167. status collapsed
  2168. \begin_layout Plain Layout
  2169. \align center
  2170. \begin_inset Graphics
  2171. filename graphics/CD4-csaw/rnaseq-compare/ensmebl-vs-entrez-star-CROP.png
  2172. lyxscale 25
  2173. width 35col%
  2174. groupId rna-comp-subfig
  2175. \end_inset
  2176. \end_layout
  2177. \begin_layout Plain Layout
  2178. \begin_inset Caption Standard
  2179. \begin_layout Plain Layout
  2180. STAR quantification, Entrez vs Ensembl gene annotation
  2181. \end_layout
  2182. \end_inset
  2183. \end_layout
  2184. \end_inset
  2185. \begin_inset space \qquad{}
  2186. \end_inset
  2187. \begin_inset Float figure
  2188. wide false
  2189. sideways false
  2190. status collapsed
  2191. \begin_layout Plain Layout
  2192. \align center
  2193. \begin_inset Graphics
  2194. filename graphics/CD4-csaw/rnaseq-compare/ensmebl-vs-entrez-shoal-CROP.png
  2195. lyxscale 25
  2196. width 35col%
  2197. groupId rna-comp-subfig
  2198. \end_inset
  2199. \end_layout
  2200. \begin_layout Plain Layout
  2201. \begin_inset Caption Standard
  2202. \begin_layout Plain Layout
  2203. Salmon+Shoal quantification, Entrez vs Ensembl gene annotation
  2204. \end_layout
  2205. \end_inset
  2206. \end_layout
  2207. \end_inset
  2208. \end_layout
  2209. \begin_layout Plain Layout
  2210. \align center
  2211. \begin_inset Float figure
  2212. wide false
  2213. sideways false
  2214. status collapsed
  2215. \begin_layout Plain Layout
  2216. \align center
  2217. \begin_inset Graphics
  2218. filename graphics/CD4-csaw/rnaseq-compare/star-vs-hisat2-CROP.png
  2219. lyxscale 25
  2220. width 35col%
  2221. groupId rna-comp-subfig
  2222. \end_inset
  2223. \end_layout
  2224. \begin_layout Plain Layout
  2225. \begin_inset Caption Standard
  2226. \begin_layout Plain Layout
  2227. STAR vs HISAT2 quantification, Ensembl gene annotation
  2228. \end_layout
  2229. \end_inset
  2230. \end_layout
  2231. \end_inset
  2232. \begin_inset space \qquad{}
  2233. \end_inset
  2234. \begin_inset Float figure
  2235. wide false
  2236. sideways false
  2237. status collapsed
  2238. \begin_layout Plain Layout
  2239. \align center
  2240. \begin_inset Graphics
  2241. filename graphics/CD4-csaw/rnaseq-compare/star-vs-salmon-CROP.png
  2242. lyxscale 25
  2243. width 35col%
  2244. groupId rna-comp-subfig
  2245. \end_inset
  2246. \end_layout
  2247. \begin_layout Plain Layout
  2248. \begin_inset Caption Standard
  2249. \begin_layout Plain Layout
  2250. Salmon vs STAR quantification, Ensembl gene annotation
  2251. \end_layout
  2252. \end_inset
  2253. \end_layout
  2254. \end_inset
  2255. \end_layout
  2256. \begin_layout Plain Layout
  2257. \align center
  2258. \begin_inset Float figure
  2259. wide false
  2260. sideways false
  2261. status collapsed
  2262. \begin_layout Plain Layout
  2263. \align center
  2264. \begin_inset Graphics
  2265. filename graphics/CD4-csaw/rnaseq-compare/salmon-vs-kallisto-CROP.png
  2266. lyxscale 25
  2267. width 35col%
  2268. groupId rna-comp-subfig
  2269. \end_inset
  2270. \end_layout
  2271. \begin_layout Plain Layout
  2272. \begin_inset Caption Standard
  2273. \begin_layout Plain Layout
  2274. Salmon vs Kallisto quantification, Ensembl gene annotation
  2275. \end_layout
  2276. \end_inset
  2277. \end_layout
  2278. \end_inset
  2279. \begin_inset space \qquad{}
  2280. \end_inset
  2281. \begin_inset Float figure
  2282. wide false
  2283. sideways false
  2284. status collapsed
  2285. \begin_layout Plain Layout
  2286. \align center
  2287. \begin_inset Graphics
  2288. filename graphics/CD4-csaw/rnaseq-compare/salmon-vs-shoal-CROP.png
  2289. lyxscale 25
  2290. width 35col%
  2291. groupId rna-comp-subfig
  2292. \end_inset
  2293. \end_layout
  2294. \begin_layout Plain Layout
  2295. \begin_inset Caption Standard
  2296. \begin_layout Plain Layout
  2297. Salmon+Shoal vs Salmon alone, Ensembl gene annotation
  2298. \end_layout
  2299. \end_inset
  2300. \end_layout
  2301. \end_inset
  2302. \end_layout
  2303. \begin_layout Plain Layout
  2304. \begin_inset Caption Standard
  2305. \begin_layout Plain Layout
  2306. \begin_inset CommandInset label
  2307. LatexCommand label
  2308. name "fig:RNA-norm-comp"
  2309. \end_inset
  2310. RNA-seq comparisons
  2311. \end_layout
  2312. \end_inset
  2313. \end_layout
  2314. \end_inset
  2315. \end_layout
  2316. \end_inset
  2317. \end_layout
  2318. \begin_layout Standard
  2319. Sequence reads were retrieved from the
  2320. \begin_inset Flex Glossary Term
  2321. status open
  2322. \begin_layout Plain Layout
  2323. SRA
  2324. \end_layout
  2325. \end_inset
  2326. \begin_inset CommandInset nomenclature
  2327. LatexCommand nomenclature
  2328. symbol "SRA"
  2329. description "Sequence Read Archive"
  2330. literal "false"
  2331. \end_inset
  2332. \begin_inset CommandInset citation
  2333. LatexCommand cite
  2334. key "Leinonen2011"
  2335. literal "false"
  2336. \end_inset
  2337. .
  2338. Five different alignment and quantification methods were tested for the
  2339. \begin_inset Flex Glossary Term
  2340. status open
  2341. \begin_layout Plain Layout
  2342. RNA-seq
  2343. \end_layout
  2344. \end_inset
  2345. data
  2346. \begin_inset CommandInset citation
  2347. LatexCommand cite
  2348. key "Dobin2012,Kim2019,Liao2014,Pimentel2016,Patro2017,gh-shoal,gh-hg38-ref"
  2349. literal "false"
  2350. \end_inset
  2351. .
  2352. Each quantification was tested with both Ensembl transcripts and UCSC known
  2353. gene annotations [CITE? Also which versions of each?].
  2354. Comparisons of downstream results from each combination of quantification
  2355. method and reference revealed that all quantifications gave broadly similar
  2356. results for most genes, so shoal with the Ensembl annotation was chosen
  2357. as the method theoretically most likely to partially mitigate some of the
  2358. batch effect in the data.
  2359. \end_layout
  2360. \begin_layout Standard
  2361. \begin_inset Float figure
  2362. wide false
  2363. sideways false
  2364. status collapsed
  2365. \begin_layout Plain Layout
  2366. \align center
  2367. \begin_inset Float figure
  2368. wide false
  2369. sideways false
  2370. status open
  2371. \begin_layout Plain Layout
  2372. \align center
  2373. \begin_inset Graphics
  2374. filename graphics/CD4-csaw/RNA-seq/PCA-no-batchsub-CROP.png
  2375. lyxscale 25
  2376. width 75col%
  2377. groupId rna-pca-subfig
  2378. \end_inset
  2379. \end_layout
  2380. \begin_layout Plain Layout
  2381. \begin_inset Caption Standard
  2382. \begin_layout Plain Layout
  2383. \series bold
  2384. \begin_inset CommandInset label
  2385. LatexCommand label
  2386. name "fig:RNA-PCA-no-batchsub"
  2387. \end_inset
  2388. Before batch correction
  2389. \end_layout
  2390. \end_inset
  2391. \end_layout
  2392. \end_inset
  2393. \end_layout
  2394. \begin_layout Plain Layout
  2395. \align center
  2396. \begin_inset Float figure
  2397. wide false
  2398. sideways false
  2399. status open
  2400. \begin_layout Plain Layout
  2401. \align center
  2402. \begin_inset Graphics
  2403. filename graphics/CD4-csaw/RNA-seq/PCA-combat-batchsub-CROP.png
  2404. lyxscale 25
  2405. width 75col%
  2406. groupId rna-pca-subfig
  2407. \end_inset
  2408. \end_layout
  2409. \begin_layout Plain Layout
  2410. \begin_inset Caption Standard
  2411. \begin_layout Plain Layout
  2412. \series bold
  2413. \begin_inset CommandInset label
  2414. LatexCommand label
  2415. name "fig:RNA-PCA-ComBat-batchsub"
  2416. \end_inset
  2417. After batch correction with ComBat
  2418. \end_layout
  2419. \end_inset
  2420. \end_layout
  2421. \end_inset
  2422. \end_layout
  2423. \begin_layout Plain Layout
  2424. \begin_inset Caption Standard
  2425. \begin_layout Plain Layout
  2426. \series bold
  2427. \begin_inset CommandInset label
  2428. LatexCommand label
  2429. name "fig:RNA-PCA"
  2430. \end_inset
  2431. PCoA plots of RNA-seq data showing effect of batch correction.
  2432. \end_layout
  2433. \end_inset
  2434. \end_layout
  2435. \end_inset
  2436. \end_layout
  2437. \begin_layout Standard
  2438. Due to an error in sample preparation, the RNA from the samples for days
  2439. 0 and 5 were sequenced using a different kit than those for days 1 and
  2440. 14.
  2441. This induced a substantial batch effect in the data due to differences
  2442. in sequencing biases between the two kits, and this batch effect is unfortunate
  2443. ly confounded with the time point variable (Figure
  2444. \begin_inset CommandInset ref
  2445. LatexCommand ref
  2446. reference "fig:RNA-PCA-no-batchsub"
  2447. plural "false"
  2448. caps "false"
  2449. noprefix "false"
  2450. \end_inset
  2451. ).
  2452. To do the best possible analysis with this data, this batch effect was
  2453. subtracted out from the data using ComBat
  2454. \begin_inset CommandInset citation
  2455. LatexCommand cite
  2456. key "Johnson2007"
  2457. literal "false"
  2458. \end_inset
  2459. , ignoring the time point variable due to the confounding with the batch
  2460. variable.
  2461. The result is a marked improvement, but the unavoidable confounding with
  2462. time point means that certain real patterns of gene expression will be
  2463. indistinguishable from the batch effect and subtracted out as a result.
  2464. Specifically, any
  2465. \begin_inset Quotes eld
  2466. \end_inset
  2467. zig-zag
  2468. \begin_inset Quotes erd
  2469. \end_inset
  2470. pattern, such as a gene whose expression goes up on day 1, down on day
  2471. 5, and back up again on day 14, will be attenuated or eliminated entirely.
  2472. In the context of a T-cell activation time course, it is unlikely that
  2473. many genes of interest will follow such an expression pattern, so this
  2474. loss was deemed an acceptable cost for correcting the batch effect.
  2475. \end_layout
  2476. \begin_layout Standard
  2477. \begin_inset Float figure
  2478. wide false
  2479. sideways false
  2480. status collapsed
  2481. \begin_layout Plain Layout
  2482. \begin_inset Flex TODO Note (inline)
  2483. status open
  2484. \begin_layout Plain Layout
  2485. Just take the top row
  2486. \end_layout
  2487. \end_inset
  2488. \end_layout
  2489. \begin_layout Plain Layout
  2490. \align center
  2491. \begin_inset Graphics
  2492. filename graphics/CD4-csaw/RNA-seq/weights-vs-covars-CROP.png
  2493. lyxscale 25
  2494. width 100col%
  2495. groupId colwidth-raster
  2496. \end_inset
  2497. \end_layout
  2498. \begin_layout Plain Layout
  2499. \begin_inset Caption Standard
  2500. \begin_layout Plain Layout
  2501. \series bold
  2502. \begin_inset CommandInset label
  2503. LatexCommand label
  2504. name "fig:RNA-seq-weights-vs-covars"
  2505. \end_inset
  2506. RNA-seq sample weights, grouped by experimental and technical covariates.
  2507. \end_layout
  2508. \end_inset
  2509. \end_layout
  2510. \end_inset
  2511. \end_layout
  2512. \begin_layout Standard
  2513. However, removing the systematic component of the batch effect still leaves
  2514. the noise component.
  2515. The gene quantifications from the first batch are substantially noisier
  2516. than those in the second batch.
  2517. This analysis corrected for this by using
  2518. \begin_inset Flex Code
  2519. status open
  2520. \begin_layout Plain Layout
  2521. limma
  2522. \end_layout
  2523. \end_inset
  2524. 's sample weighting method to assign lower weights to the noisy samples
  2525. of batch 1
  2526. \begin_inset CommandInset citation
  2527. LatexCommand cite
  2528. key "Ritchie2006,Liu2015"
  2529. literal "false"
  2530. \end_inset
  2531. .
  2532. The resulting analysis gives an accurate assessment of statistical significance
  2533. for all comparisons, which unfortunately means a loss of statistical power
  2534. for comparisons involving samples in batch 1.
  2535. \end_layout
  2536. \begin_layout Standard
  2537. In any case, the
  2538. \begin_inset Flex Glossary Term
  2539. status open
  2540. \begin_layout Plain Layout
  2541. RNA-seq
  2542. \end_layout
  2543. \end_inset
  2544. counts were first normalized using
  2545. \begin_inset Flex Glossary Term
  2546. status open
  2547. \begin_layout Plain Layout
  2548. TMM
  2549. \end_layout
  2550. \end_inset
  2551. \begin_inset CommandInset nomenclature
  2552. LatexCommand nomenclature
  2553. symbol "TMM"
  2554. description "trimmed mean of M-values"
  2555. literal "false"
  2556. \end_inset
  2557. \begin_inset CommandInset citation
  2558. LatexCommand cite
  2559. key "Robinson2010"
  2560. literal "false"
  2561. \end_inset
  2562. , converted to normalized
  2563. \begin_inset Flex Glossary Term
  2564. status open
  2565. \begin_layout Plain Layout
  2566. logCPM
  2567. \end_layout
  2568. \end_inset
  2569. \begin_inset CommandInset nomenclature
  2570. LatexCommand nomenclature
  2571. symbol "logCPM"
  2572. description "$\\log_2$ counts per million"
  2573. literal "false"
  2574. \end_inset
  2575. with quality weights using
  2576. \begin_inset Flex Code
  2577. status open
  2578. \begin_layout Plain Layout
  2579. voomWithQualityWeights
  2580. \end_layout
  2581. \end_inset
  2582. \begin_inset CommandInset citation
  2583. LatexCommand cite
  2584. key "Law2013,Liu2015"
  2585. literal "false"
  2586. \end_inset
  2587. , and batch-corrected at this point using ComBat.
  2588. A linear model was fit to the batch-corrected, quality-weighted data for
  2589. each gene using
  2590. \begin_inset Flex Code
  2591. status open
  2592. \begin_layout Plain Layout
  2593. limma
  2594. \end_layout
  2595. \end_inset
  2596. , and each gene was tested for differential expression using
  2597. \begin_inset Flex Code
  2598. status open
  2599. \begin_layout Plain Layout
  2600. limma
  2601. \end_layout
  2602. \end_inset
  2603. 's empirical Bayes moderated
  2604. \begin_inset Formula $t$
  2605. \end_inset
  2606. -test
  2607. \begin_inset CommandInset citation
  2608. LatexCommand cite
  2609. key "Smyth2005,Law2013,Phipson2013"
  2610. literal "false"
  2611. \end_inset
  2612. .
  2613. P-values were corrected for multiple testing using the
  2614. \begin_inset Flex Glossary Term
  2615. status open
  2616. \begin_layout Plain Layout
  2617. BH
  2618. \end_layout
  2619. \end_inset
  2620. \begin_inset CommandInset nomenclature
  2621. LatexCommand nomenclature
  2622. symbol "BH"
  2623. description "Benjamini-Hochberg"
  2624. literal "false"
  2625. \end_inset
  2626. procedure for
  2627. \begin_inset Flex Glossary Term
  2628. status open
  2629. \begin_layout Plain Layout
  2630. FDR
  2631. \end_layout
  2632. \end_inset
  2633. control
  2634. \begin_inset CommandInset citation
  2635. LatexCommand cite
  2636. key "Benjamini1995"
  2637. literal "false"
  2638. \end_inset
  2639. .
  2640. \end_layout
  2641. \begin_layout Subsection
  2642. ChIP-seq differential modification analysis
  2643. \end_layout
  2644. \begin_layout Standard
  2645. \begin_inset Float figure
  2646. wide false
  2647. sideways false
  2648. status collapsed
  2649. \begin_layout Plain Layout
  2650. \align center
  2651. \begin_inset Float figure
  2652. wide false
  2653. sideways false
  2654. status open
  2655. \begin_layout Plain Layout
  2656. \align center
  2657. \begin_inset Graphics
  2658. filename graphics/CD4-csaw/csaw/CCF-plots-noBL-PAGE2-CROP.pdf
  2659. lyxscale 50
  2660. height 40theight%
  2661. groupId ccf-subfig
  2662. \end_inset
  2663. \end_layout
  2664. \begin_layout Plain Layout
  2665. \begin_inset Caption Standard
  2666. \begin_layout Plain Layout
  2667. \series bold
  2668. \begin_inset CommandInset label
  2669. LatexCommand label
  2670. name "fig:CCF-without-blacklist"
  2671. \end_inset
  2672. Cross-correlation plots without removing blacklisted reads.
  2673. \series default
  2674. Without blacklisting, many artifactual peaks are visible in the cross-correlatio
  2675. ns of the ChIP-seq samples, and the peak at the true fragment size (147
  2676. \begin_inset space ~
  2677. \end_inset
  2678. bp) is frequently overshadowed by the artifactual peak at the read length
  2679. (100
  2680. \begin_inset space ~
  2681. \end_inset
  2682. bp).
  2683. \end_layout
  2684. \end_inset
  2685. \end_layout
  2686. \end_inset
  2687. \end_layout
  2688. \begin_layout Plain Layout
  2689. \align center
  2690. \begin_inset Float figure
  2691. wide false
  2692. sideways false
  2693. status open
  2694. \begin_layout Plain Layout
  2695. \align center
  2696. \begin_inset Graphics
  2697. filename graphics/CD4-csaw/csaw/CCF-plots-PAGE2-CROP.pdf
  2698. lyxscale 50
  2699. height 40theight%
  2700. groupId ccf-subfig
  2701. \end_inset
  2702. \end_layout
  2703. \begin_layout Plain Layout
  2704. \begin_inset Caption Standard
  2705. \begin_layout Plain Layout
  2706. \series bold
  2707. \begin_inset CommandInset label
  2708. LatexCommand label
  2709. name "fig:CCF-with-blacklist"
  2710. \end_inset
  2711. Cross-correlation plots with blacklisted reads removed.
  2712. \series default
  2713. After blacklisting, most ChIP-seq samples have clean-looking periodic cross-cor
  2714. relation plots, with the largest peak around 147
  2715. \begin_inset space ~
  2716. \end_inset
  2717. bp, the expected size for a fragment of DNA from a single nucleosome, and
  2718. little to no peak at the read length, 100
  2719. \begin_inset space ~
  2720. \end_inset
  2721. bp.
  2722. \end_layout
  2723. \end_inset
  2724. \end_layout
  2725. \end_inset
  2726. \end_layout
  2727. \begin_layout Plain Layout
  2728. \begin_inset Caption Standard
  2729. \begin_layout Plain Layout
  2730. \series bold
  2731. \begin_inset CommandInset label
  2732. LatexCommand label
  2733. name "fig:CCF-master"
  2734. \end_inset
  2735. Strand cross-correlation plots for ChIP-seq data, before and after blacklisting.
  2736. \end_layout
  2737. \end_inset
  2738. \end_layout
  2739. \end_inset
  2740. \end_layout
  2741. \begin_layout Standard
  2742. \begin_inset Note Note
  2743. status open
  2744. \begin_layout Plain Layout
  2745. \begin_inset Float figure
  2746. wide false
  2747. sideways false
  2748. status collapsed
  2749. \begin_layout Plain Layout
  2750. \align center
  2751. \begin_inset Graphics
  2752. filename graphics/CD4-csaw/ChIP-seq/H3K4me2-sample-MAplot-bins-CROP.png
  2753. lyxscale 25
  2754. width 100col%
  2755. groupId colwidth-raster
  2756. \end_inset
  2757. \end_layout
  2758. \begin_layout Plain Layout
  2759. \begin_inset Caption Standard
  2760. \begin_layout Plain Layout
  2761. \series bold
  2762. \begin_inset CommandInset label
  2763. LatexCommand label
  2764. name "fig:MA-plot-bigbins"
  2765. \end_inset
  2766. MA plot of H3K4me2 read counts in 10kb bins for two arbitrary samples.
  2767. \end_layout
  2768. \end_inset
  2769. \end_layout
  2770. \end_inset
  2771. \end_layout
  2772. \end_inset
  2773. \end_layout
  2774. \begin_layout Standard
  2775. \begin_inset Flex TODO Note (inline)
  2776. status open
  2777. \begin_layout Plain Layout
  2778. Be consistent about use of
  2779. \begin_inset Quotes eld
  2780. \end_inset
  2781. differential binding
  2782. \begin_inset Quotes erd
  2783. \end_inset
  2784. vs
  2785. \begin_inset Quotes eld
  2786. \end_inset
  2787. differential modification
  2788. \begin_inset Quotes erd
  2789. \end_inset
  2790. throughout this chapter.
  2791. The latter is usually preferred.
  2792. \end_layout
  2793. \end_inset
  2794. \end_layout
  2795. \begin_layout Standard
  2796. Sequence reads were retrieved from
  2797. \begin_inset Flex Glossary Term
  2798. status open
  2799. \begin_layout Plain Layout
  2800. SRA
  2801. \end_layout
  2802. \end_inset
  2803. \begin_inset CommandInset citation
  2804. LatexCommand cite
  2805. key "Leinonen2011"
  2806. literal "false"
  2807. \end_inset
  2808. .
  2809. \begin_inset Flex Glossary Term (Capital)
  2810. status open
  2811. \begin_layout Plain Layout
  2812. ChIP-seq
  2813. \end_layout
  2814. \end_inset
  2815. (and input) reads were aligned to GRCh38 genome assembly using Bowtie 2
  2816. \begin_inset CommandInset citation
  2817. LatexCommand cite
  2818. key "Langmead2012,Schneider2017,gh-hg38-ref"
  2819. literal "false"
  2820. \end_inset
  2821. .
  2822. Artifact regions were annotated using a custom implementation of the
  2823. \begin_inset Flex Code
  2824. status open
  2825. \begin_layout Plain Layout
  2826. GreyListChIP
  2827. \end_layout
  2828. \end_inset
  2829. algorithm, and these
  2830. \begin_inset Quotes eld
  2831. \end_inset
  2832. greylists
  2833. \begin_inset Quotes erd
  2834. \end_inset
  2835. were merged with the published
  2836. \begin_inset Flex Glossary Term
  2837. status open
  2838. \begin_layout Plain Layout
  2839. ENCODE
  2840. \end_layout
  2841. \end_inset
  2842. blacklists
  2843. \begin_inset CommandInset citation
  2844. LatexCommand cite
  2845. key "greylistchip,Amemiya2019,Dunham2012,gh-cd4-csaw"
  2846. literal "false"
  2847. \end_inset
  2848. .
  2849. Any read or called peak overlapping one of these regions was regarded as
  2850. artifactual and excluded from downstream analyses.
  2851. Figure
  2852. \begin_inset CommandInset ref
  2853. LatexCommand ref
  2854. reference "fig:CCF-master"
  2855. plural "false"
  2856. caps "false"
  2857. noprefix "false"
  2858. \end_inset
  2859. shows the improvement after blacklisting in the strand cross-correlation
  2860. plots, a common quality control plot for
  2861. \begin_inset Flex Glossary Term
  2862. status open
  2863. \begin_layout Plain Layout
  2864. ChIP-seq
  2865. \end_layout
  2866. \end_inset
  2867. data.
  2868. Peaks were called using
  2869. \begin_inset Flex Code
  2870. status open
  2871. \begin_layout Plain Layout
  2872. epic
  2873. \end_layout
  2874. \end_inset
  2875. , an implementation of the
  2876. \begin_inset Flex Glossary Term
  2877. status open
  2878. \begin_layout Plain Layout
  2879. SICER
  2880. \end_layout
  2881. \end_inset
  2882. algorithm
  2883. \begin_inset CommandInset citation
  2884. LatexCommand cite
  2885. key "Zang2009,gh-epic"
  2886. literal "false"
  2887. \end_inset
  2888. .
  2889. Peaks were also called separately using
  2890. \begin_inset Flex Glossary Term
  2891. status open
  2892. \begin_layout Plain Layout
  2893. MACS
  2894. \end_layout
  2895. \end_inset
  2896. , but
  2897. \begin_inset Flex Glossary Term
  2898. status open
  2899. \begin_layout Plain Layout
  2900. MACS
  2901. \end_layout
  2902. \end_inset
  2903. was determined to be a poor fit for the data, and these peak calls are
  2904. not used in any further analyses
  2905. \begin_inset CommandInset citation
  2906. LatexCommand cite
  2907. key "Zhang2008"
  2908. literal "false"
  2909. \end_inset
  2910. .
  2911. Consensus peaks were determined by applying the
  2912. \begin_inset Flex Glossary Term
  2913. status open
  2914. \begin_layout Plain Layout
  2915. IDR
  2916. \end_layout
  2917. \end_inset
  2918. framework
  2919. \begin_inset CommandInset citation
  2920. LatexCommand cite
  2921. key "Li2006,gh-idr"
  2922. literal "false"
  2923. \end_inset
  2924. to find peaks consistently called in the same locations across all 4 donors.
  2925. \end_layout
  2926. \begin_layout Standard
  2927. Promoters were defined by computing the distance from each annotated
  2928. \begin_inset Flex Glossary Term
  2929. status open
  2930. \begin_layout Plain Layout
  2931. TSS
  2932. \end_layout
  2933. \end_inset
  2934. to the nearest called peak and examining the distribution of distances,
  2935. observing that peaks for each histone mark were enriched within a certain
  2936. distance of the
  2937. \begin_inset Flex Glossary Term
  2938. status open
  2939. \begin_layout Plain Layout
  2940. TSS
  2941. \end_layout
  2942. \end_inset
  2943. .
  2944. For H3K4me2 and H3K4me3, this distance was about 1
  2945. \begin_inset space ~
  2946. \end_inset
  2947. kb, while for H3K27me3 it was 2.5
  2948. \begin_inset space ~
  2949. \end_inset
  2950. kb.
  2951. These distances were used as an
  2952. \begin_inset Quotes eld
  2953. \end_inset
  2954. effective promoter radius
  2955. \begin_inset Quotes erd
  2956. \end_inset
  2957. for each mark.
  2958. The promoter region for each gene was defined as the region of the genome
  2959. within this distance upstream or downstream of the gene's annotated
  2960. \begin_inset Flex Glossary Term
  2961. status open
  2962. \begin_layout Plain Layout
  2963. TSS
  2964. \end_layout
  2965. \end_inset
  2966. .
  2967. For genes with multiple annotated
  2968. \begin_inset ERT
  2969. status open
  2970. \begin_layout Plain Layout
  2971. \backslash
  2972. glspl*{TSS}
  2973. \end_layout
  2974. \end_inset
  2975. , a promoter region was defined for each
  2976. \begin_inset Flex Glossary Term
  2977. status open
  2978. \begin_layout Plain Layout
  2979. TSS
  2980. \end_layout
  2981. \end_inset
  2982. individually, and any promoters that overlapped (due to multiple
  2983. \begin_inset ERT
  2984. status open
  2985. \begin_layout Plain Layout
  2986. \backslash
  2987. glspl*{TSS}
  2988. \end_layout
  2989. \end_inset
  2990. being closer than 2 times the radius) were merged into one large promoter.
  2991. Thus, some genes had multiple promoters defined, which were each analyzed
  2992. separately for differential modification.
  2993. \end_layout
  2994. \begin_layout Standard
  2995. \begin_inset Float figure
  2996. wide false
  2997. sideways false
  2998. status collapsed
  2999. \begin_layout Plain Layout
  3000. \begin_inset Float figure
  3001. wide false
  3002. sideways false
  3003. status collapsed
  3004. \begin_layout Plain Layout
  3005. \align center
  3006. \begin_inset Graphics
  3007. filename graphics/CD4-csaw/ChIP-seq/H3K4me2-PCA-raw-CROP.png
  3008. lyxscale 25
  3009. width 45col%
  3010. groupId pcoa-subfig
  3011. \end_inset
  3012. \end_layout
  3013. \begin_layout Plain Layout
  3014. \begin_inset Caption Standard
  3015. \begin_layout Plain Layout
  3016. \series bold
  3017. \begin_inset CommandInset label
  3018. LatexCommand label
  3019. name "fig:PCoA-H3K4me2-bad"
  3020. \end_inset
  3021. H3K4me2, no correction
  3022. \end_layout
  3023. \end_inset
  3024. \end_layout
  3025. \end_inset
  3026. \begin_inset space \hfill{}
  3027. \end_inset
  3028. \begin_inset Float figure
  3029. wide false
  3030. sideways false
  3031. status collapsed
  3032. \begin_layout Plain Layout
  3033. \align center
  3034. \begin_inset Graphics
  3035. filename graphics/CD4-csaw/ChIP-seq/H3K4me2-PCA-SVsub-CROP.png
  3036. lyxscale 25
  3037. width 45col%
  3038. groupId pcoa-subfig
  3039. \end_inset
  3040. \end_layout
  3041. \begin_layout Plain Layout
  3042. \begin_inset Caption Standard
  3043. \begin_layout Plain Layout
  3044. \series bold
  3045. \begin_inset CommandInset label
  3046. LatexCommand label
  3047. name "fig:PCoA-H3K4me2-good"
  3048. \end_inset
  3049. H3K4me2, SVs subtracted
  3050. \end_layout
  3051. \end_inset
  3052. \end_layout
  3053. \end_inset
  3054. \end_layout
  3055. \begin_layout Plain Layout
  3056. \begin_inset Float figure
  3057. wide false
  3058. sideways false
  3059. status collapsed
  3060. \begin_layout Plain Layout
  3061. \align center
  3062. \begin_inset Graphics
  3063. filename graphics/CD4-csaw/ChIP-seq/H3K4me3-PCA-raw-CROP.png
  3064. lyxscale 25
  3065. width 45col%
  3066. groupId pcoa-subfig
  3067. \end_inset
  3068. \end_layout
  3069. \begin_layout Plain Layout
  3070. \begin_inset Caption Standard
  3071. \begin_layout Plain Layout
  3072. \series bold
  3073. \begin_inset CommandInset label
  3074. LatexCommand label
  3075. name "fig:PCoA-H3K4me3-bad"
  3076. \end_inset
  3077. H3K4me3, no correction
  3078. \end_layout
  3079. \end_inset
  3080. \end_layout
  3081. \end_inset
  3082. \begin_inset space \hfill{}
  3083. \end_inset
  3084. \begin_inset Float figure
  3085. wide false
  3086. sideways false
  3087. status collapsed
  3088. \begin_layout Plain Layout
  3089. \align center
  3090. \begin_inset Graphics
  3091. filename graphics/CD4-csaw/ChIP-seq/H3K4me3-PCA-SVsub-CROP.png
  3092. lyxscale 25
  3093. width 45col%
  3094. groupId pcoa-subfig
  3095. \end_inset
  3096. \end_layout
  3097. \begin_layout Plain Layout
  3098. \begin_inset Caption Standard
  3099. \begin_layout Plain Layout
  3100. \series bold
  3101. \begin_inset CommandInset label
  3102. LatexCommand label
  3103. name "fig:PCoA-H3K4me3-good"
  3104. \end_inset
  3105. H3K4me3, SVs subtracted
  3106. \end_layout
  3107. \end_inset
  3108. \end_layout
  3109. \end_inset
  3110. \end_layout
  3111. \begin_layout Plain Layout
  3112. \begin_inset Float figure
  3113. wide false
  3114. sideways false
  3115. status collapsed
  3116. \begin_layout Plain Layout
  3117. \align center
  3118. \begin_inset Graphics
  3119. filename graphics/CD4-csaw/ChIP-seq/H3K27me3-PCA-raw-CROP.png
  3120. lyxscale 25
  3121. width 45col%
  3122. groupId pcoa-subfig
  3123. \end_inset
  3124. \end_layout
  3125. \begin_layout Plain Layout
  3126. \begin_inset Caption Standard
  3127. \begin_layout Plain Layout
  3128. \series bold
  3129. \begin_inset CommandInset label
  3130. LatexCommand label
  3131. name "fig:PCoA-H3K27me3-bad"
  3132. \end_inset
  3133. H3K27me3, no correction
  3134. \end_layout
  3135. \end_inset
  3136. \end_layout
  3137. \end_inset
  3138. \begin_inset space \hfill{}
  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/ChIP-seq/H3K27me3-PCA-SVsub-CROP.png
  3148. lyxscale 25
  3149. width 45col%
  3150. groupId pcoa-subfig
  3151. \end_inset
  3152. \end_layout
  3153. \begin_layout Plain Layout
  3154. \begin_inset Caption Standard
  3155. \begin_layout Plain Layout
  3156. \series bold
  3157. \begin_inset CommandInset label
  3158. LatexCommand label
  3159. name "fig:PCoA-H3K27me3-good"
  3160. \end_inset
  3161. H3K27me3, SVs subtracted
  3162. \end_layout
  3163. \end_inset
  3164. \end_layout
  3165. \end_inset
  3166. \end_layout
  3167. \begin_layout Plain Layout
  3168. \begin_inset Caption Standard
  3169. \begin_layout Plain Layout
  3170. \series bold
  3171. \begin_inset CommandInset label
  3172. LatexCommand label
  3173. name "fig:PCoA-ChIP"
  3174. \end_inset
  3175. PCoA plots of ChIP-seq sliding window data, before and after subtracting
  3176. surrogate variables (SVs).
  3177. \end_layout
  3178. \end_inset
  3179. \end_layout
  3180. \end_inset
  3181. \end_layout
  3182. \begin_layout Standard
  3183. Reads in promoters, peaks, and sliding windows across the genome were counted
  3184. and normalized using
  3185. \begin_inset Flex Code
  3186. status open
  3187. \begin_layout Plain Layout
  3188. csaw
  3189. \end_layout
  3190. \end_inset
  3191. and analyzed for differential modification using
  3192. \begin_inset Flex Code
  3193. status open
  3194. \begin_layout Plain Layout
  3195. edgeR
  3196. \end_layout
  3197. \end_inset
  3198. \begin_inset CommandInset citation
  3199. LatexCommand cite
  3200. key "Lun2014,Lun2015a,Lund2012,Phipson2016"
  3201. literal "false"
  3202. \end_inset
  3203. .
  3204. Unobserved confounding factors in the
  3205. \begin_inset Flex Glossary Term
  3206. status open
  3207. \begin_layout Plain Layout
  3208. ChIP-seq
  3209. \end_layout
  3210. \end_inset
  3211. data were corrected using
  3212. \begin_inset Flex Glossary Term
  3213. status open
  3214. \begin_layout Plain Layout
  3215. SVA
  3216. \end_layout
  3217. \end_inset
  3218. \begin_inset CommandInset citation
  3219. LatexCommand cite
  3220. key "Leek2007,Leek2014"
  3221. literal "false"
  3222. \end_inset
  3223. .
  3224. Principal coordinate plots of the promoter count data for each histone
  3225. mark before and after subtracting surrogate variable effects are shown
  3226. in Figure
  3227. \begin_inset CommandInset ref
  3228. LatexCommand ref
  3229. reference "fig:PCoA-ChIP"
  3230. plural "false"
  3231. caps "false"
  3232. noprefix "false"
  3233. \end_inset
  3234. .
  3235. \end_layout
  3236. \begin_layout Standard
  3237. To investigate whether the location of a peak within the promoter region
  3238. was important,
  3239. \begin_inset Quotes eld
  3240. \end_inset
  3241. relative coverage profiles
  3242. \begin_inset Quotes erd
  3243. \end_inset
  3244. were generated.
  3245. First, 500-bp sliding windows were tiled around each annotated
  3246. \begin_inset Flex Glossary Term
  3247. status open
  3248. \begin_layout Plain Layout
  3249. TSS
  3250. \end_layout
  3251. \end_inset
  3252. : one window centered on the
  3253. \begin_inset Flex Glossary Term
  3254. status open
  3255. \begin_layout Plain Layout
  3256. TSS
  3257. \end_layout
  3258. \end_inset
  3259. itself, and 10 windows each upstream and downstream, thus covering a 10.5-kb
  3260. region centered on the
  3261. \begin_inset Flex Glossary Term
  3262. status open
  3263. \begin_layout Plain Layout
  3264. TSS
  3265. \end_layout
  3266. \end_inset
  3267. with 21 windows.
  3268. Reads in each window for each
  3269. \begin_inset Flex Glossary Term
  3270. status open
  3271. \begin_layout Plain Layout
  3272. TSS
  3273. \end_layout
  3274. \end_inset
  3275. were counted in each sample, and the counts were normalized and converted
  3276. to
  3277. \begin_inset Flex Glossary Term
  3278. status open
  3279. \begin_layout Plain Layout
  3280. logCPM
  3281. \end_layout
  3282. \end_inset
  3283. as in the differential modification analysis.
  3284. Then, the
  3285. \begin_inset Flex Glossary Term
  3286. status open
  3287. \begin_layout Plain Layout
  3288. logCPM
  3289. \end_layout
  3290. \end_inset
  3291. values within each promoter were normalized to an average of zero, such
  3292. that each window's normalized abundance now represents the relative read
  3293. depth of that window compared to all other windows in the same promoter.
  3294. The normalized abundance values for each window in a promoter are collectively
  3295. referred to as that promoter's
  3296. \begin_inset Quotes eld
  3297. \end_inset
  3298. relative coverage profile
  3299. \begin_inset Quotes erd
  3300. \end_inset
  3301. .
  3302. \end_layout
  3303. \begin_layout Subsection
  3304. MOFA recovers biologically relevant variation from blind analysis by correlating
  3305. across datasets
  3306. \end_layout
  3307. \begin_layout Standard
  3308. \begin_inset ERT
  3309. status open
  3310. \begin_layout Plain Layout
  3311. \backslash
  3312. afterpage{
  3313. \end_layout
  3314. \begin_layout Plain Layout
  3315. \backslash
  3316. begin{landscape}
  3317. \end_layout
  3318. \end_inset
  3319. \end_layout
  3320. \begin_layout Standard
  3321. \begin_inset Float figure
  3322. wide false
  3323. sideways false
  3324. status open
  3325. \begin_layout Plain Layout
  3326. \begin_inset Float figure
  3327. wide false
  3328. sideways false
  3329. status open
  3330. \begin_layout Plain Layout
  3331. \align center
  3332. \begin_inset Graphics
  3333. filename graphics/CD4-csaw/MOFA-varExplaiend-matrix-CROP.png
  3334. lyxscale 25
  3335. width 45col%
  3336. groupId mofa-subfig
  3337. \end_inset
  3338. \end_layout
  3339. \begin_layout Plain Layout
  3340. \begin_inset Caption Standard
  3341. \begin_layout Plain Layout
  3342. \series bold
  3343. \begin_inset CommandInset label
  3344. LatexCommand label
  3345. name "fig:mofa-varexplained"
  3346. \end_inset
  3347. Variance explained in each data set by each latent factor estimated by MOFA.
  3348. \series default
  3349. For each LF learned by MOFA, the variance explained by that factor in each
  3350. data set (
  3351. \begin_inset Quotes eld
  3352. \end_inset
  3353. view
  3354. \begin_inset Quotes erd
  3355. \end_inset
  3356. ) is shown by the shading of the cells in the lower section.
  3357. The upper section shows the total fraction of each data set's variance
  3358. that is explained by all LFs combined.
  3359. \end_layout
  3360. \end_inset
  3361. \end_layout
  3362. \end_inset
  3363. \begin_inset space \hfill{}
  3364. \end_inset
  3365. \begin_inset Float figure
  3366. wide false
  3367. sideways false
  3368. status open
  3369. \begin_layout Plain Layout
  3370. \align center
  3371. \begin_inset Graphics
  3372. filename graphics/CD4-csaw/MOFA-LF-scatter-CROP.png
  3373. lyxscale 25
  3374. width 45col%
  3375. groupId mofa-subfig
  3376. \end_inset
  3377. \end_layout
  3378. \begin_layout Plain Layout
  3379. \begin_inset Caption Standard
  3380. \begin_layout Plain Layout
  3381. \series bold
  3382. \begin_inset CommandInset label
  3383. LatexCommand label
  3384. name "fig:mofa-lf-scatter"
  3385. \end_inset
  3386. Scatter plots of specific pairs of MOFA latent factors.
  3387. \series default
  3388. LFs 1, 4, and 5 explain substantial variation in all data sets, so they
  3389. are plotted against each other in order to reveal patterns of variation
  3390. that are shared across all data sets.
  3391. \end_layout
  3392. \end_inset
  3393. \end_layout
  3394. \end_inset
  3395. \end_layout
  3396. \begin_layout Plain Layout
  3397. \begin_inset Caption Standard
  3398. \begin_layout Plain Layout
  3399. \series bold
  3400. \begin_inset CommandInset label
  3401. LatexCommand label
  3402. name "fig:MOFA-master"
  3403. \end_inset
  3404. MOFA latent factors separate technical confounders from
  3405. \end_layout
  3406. \end_inset
  3407. \end_layout
  3408. \end_inset
  3409. \end_layout
  3410. \begin_layout Standard
  3411. \begin_inset ERT
  3412. status open
  3413. \begin_layout Plain Layout
  3414. \backslash
  3415. end{landscape}
  3416. \end_layout
  3417. \begin_layout Plain Layout
  3418. }
  3419. \end_layout
  3420. \end_inset
  3421. \end_layout
  3422. \begin_layout Standard
  3423. \begin_inset Flex Glossary Term
  3424. status open
  3425. \begin_layout Plain Layout
  3426. MOFA
  3427. \end_layout
  3428. \end_inset
  3429. \begin_inset CommandInset nomenclature
  3430. LatexCommand nomenclature
  3431. symbol "MOFA"
  3432. description "Multi-Omics Factor Analysis"
  3433. literal "false"
  3434. \end_inset
  3435. was run on all the
  3436. \begin_inset Flex Glossary Term
  3437. status open
  3438. \begin_layout Plain Layout
  3439. ChIP-seq
  3440. \end_layout
  3441. \end_inset
  3442. windows overlapping consensus peaks for each histone mark, as well as the
  3443. \begin_inset Flex Glossary Term
  3444. status open
  3445. \begin_layout Plain Layout
  3446. RNA-seq
  3447. \end_layout
  3448. \end_inset
  3449. data, in order to identify patterns of coordinated variation across all
  3450. data sets
  3451. \begin_inset CommandInset citation
  3452. LatexCommand cite
  3453. key "Argelaguet2018"
  3454. literal "false"
  3455. \end_inset
  3456. .
  3457. The results are summarized in Figure
  3458. \begin_inset CommandInset ref
  3459. LatexCommand ref
  3460. reference "fig:MOFA-master"
  3461. plural "false"
  3462. caps "false"
  3463. noprefix "false"
  3464. \end_inset
  3465. .
  3466. \begin_inset ERT
  3467. status open
  3468. \begin_layout Plain Layout
  3469. \backslash
  3470. Glspl*{LF}
  3471. \end_layout
  3472. \end_inset
  3473. \begin_inset CommandInset nomenclature
  3474. LatexCommand nomenclature
  3475. symbol "LF"
  3476. description "latent factor"
  3477. literal "false"
  3478. \end_inset
  3479. 1, 4, and 5 were determined to explain the most variation consistently
  3480. across all data sets (Figure
  3481. \begin_inset CommandInset ref
  3482. LatexCommand ref
  3483. reference "fig:mofa-varexplained"
  3484. plural "false"
  3485. caps "false"
  3486. noprefix "false"
  3487. \end_inset
  3488. ), and scatter plots of these factors show that they also correlate best
  3489. with the experimental factors (Figure
  3490. \begin_inset CommandInset ref
  3491. LatexCommand ref
  3492. reference "fig:mofa-lf-scatter"
  3493. plural "false"
  3494. caps "false"
  3495. noprefix "false"
  3496. \end_inset
  3497. ).
  3498. \begin_inset Flex Glossary Term
  3499. status open
  3500. \begin_layout Plain Layout
  3501. LF
  3502. \end_layout
  3503. \end_inset
  3504. 2 captures the batch effect in the
  3505. \begin_inset Flex Glossary Term
  3506. status open
  3507. \begin_layout Plain Layout
  3508. RNA-seq
  3509. \end_layout
  3510. \end_inset
  3511. data.
  3512. Removing the effect of
  3513. \begin_inset Flex Glossary Term
  3514. status open
  3515. \begin_layout Plain Layout
  3516. LF
  3517. \end_layout
  3518. \end_inset
  3519. 2 using
  3520. \begin_inset Flex Glossary Term
  3521. status open
  3522. \begin_layout Plain Layout
  3523. MOFA
  3524. \end_layout
  3525. \end_inset
  3526. theoretically yields a batch correction that does not depend on knowing
  3527. the experimental factors.
  3528. When this was attempted, the resulting batch correction was comparable
  3529. to ComBat (see Figure
  3530. \begin_inset CommandInset ref
  3531. LatexCommand ref
  3532. reference "fig:RNA-PCA-ComBat-batchsub"
  3533. plural "false"
  3534. caps "false"
  3535. noprefix "false"
  3536. \end_inset
  3537. ), indicating that the ComBat-based batch correction has little room for
  3538. improvement given the problems with the data set.
  3539. \end_layout
  3540. \begin_layout Standard
  3541. \begin_inset Note Note
  3542. status collapsed
  3543. \begin_layout Plain Layout
  3544. \begin_inset Float figure
  3545. wide false
  3546. sideways false
  3547. status open
  3548. \begin_layout Plain Layout
  3549. \align center
  3550. \begin_inset Graphics
  3551. filename graphics/CD4-csaw/MOFA-batch-correct-CROP.png
  3552. lyxscale 25
  3553. width 100col%
  3554. groupId colwidth-raster
  3555. \end_inset
  3556. \end_layout
  3557. \begin_layout Plain Layout
  3558. \begin_inset Caption Standard
  3559. \begin_layout Plain Layout
  3560. \series bold
  3561. \begin_inset CommandInset label
  3562. LatexCommand label
  3563. name "fig:mofa-batchsub"
  3564. \end_inset
  3565. Result of RNA-seq batch-correction using MOFA latent factors
  3566. \end_layout
  3567. \end_inset
  3568. \end_layout
  3569. \end_inset
  3570. \end_layout
  3571. \end_inset
  3572. \end_layout
  3573. \begin_layout Standard
  3574. \begin_inset Note Note
  3575. status open
  3576. \begin_layout Plain Layout
  3577. Placing these floats is a challenge
  3578. \end_layout
  3579. \end_inset
  3580. \end_layout
  3581. \begin_layout Standard
  3582. \begin_inset Float table
  3583. wide false
  3584. sideways false
  3585. status collapsed
  3586. \begin_layout Plain Layout
  3587. \align center
  3588. \begin_inset Tabular
  3589. <lyxtabular version="3" rows="11" columns="3">
  3590. <features tabularvalignment="middle">
  3591. <column alignment="center" valignment="top">
  3592. <column alignment="center" valignment="top">
  3593. <column alignment="center" valignment="top">
  3594. <row>
  3595. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  3596. \begin_inset Text
  3597. \begin_layout Plain Layout
  3598. Test
  3599. \end_layout
  3600. \end_inset
  3601. </cell>
  3602. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  3603. \begin_inset Text
  3604. \begin_layout Plain Layout
  3605. Est.
  3606. non-null
  3607. \end_layout
  3608. \end_inset
  3609. </cell>
  3610. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  3611. \begin_inset Text
  3612. \begin_layout Plain Layout
  3613. \begin_inset Formula $\mathrm{FDR}\le10\%$
  3614. \end_inset
  3615. \end_layout
  3616. \end_inset
  3617. </cell>
  3618. </row>
  3619. <row>
  3620. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  3621. \begin_inset Text
  3622. \begin_layout Plain Layout
  3623. Naïve Day 0 vs Day 1
  3624. \end_layout
  3625. \end_inset
  3626. </cell>
  3627. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  3628. \begin_inset Text
  3629. \begin_layout Plain Layout
  3630. 5992
  3631. \end_layout
  3632. \end_inset
  3633. </cell>
  3634. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  3635. \begin_inset Text
  3636. \begin_layout Plain Layout
  3637. 1613
  3638. \end_layout
  3639. \end_inset
  3640. </cell>
  3641. </row>
  3642. <row>
  3643. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  3644. \begin_inset Text
  3645. \begin_layout Plain Layout
  3646. Naïve Day 0 vs Day 5
  3647. \end_layout
  3648. \end_inset
  3649. </cell>
  3650. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  3651. \begin_inset Text
  3652. \begin_layout Plain Layout
  3653. 3038
  3654. \end_layout
  3655. \end_inset
  3656. </cell>
  3657. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  3658. \begin_inset Text
  3659. \begin_layout Plain Layout
  3660. 32
  3661. \end_layout
  3662. \end_inset
  3663. </cell>
  3664. </row>
  3665. <row>
  3666. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  3667. \begin_inset Text
  3668. \begin_layout Plain Layout
  3669. Naïve Day 0 vs Day 14
  3670. \end_layout
  3671. \end_inset
  3672. </cell>
  3673. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  3674. \begin_inset Text
  3675. \begin_layout Plain Layout
  3676. 1870
  3677. \end_layout
  3678. \end_inset
  3679. </cell>
  3680. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  3681. \begin_inset Text
  3682. \begin_layout Plain Layout
  3683. 190
  3684. \end_layout
  3685. \end_inset
  3686. </cell>
  3687. </row>
  3688. <row>
  3689. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  3690. \begin_inset Text
  3691. \begin_layout Plain Layout
  3692. Memory Day 0 vs Day 1
  3693. \end_layout
  3694. \end_inset
  3695. </cell>
  3696. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  3697. \begin_inset Text
  3698. \begin_layout Plain Layout
  3699. 3195
  3700. \end_layout
  3701. \end_inset
  3702. </cell>
  3703. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  3704. \begin_inset Text
  3705. \begin_layout Plain Layout
  3706. 411
  3707. \end_layout
  3708. \end_inset
  3709. </cell>
  3710. </row>
  3711. <row>
  3712. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  3713. \begin_inset Text
  3714. \begin_layout Plain Layout
  3715. Memory Day 0 vs Day 5
  3716. \end_layout
  3717. \end_inset
  3718. </cell>
  3719. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  3720. \begin_inset Text
  3721. \begin_layout Plain Layout
  3722. 2688
  3723. \end_layout
  3724. \end_inset
  3725. </cell>
  3726. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  3727. \begin_inset Text
  3728. \begin_layout Plain Layout
  3729. 18
  3730. \end_layout
  3731. \end_inset
  3732. </cell>
  3733. </row>
  3734. <row>
  3735. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  3736. \begin_inset Text
  3737. \begin_layout Plain Layout
  3738. Memory Day 0 vs Day 14
  3739. \end_layout
  3740. \end_inset
  3741. </cell>
  3742. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  3743. \begin_inset Text
  3744. \begin_layout Plain Layout
  3745. 1911
  3746. \end_layout
  3747. \end_inset
  3748. </cell>
  3749. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  3750. \begin_inset Text
  3751. \begin_layout Plain Layout
  3752. 227
  3753. \end_layout
  3754. \end_inset
  3755. </cell>
  3756. </row>
  3757. <row>
  3758. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  3759. \begin_inset Text
  3760. \begin_layout Plain Layout
  3761. Day 0 Naïve vs Memory
  3762. \end_layout
  3763. \end_inset
  3764. </cell>
  3765. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  3766. \begin_inset Text
  3767. \begin_layout Plain Layout
  3768. 0
  3769. \end_layout
  3770. \end_inset
  3771. </cell>
  3772. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  3773. \begin_inset Text
  3774. \begin_layout Plain Layout
  3775. 2
  3776. \end_layout
  3777. \end_inset
  3778. </cell>
  3779. </row>
  3780. <row>
  3781. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  3782. \begin_inset Text
  3783. \begin_layout Plain Layout
  3784. Day 1 Naïve vs Memory
  3785. \end_layout
  3786. \end_inset
  3787. </cell>
  3788. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  3789. \begin_inset Text
  3790. \begin_layout Plain Layout
  3791. 9167
  3792. \end_layout
  3793. \end_inset
  3794. </cell>
  3795. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  3796. \begin_inset Text
  3797. \begin_layout Plain Layout
  3798. 5532
  3799. \end_layout
  3800. \end_inset
  3801. </cell>
  3802. </row>
  3803. <row>
  3804. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  3805. \begin_inset Text
  3806. \begin_layout Plain Layout
  3807. Day 5 Naïve vs Memory
  3808. \end_layout
  3809. \end_inset
  3810. </cell>
  3811. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  3812. \begin_inset Text
  3813. \begin_layout Plain Layout
  3814. 0
  3815. \end_layout
  3816. \end_inset
  3817. </cell>
  3818. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  3819. \begin_inset Text
  3820. \begin_layout Plain Layout
  3821. 0
  3822. \end_layout
  3823. \end_inset
  3824. </cell>
  3825. </row>
  3826. <row>
  3827. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  3828. \begin_inset Text
  3829. \begin_layout Plain Layout
  3830. Day 14 Naïve vs Memory
  3831. \end_layout
  3832. \end_inset
  3833. </cell>
  3834. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  3835. \begin_inset Text
  3836. \begin_layout Plain Layout
  3837. 6446
  3838. \end_layout
  3839. \end_inset
  3840. </cell>
  3841. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  3842. \begin_inset Text
  3843. \begin_layout Plain Layout
  3844. 2319
  3845. \end_layout
  3846. \end_inset
  3847. </cell>
  3848. </row>
  3849. </lyxtabular>
  3850. \end_inset
  3851. \end_layout
  3852. \begin_layout Plain Layout
  3853. \begin_inset Caption Standard
  3854. \begin_layout Plain Layout
  3855. \series bold
  3856. \begin_inset CommandInset label
  3857. LatexCommand label
  3858. name "tab:Estimated-and-detected-rnaseq"
  3859. \end_inset
  3860. Estimated and detected differentially expressed genes.
  3861. \series default
  3862. \begin_inset Quotes eld
  3863. \end_inset
  3864. Test
  3865. \begin_inset Quotes erd
  3866. \end_inset
  3867. : Which sample groups were compared;
  3868. \begin_inset Quotes eld
  3869. \end_inset
  3870. Est non-null
  3871. \begin_inset Quotes erd
  3872. \end_inset
  3873. : Estimated number of differentially expressed genes, using the method of
  3874. averaging local FDR values
  3875. \begin_inset CommandInset citation
  3876. LatexCommand cite
  3877. key "Phipson2013Thesis"
  3878. literal "false"
  3879. \end_inset
  3880. ;
  3881. \begin_inset Quotes eld
  3882. \end_inset
  3883. \begin_inset Formula $\mathrm{FDR}\le10\%$
  3884. \end_inset
  3885. \begin_inset Quotes erd
  3886. \end_inset
  3887. : Number of significantly differentially expressed genes at an FDR threshold
  3888. of 10%.
  3889. The total number of genes tested was 16707.
  3890. \end_layout
  3891. \end_inset
  3892. \end_layout
  3893. \end_inset
  3894. \end_layout
  3895. \begin_layout Section
  3896. Results
  3897. \end_layout
  3898. \begin_layout Standard
  3899. \begin_inset Flex TODO Note (inline)
  3900. status open
  3901. \begin_layout Plain Layout
  3902. Focus on what hypotheses were tested, then select figures that show how
  3903. those hypotheses were tested, even if the result is a negative.
  3904. Not every interesting result needs to be in here.
  3905. Chapter should tell a story.
  3906. \end_layout
  3907. \end_inset
  3908. \end_layout
  3909. \begin_layout Standard
  3910. \begin_inset Flex TODO Note (inline)
  3911. status open
  3912. \begin_layout Plain Layout
  3913. Maybe reorder these sections to do RNA-seq, then ChIP-seq, then combined
  3914. analyses?
  3915. \end_layout
  3916. \end_inset
  3917. \end_layout
  3918. \begin_layout Subsection
  3919. Interpretation of RNA-seq analysis is limited by a major confounding factor
  3920. \end_layout
  3921. \begin_layout Standard
  3922. \begin_inset Note Note
  3923. status open
  3924. \begin_layout Plain Layout
  3925. Putting a float here causes an error.
  3926. No idea why.
  3927. See above for the floats that should be placed here.
  3928. \end_layout
  3929. \end_inset
  3930. \end_layout
  3931. \begin_layout Standard
  3932. Genes called as present in the
  3933. \begin_inset Flex Glossary Term
  3934. status open
  3935. \begin_layout Plain Layout
  3936. RNA-seq
  3937. \end_layout
  3938. \end_inset
  3939. data were tested for differential expression between all time points and
  3940. cell types.
  3941. The counts of differentially expressed genes are shown in Table
  3942. \begin_inset CommandInset ref
  3943. LatexCommand ref
  3944. reference "tab:Estimated-and-detected-rnaseq"
  3945. plural "false"
  3946. caps "false"
  3947. noprefix "false"
  3948. \end_inset
  3949. .
  3950. Notably, all the results for Day 0 and Day 5 have substantially fewer genes
  3951. called differentially expressed than any of the results for other time
  3952. points.
  3953. This is an unfortunate result of the difference in sample quality between
  3954. the two batches of
  3955. \begin_inset Flex Glossary Term
  3956. status open
  3957. \begin_layout Plain Layout
  3958. RNA-seq
  3959. \end_layout
  3960. \end_inset
  3961. data.
  3962. All the samples in Batch 1, which includes all the samples from Days 0
  3963. and 5, have substantially more variability than the samples in Batch 2,
  3964. which includes the other time points.
  3965. This is reflected in the substantially higher weights assigned to Batch
  3966. 2 (Figure
  3967. \begin_inset CommandInset ref
  3968. LatexCommand ref
  3969. reference "fig:RNA-seq-weights-vs-covars"
  3970. plural "false"
  3971. caps "false"
  3972. noprefix "false"
  3973. \end_inset
  3974. ).
  3975. The batch effect has both a systematic component and a random noise component.
  3976. While the systematic component was subtracted out using ComBat (Figure
  3977. \begin_inset CommandInset ref
  3978. LatexCommand ref
  3979. reference "fig:RNA-PCA"
  3980. plural "false"
  3981. caps "false"
  3982. noprefix "false"
  3983. \end_inset
  3984. ), no such correction is possible for the noise component: Batch 1 simply
  3985. has substantially more random noise in it, which reduces the statistical
  3986. power for any differential expression tests involving samples in that batch.
  3987. \end_layout
  3988. \begin_layout Standard
  3989. \begin_inset Float figure
  3990. wide false
  3991. sideways false
  3992. status collapsed
  3993. \begin_layout Plain Layout
  3994. \align center
  3995. \begin_inset Graphics
  3996. filename graphics/CD4-csaw/RNA-seq/PCA-final-12-CROP.png
  3997. lyxscale 25
  3998. width 100col%
  3999. groupId colwidth-raster
  4000. \end_inset
  4001. \end_layout
  4002. \begin_layout Plain Layout
  4003. \begin_inset Caption Standard
  4004. \begin_layout Plain Layout
  4005. \series bold
  4006. \begin_inset CommandInset label
  4007. LatexCommand label
  4008. name "fig:rna-pca-final"
  4009. \end_inset
  4010. PCoA plot of RNA-seq samples after ComBat batch correction.
  4011. \series default
  4012. Each point represents an individual sample.
  4013. Samples with the same combination of cell type and time point are encircled
  4014. with a shaded region to aid in visual identification of the sample groups.
  4015. Samples with of same cell type from the same donor are connected by lines
  4016. to indicate the
  4017. \begin_inset Quotes eld
  4018. \end_inset
  4019. trajectory
  4020. \begin_inset Quotes erd
  4021. \end_inset
  4022. of each donor's cells over time in PCoA space.
  4023. \end_layout
  4024. \end_inset
  4025. \end_layout
  4026. \end_inset
  4027. \end_layout
  4028. \begin_layout Standard
  4029. Despite the difficulty in detecting specific differentially expressed genes,
  4030. there is still evidence that differential expression is present for these
  4031. time points.
  4032. In Figure
  4033. \begin_inset CommandInset ref
  4034. LatexCommand ref
  4035. reference "fig:rna-pca-final"
  4036. plural "false"
  4037. caps "false"
  4038. noprefix "false"
  4039. \end_inset
  4040. , there is a clear separation between naïve and memory samples at Day 0,
  4041. despite the fact that only 2 genes were significantly differentially expressed
  4042. for this comparison.
  4043. Similarly, the small numbers of genes detected for the Day 0 vs Day 5 compariso
  4044. ns do not reflect the large separation between these time points in Figure
  4045. \begin_inset CommandInset ref
  4046. LatexCommand ref
  4047. reference "fig:rna-pca-final"
  4048. plural "false"
  4049. caps "false"
  4050. noprefix "false"
  4051. \end_inset
  4052. .
  4053. In addition, the
  4054. \begin_inset Flex Glossary Term
  4055. status open
  4056. \begin_layout Plain Layout
  4057. MOFA
  4058. \end_layout
  4059. \end_inset
  4060. \begin_inset Flex Glossary Term
  4061. status open
  4062. \begin_layout Plain Layout
  4063. LF
  4064. \end_layout
  4065. \end_inset
  4066. plots in Figure
  4067. \begin_inset CommandInset ref
  4068. LatexCommand ref
  4069. reference "fig:mofa-lf-scatter"
  4070. plural "false"
  4071. caps "false"
  4072. noprefix "false"
  4073. \end_inset
  4074. .
  4075. This suggests that there is indeed a differential expression signal present
  4076. in the data for these comparisons, but the large variability in the Batch
  4077. 1 samples obfuscates this signal at the individual gene level.
  4078. As a result, it is impossible to make any meaningful statements about the
  4079. \begin_inset Quotes eld
  4080. \end_inset
  4081. size
  4082. \begin_inset Quotes erd
  4083. \end_inset
  4084. of the gene signature for any time point, since the number of significant
  4085. genes as well as the estimated number of differentially expressed genes
  4086. depends so strongly on the variations in sample quality in addition to
  4087. the size of the differential expression signal in the data.
  4088. Gene-set enrichment analyses are similarly impractical.
  4089. However, analyses looking at genome-wide patterns of expression are still
  4090. practical.
  4091. \end_layout
  4092. \begin_layout Subsection
  4093. H3K4 and H3K27 methylation occur in broad regions and are enriched near
  4094. promoters
  4095. \end_layout
  4096. \begin_layout Standard
  4097. \begin_inset Float table
  4098. wide false
  4099. sideways false
  4100. status collapsed
  4101. \begin_layout Plain Layout
  4102. \align center
  4103. \begin_inset Flex TODO Note (inline)
  4104. status open
  4105. \begin_layout Plain Layout
  4106. Also get
  4107. \emph on
  4108. median
  4109. \emph default
  4110. peak width and maybe other quantiles (25%, 75%)
  4111. \end_layout
  4112. \end_inset
  4113. \end_layout
  4114. \begin_layout Plain Layout
  4115. \align center
  4116. \begin_inset Tabular
  4117. <lyxtabular version="3" rows="4" columns="5">
  4118. <features tabularvalignment="middle">
  4119. <column alignment="center" valignment="top">
  4120. <column alignment="center" valignment="top">
  4121. <column alignment="center" valignment="top">
  4122. <column alignment="center" valignment="top">
  4123. <column alignment="center" valignment="top">
  4124. <row>
  4125. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  4126. \begin_inset Text
  4127. \begin_layout Plain Layout
  4128. Histone Mark
  4129. \end_layout
  4130. \end_inset
  4131. </cell>
  4132. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  4133. \begin_inset Text
  4134. \begin_layout Plain Layout
  4135. # Peaks
  4136. \end_layout
  4137. \end_inset
  4138. </cell>
  4139. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  4140. \begin_inset Text
  4141. \begin_layout Plain Layout
  4142. Mean peak width
  4143. \end_layout
  4144. \end_inset
  4145. </cell>
  4146. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  4147. \begin_inset Text
  4148. \begin_layout Plain Layout
  4149. genome coverage
  4150. \end_layout
  4151. \end_inset
  4152. </cell>
  4153. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  4154. \begin_inset Text
  4155. \begin_layout Plain Layout
  4156. FRiP
  4157. \end_layout
  4158. \end_inset
  4159. </cell>
  4160. </row>
  4161. <row>
  4162. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  4163. \begin_inset Text
  4164. \begin_layout Plain Layout
  4165. H3K4me2
  4166. \end_layout
  4167. \end_inset
  4168. </cell>
  4169. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  4170. \begin_inset Text
  4171. \begin_layout Plain Layout
  4172. 14965
  4173. \end_layout
  4174. \end_inset
  4175. </cell>
  4176. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  4177. \begin_inset Text
  4178. \begin_layout Plain Layout
  4179. 3970
  4180. \end_layout
  4181. \end_inset
  4182. </cell>
  4183. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  4184. \begin_inset Text
  4185. \begin_layout Plain Layout
  4186. 1.92%
  4187. \end_layout
  4188. \end_inset
  4189. </cell>
  4190. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  4191. \begin_inset Text
  4192. \begin_layout Plain Layout
  4193. 14.2%
  4194. \end_layout
  4195. \end_inset
  4196. </cell>
  4197. </row>
  4198. <row>
  4199. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  4200. \begin_inset Text
  4201. \begin_layout Plain Layout
  4202. H3K4me3
  4203. \end_layout
  4204. \end_inset
  4205. </cell>
  4206. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  4207. \begin_inset Text
  4208. \begin_layout Plain Layout
  4209. 6163
  4210. \end_layout
  4211. \end_inset
  4212. </cell>
  4213. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  4214. \begin_inset Text
  4215. \begin_layout Plain Layout
  4216. 2946
  4217. \end_layout
  4218. \end_inset
  4219. </cell>
  4220. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  4221. \begin_inset Text
  4222. \begin_layout Plain Layout
  4223. 0.588%
  4224. \end_layout
  4225. \end_inset
  4226. </cell>
  4227. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  4228. \begin_inset Text
  4229. \begin_layout Plain Layout
  4230. 6.57%
  4231. \end_layout
  4232. \end_inset
  4233. </cell>
  4234. </row>
  4235. <row>
  4236. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  4237. \begin_inset Text
  4238. \begin_layout Plain Layout
  4239. H3K27me3
  4240. \end_layout
  4241. \end_inset
  4242. </cell>
  4243. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  4244. \begin_inset Text
  4245. \begin_layout Plain Layout
  4246. 18139
  4247. \end_layout
  4248. \end_inset
  4249. </cell>
  4250. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  4251. \begin_inset Text
  4252. \begin_layout Plain Layout
  4253. 18967
  4254. \end_layout
  4255. \end_inset
  4256. </cell>
  4257. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  4258. \begin_inset Text
  4259. \begin_layout Plain Layout
  4260. 11.1%
  4261. \end_layout
  4262. \end_inset
  4263. </cell>
  4264. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  4265. \begin_inset Text
  4266. \begin_layout Plain Layout
  4267. 22.5%
  4268. \end_layout
  4269. \end_inset
  4270. </cell>
  4271. </row>
  4272. </lyxtabular>
  4273. \end_inset
  4274. \end_layout
  4275. \begin_layout Plain Layout
  4276. \begin_inset Flex TODO Note (inline)
  4277. status open
  4278. \begin_layout Plain Layout
  4279. Get the IDR threshold
  4280. \end_layout
  4281. \end_inset
  4282. \end_layout
  4283. \begin_layout Plain Layout
  4284. \begin_inset Caption Standard
  4285. \begin_layout Plain Layout
  4286. \series bold
  4287. \begin_inset CommandInset label
  4288. LatexCommand label
  4289. name "tab:peak-calling-summary"
  4290. \end_inset
  4291. Peak-calling summary.
  4292. \series default
  4293. For each histone mark, the number of peaks called using SICER at an IDR
  4294. threshold of ???, the mean width of those peaks, the fraction of the genome
  4295. covered by peaks, and the fraction of reads in peaks (FRiP).
  4296. \end_layout
  4297. \end_inset
  4298. \end_layout
  4299. \end_inset
  4300. \end_layout
  4301. \begin_layout Standard
  4302. Table
  4303. \begin_inset CommandInset ref
  4304. LatexCommand ref
  4305. reference "tab:peak-calling-summary"
  4306. plural "false"
  4307. caps "false"
  4308. noprefix "false"
  4309. \end_inset
  4310. gives a summary of the peak calling statistics for each histone mark.
  4311. Consistent with previous observations [CITATION NEEDED], all 3 histone
  4312. marks occur in broad regions spanning many consecutive nucleosomes, rather
  4313. than in sharp peaks as would be expected for a transcription factor or
  4314. other molecule that binds to specific sites.
  4315. This conclusion is further supported by Figure
  4316. \begin_inset CommandInset ref
  4317. LatexCommand ref
  4318. reference "fig:CCF-with-blacklist"
  4319. plural "false"
  4320. caps "false"
  4321. noprefix "false"
  4322. \end_inset
  4323. , in which a clear nucleosome-sized periodicity is visible in the cross-correlat
  4324. ion value for each sample, indicating that each time a given mark is present
  4325. on one histone, it is also likely to be found on adjacent histones as well.
  4326. H3K27me3 enrichment in particular is substantially more broad than either
  4327. H3K4 mark, with a mean peak width of almost 19,000 bp.
  4328. This is also reflected in the periodicity observed in Figure
  4329. \begin_inset CommandInset ref
  4330. LatexCommand ref
  4331. reference "fig:CCF-with-blacklist"
  4332. plural "false"
  4333. caps "false"
  4334. noprefix "false"
  4335. \end_inset
  4336. , which remains strong much farther out for H3K27me3 than the other marks,
  4337. showing H3K27me3 especially tends to be found on long runs of consecutive
  4338. histones.
  4339. \end_layout
  4340. \begin_layout Standard
  4341. \begin_inset Float figure
  4342. wide false
  4343. sideways false
  4344. status open
  4345. \begin_layout Plain Layout
  4346. \begin_inset Flex TODO Note (inline)
  4347. status open
  4348. \begin_layout Plain Layout
  4349. Ensure this figure uses the peak calls from the new analysis.
  4350. \end_layout
  4351. \end_inset
  4352. \end_layout
  4353. \begin_layout Plain Layout
  4354. \begin_inset Flex TODO Note (inline)
  4355. status open
  4356. \begin_layout Plain Layout
  4357. Need a control: shuffle all peaks and repeat, N times.
  4358. Do real vs shuffled control both in a top/bottom arrangement.
  4359. \end_layout
  4360. \end_inset
  4361. \end_layout
  4362. \begin_layout Plain Layout
  4363. \begin_inset Flex TODO Note (inline)
  4364. status open
  4365. \begin_layout Plain Layout
  4366. Consider counting TSS inside peaks as negative number indicating how far
  4367. \emph on
  4368. inside
  4369. \emph default
  4370. the peak the TSS is (i.e.
  4371. distance to nearest non-peak area).
  4372. \end_layout
  4373. \end_inset
  4374. \end_layout
  4375. \begin_layout Plain Layout
  4376. \begin_inset Flex TODO Note (inline)
  4377. status open
  4378. \begin_layout Plain Layout
  4379. The H3K4 part of this figure is included in
  4380. \begin_inset CommandInset citation
  4381. LatexCommand cite
  4382. key "LaMere2016"
  4383. literal "false"
  4384. \end_inset
  4385. as Fig.
  4386. S2.
  4387. Do I need to do anything about that?
  4388. \end_layout
  4389. \end_inset
  4390. \end_layout
  4391. \begin_layout Plain Layout
  4392. \align center
  4393. \begin_inset Graphics
  4394. filename graphics/CD4-csaw/Promoter Peak Distance Profile-PAGE1-CROP.pdf
  4395. lyxscale 50
  4396. width 80col%
  4397. \end_inset
  4398. \end_layout
  4399. \begin_layout Plain Layout
  4400. \begin_inset Caption Standard
  4401. \begin_layout Plain Layout
  4402. \series bold
  4403. \begin_inset CommandInset label
  4404. LatexCommand label
  4405. name "fig:near-promoter-peak-enrich"
  4406. \end_inset
  4407. Enrichment of peaks in promoter neighborhoods.
  4408. \series default
  4409. This plot shows the distribution of distances from each annotated transcription
  4410. start site in the genome to the nearest called peak.
  4411. Each line represents one combination of histone mark, cell type, and time
  4412. point.
  4413. Distributions are smoothed using kernel density estimation [CITE? see ggplot2
  4414. stat_density()].
  4415. Transcription start sites that occur
  4416. \emph on
  4417. within
  4418. \emph default
  4419. peaks were excluded from this plot to avoid a large spike at zero that
  4420. would overshadow the rest of the distribution.
  4421. \end_layout
  4422. \end_inset
  4423. \end_layout
  4424. \end_inset
  4425. \end_layout
  4426. \begin_layout Standard
  4427. \begin_inset Float table
  4428. wide false
  4429. sideways false
  4430. status collapsed
  4431. \begin_layout Plain Layout
  4432. \align center
  4433. \begin_inset Tabular
  4434. <lyxtabular version="3" rows="4" columns="2">
  4435. <features tabularvalignment="middle">
  4436. <column alignment="center" valignment="top">
  4437. <column alignment="center" valignment="top">
  4438. <row>
  4439. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  4440. \begin_inset Text
  4441. \begin_layout Plain Layout
  4442. Histone mark
  4443. \end_layout
  4444. \end_inset
  4445. </cell>
  4446. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  4447. \begin_inset Text
  4448. \begin_layout Plain Layout
  4449. Effective promoter radius
  4450. \end_layout
  4451. \end_inset
  4452. </cell>
  4453. </row>
  4454. <row>
  4455. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  4456. \begin_inset Text
  4457. \begin_layout Plain Layout
  4458. H3K4me2
  4459. \end_layout
  4460. \end_inset
  4461. </cell>
  4462. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  4463. \begin_inset Text
  4464. \begin_layout Plain Layout
  4465. 1 kb
  4466. \end_layout
  4467. \end_inset
  4468. </cell>
  4469. </row>
  4470. <row>
  4471. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  4472. \begin_inset Text
  4473. \begin_layout Plain Layout
  4474. H3K4me3
  4475. \end_layout
  4476. \end_inset
  4477. </cell>
  4478. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  4479. \begin_inset Text
  4480. \begin_layout Plain Layout
  4481. 1 kb
  4482. \end_layout
  4483. \end_inset
  4484. </cell>
  4485. </row>
  4486. <row>
  4487. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  4488. \begin_inset Text
  4489. \begin_layout Plain Layout
  4490. H3K27me3
  4491. \end_layout
  4492. \end_inset
  4493. </cell>
  4494. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  4495. \begin_inset Text
  4496. \begin_layout Plain Layout
  4497. 2.5 kb
  4498. \end_layout
  4499. \end_inset
  4500. </cell>
  4501. </row>
  4502. </lyxtabular>
  4503. \end_inset
  4504. \end_layout
  4505. \begin_layout Plain Layout
  4506. \begin_inset Caption Standard
  4507. \begin_layout Plain Layout
  4508. \series bold
  4509. \begin_inset CommandInset label
  4510. LatexCommand label
  4511. name "tab:effective-promoter-radius"
  4512. \end_inset
  4513. Effective promoter radius for each histone mark.
  4514. \series default
  4515. These values represent the approximate distance from transcription start
  4516. site positions within which an excess of peaks are found, as shown in Figure
  4517. \begin_inset CommandInset ref
  4518. LatexCommand ref
  4519. reference "fig:near-promoter-peak-enrich"
  4520. plural "false"
  4521. caps "false"
  4522. noprefix "false"
  4523. \end_inset
  4524. .
  4525. \end_layout
  4526. \end_inset
  4527. \end_layout
  4528. \begin_layout Plain Layout
  4529. \end_layout
  4530. \end_inset
  4531. \end_layout
  4532. \begin_layout Standard
  4533. All 3 histone marks tend to occur more often near promoter regions, as shown
  4534. in Figure
  4535. \begin_inset CommandInset ref
  4536. LatexCommand ref
  4537. reference "fig:near-promoter-peak-enrich"
  4538. plural "false"
  4539. caps "false"
  4540. noprefix "false"
  4541. \end_inset
  4542. .
  4543. The majority of each density distribution is flat, representing the background
  4544. density of peaks genome-wide.
  4545. Each distribution has a peak near zero, representing an enrichment of peaks
  4546. close to
  4547. \begin_inset Flex Glossary Term
  4548. status open
  4549. \begin_layout Plain Layout
  4550. TSS
  4551. \end_layout
  4552. \end_inset
  4553. positions relative to the remainder of the genome.
  4554. Interestingly, the
  4555. \begin_inset Quotes eld
  4556. \end_inset
  4557. radius
  4558. \begin_inset Quotes erd
  4559. \end_inset
  4560. within which this enrichment occurs is not the same for every histone mark
  4561. (Table
  4562. \begin_inset CommandInset ref
  4563. LatexCommand ref
  4564. reference "tab:effective-promoter-radius"
  4565. plural "false"
  4566. caps "false"
  4567. noprefix "false"
  4568. \end_inset
  4569. ).
  4570. For H3K4me2 and H3K4me3, peaks are most enriched within 1
  4571. \begin_inset space ~
  4572. \end_inset
  4573. kbp of
  4574. \begin_inset Flex Glossary Term
  4575. status open
  4576. \begin_layout Plain Layout
  4577. TSS
  4578. \end_layout
  4579. \end_inset
  4580. positions, while for H3K27me3, enrichment is broader, extending to 2.5
  4581. \begin_inset space ~
  4582. \end_inset
  4583. kbp.
  4584. These
  4585. \begin_inset Quotes eld
  4586. \end_inset
  4587. effective promoter radii
  4588. \begin_inset Quotes erd
  4589. \end_inset
  4590. remain approximately the same across all combinations of experimental condition
  4591. (cell type, time point, and donor), so they appear to be a property of
  4592. the histone mark itself.
  4593. Hence, these radii were used to define the promoter regions for each histone
  4594. mark in all further analyses.
  4595. \end_layout
  4596. \begin_layout Standard
  4597. \begin_inset Flex TODO Note (inline)
  4598. status open
  4599. \begin_layout Plain Layout
  4600. Consider also showing figure for distance to nearest peak center, and reference
  4601. median peak size once that is known.
  4602. \end_layout
  4603. \end_inset
  4604. \end_layout
  4605. \begin_layout Subsection
  4606. H3K4 and H3K27 promoter methylation has broadly the expected correlation
  4607. with gene expression
  4608. \end_layout
  4609. \begin_layout Standard
  4610. \begin_inset Float figure
  4611. wide false
  4612. sideways false
  4613. status collapsed
  4614. \begin_layout Plain Layout
  4615. \begin_inset Flex TODO Note (inline)
  4616. status open
  4617. \begin_layout Plain Layout
  4618. This figure is generated from the old analysis.
  4619. Either note that in some way or re-generate it from the new peak calls.
  4620. \end_layout
  4621. \end_inset
  4622. \end_layout
  4623. \begin_layout Plain Layout
  4624. \align center
  4625. \begin_inset Graphics
  4626. filename graphics/CD4-csaw/FPKM by Peak Violin Plots-CROP.pdf
  4627. lyxscale 50
  4628. width 100col%
  4629. \end_inset
  4630. \end_layout
  4631. \begin_layout Plain Layout
  4632. \begin_inset Caption Standard
  4633. \begin_layout Plain Layout
  4634. \series bold
  4635. \begin_inset CommandInset label
  4636. LatexCommand label
  4637. name "fig:fpkm-by-peak"
  4638. \end_inset
  4639. Expression distributions of genes with and without promoter peaks.
  4640. \end_layout
  4641. \end_inset
  4642. \end_layout
  4643. \end_inset
  4644. \end_layout
  4645. \begin_layout Standard
  4646. H3K4me2 and H3K4me2 have previously been reported as activating marks whose
  4647. presence in a gene's promoter is associated with higher gene expression,
  4648. while H3K27me3 has been reported as inactivating [CITE].
  4649. The data are consistent with this characterization: genes whose promoters
  4650. (as defined by the radii for each histone mark listed in
  4651. \begin_inset CommandInset ref
  4652. LatexCommand ref
  4653. reference "tab:effective-promoter-radius"
  4654. plural "false"
  4655. caps "false"
  4656. noprefix "false"
  4657. \end_inset
  4658. ) overlap with a H3K4me2 or H3K4me3 peak tend to have higher expression
  4659. than those that don't, while H3K27me3 is likewise associated with lower
  4660. gene expression, as shown in
  4661. \begin_inset CommandInset ref
  4662. LatexCommand ref
  4663. reference "fig:fpkm-by-peak"
  4664. plural "false"
  4665. caps "false"
  4666. noprefix "false"
  4667. \end_inset
  4668. .
  4669. This pattern holds across all combinations of cell type and time point
  4670. (Welch's
  4671. \emph on
  4672. t
  4673. \emph default
  4674. -test, all
  4675. \begin_inset Formula $p\mathrm{-values}\ll2.2\times10^{-16}$
  4676. \end_inset
  4677. ).
  4678. The difference in average
  4679. \begin_inset Formula $\log_{2}$
  4680. \end_inset
  4681. \begin_inset Flex Glossary Term
  4682. status open
  4683. \begin_layout Plain Layout
  4684. FPKM
  4685. \end_layout
  4686. \end_inset
  4687. \begin_inset CommandInset nomenclature
  4688. LatexCommand nomenclature
  4689. symbol "FPKM"
  4690. description "fragments per kilobase per million fragments"
  4691. literal "false"
  4692. \end_inset
  4693. values when a peak overlaps the promoter is about
  4694. \begin_inset Formula $+5.67$
  4695. \end_inset
  4696. for H3K4me2,
  4697. \begin_inset Formula $+5.76$
  4698. \end_inset
  4699. for H3K4me2, and
  4700. \begin_inset Formula $-4.00$
  4701. \end_inset
  4702. for H3K27me3.
  4703. \end_layout
  4704. \begin_layout Standard
  4705. \begin_inset Flex TODO Note (inline)
  4706. status open
  4707. \begin_layout Plain Layout
  4708. I also have some figures looking at interactions between marks (e.g.
  4709. what if a promoter has both H3K4me3 and H3K27me3), but I don't know if
  4710. that much detail is warranted here, since all the effects just seem approximate
  4711. ly additive anyway.
  4712. \end_layout
  4713. \end_inset
  4714. \end_layout
  4715. \begin_layout Subsection
  4716. Gene expression and promoter histone methylation patterns in naïve and memory
  4717. show convergence at day 14
  4718. \end_layout
  4719. \begin_layout Standard
  4720. \begin_inset ERT
  4721. status open
  4722. \begin_layout Plain Layout
  4723. \backslash
  4724. afterpage{
  4725. \end_layout
  4726. \begin_layout Plain Layout
  4727. \backslash
  4728. begin{landscape}
  4729. \end_layout
  4730. \end_inset
  4731. \end_layout
  4732. \begin_layout Standard
  4733. \begin_inset Float table
  4734. wide false
  4735. sideways false
  4736. status open
  4737. \begin_layout Plain Layout
  4738. \align center
  4739. \begin_inset Tabular
  4740. <lyxtabular version="3" rows="6" columns="7">
  4741. <features tabularvalignment="middle">
  4742. <column alignment="center" valignment="top">
  4743. <column alignment="center" valignment="top">
  4744. <column alignment="center" valignment="top">
  4745. <column alignment="center" valignment="top">
  4746. <column alignment="center" valignment="top">
  4747. <column alignment="center" valignment="top">
  4748. <column alignment="center" valignment="top">
  4749. <row>
  4750. <cell alignment="center" valignment="top" usebox="none">
  4751. \begin_inset Text
  4752. \begin_layout Plain Layout
  4753. \end_layout
  4754. \end_inset
  4755. </cell>
  4756. <cell multicolumn="1" alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  4757. \begin_inset Text
  4758. \begin_layout Plain Layout
  4759. Number of significant promoters
  4760. \end_layout
  4761. \end_inset
  4762. </cell>
  4763. <cell multicolumn="2" alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  4764. \begin_inset Text
  4765. \begin_layout Plain Layout
  4766. \end_layout
  4767. \end_inset
  4768. </cell>
  4769. <cell multicolumn="2" alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  4770. \begin_inset Text
  4771. \begin_layout Plain Layout
  4772. \end_layout
  4773. \end_inset
  4774. </cell>
  4775. <cell multicolumn="1" alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  4776. \begin_inset Text
  4777. \begin_layout Plain Layout
  4778. Est.
  4779. differentially modified promoters
  4780. \end_layout
  4781. \end_inset
  4782. </cell>
  4783. <cell multicolumn="2" alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  4784. \begin_inset Text
  4785. \begin_layout Plain Layout
  4786. \end_layout
  4787. \end_inset
  4788. </cell>
  4789. <cell multicolumn="2" alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  4790. \begin_inset Text
  4791. \begin_layout Plain Layout
  4792. \end_layout
  4793. \end_inset
  4794. </cell>
  4795. </row>
  4796. <row>
  4797. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  4798. \begin_inset Text
  4799. \begin_layout Plain Layout
  4800. Time Point
  4801. \end_layout
  4802. \end_inset
  4803. </cell>
  4804. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  4805. \begin_inset Text
  4806. \begin_layout Plain Layout
  4807. H3K4me2
  4808. \end_layout
  4809. \end_inset
  4810. </cell>
  4811. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  4812. \begin_inset Text
  4813. \begin_layout Plain Layout
  4814. H3K4me3
  4815. \end_layout
  4816. \end_inset
  4817. </cell>
  4818. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  4819. \begin_inset Text
  4820. \begin_layout Plain Layout
  4821. H3K27me3
  4822. \end_layout
  4823. \end_inset
  4824. </cell>
  4825. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  4826. \begin_inset Text
  4827. \begin_layout Plain Layout
  4828. H3K4me2
  4829. \end_layout
  4830. \end_inset
  4831. </cell>
  4832. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  4833. \begin_inset Text
  4834. \begin_layout Plain Layout
  4835. H3K4me3
  4836. \end_layout
  4837. \end_inset
  4838. </cell>
  4839. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  4840. \begin_inset Text
  4841. \begin_layout Plain Layout
  4842. H3K27me3
  4843. \end_layout
  4844. \end_inset
  4845. </cell>
  4846. </row>
  4847. <row>
  4848. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  4849. \begin_inset Text
  4850. \begin_layout Plain Layout
  4851. Day 0
  4852. \end_layout
  4853. \end_inset
  4854. </cell>
  4855. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  4856. \begin_inset Text
  4857. \begin_layout Plain Layout
  4858. 4553
  4859. \end_layout
  4860. \end_inset
  4861. </cell>
  4862. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  4863. \begin_inset Text
  4864. \begin_layout Plain Layout
  4865. 927
  4866. \end_layout
  4867. \end_inset
  4868. </cell>
  4869. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  4870. \begin_inset Text
  4871. \begin_layout Plain Layout
  4872. 6
  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. 9967
  4880. \end_layout
  4881. \end_inset
  4882. </cell>
  4883. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  4884. \begin_inset Text
  4885. \begin_layout Plain Layout
  4886. 4149
  4887. \end_layout
  4888. \end_inset
  4889. </cell>
  4890. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  4891. \begin_inset Text
  4892. \begin_layout Plain Layout
  4893. 2404
  4894. \end_layout
  4895. \end_inset
  4896. </cell>
  4897. </row>
  4898. <row>
  4899. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  4900. \begin_inset Text
  4901. \begin_layout Plain Layout
  4902. Day 1
  4903. \end_layout
  4904. \end_inset
  4905. </cell>
  4906. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  4907. \begin_inset Text
  4908. \begin_layout Plain Layout
  4909. 567
  4910. \end_layout
  4911. \end_inset
  4912. </cell>
  4913. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  4914. \begin_inset Text
  4915. \begin_layout Plain Layout
  4916. 278
  4917. \end_layout
  4918. \end_inset
  4919. </cell>
  4920. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  4921. \begin_inset Text
  4922. \begin_layout Plain Layout
  4923. 1570
  4924. \end_layout
  4925. \end_inset
  4926. </cell>
  4927. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  4928. \begin_inset Text
  4929. \begin_layout Plain Layout
  4930. 4370
  4931. \end_layout
  4932. \end_inset
  4933. </cell>
  4934. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  4935. \begin_inset Text
  4936. \begin_layout Plain Layout
  4937. 2145
  4938. \end_layout
  4939. \end_inset
  4940. </cell>
  4941. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  4942. \begin_inset Text
  4943. \begin_layout Plain Layout
  4944. 6598
  4945. \end_layout
  4946. \end_inset
  4947. </cell>
  4948. </row>
  4949. <row>
  4950. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  4951. \begin_inset Text
  4952. \begin_layout Plain Layout
  4953. Day 5
  4954. \end_layout
  4955. \end_inset
  4956. </cell>
  4957. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  4958. \begin_inset Text
  4959. \begin_layout Plain Layout
  4960. 2313
  4961. \end_layout
  4962. \end_inset
  4963. </cell>
  4964. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  4965. \begin_inset Text
  4966. \begin_layout Plain Layout
  4967. 139
  4968. \end_layout
  4969. \end_inset
  4970. </cell>
  4971. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  4972. \begin_inset Text
  4973. \begin_layout Plain Layout
  4974. 490
  4975. \end_layout
  4976. \end_inset
  4977. </cell>
  4978. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  4979. \begin_inset Text
  4980. \begin_layout Plain Layout
  4981. 9450
  4982. \end_layout
  4983. \end_inset
  4984. </cell>
  4985. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  4986. \begin_inset Text
  4987. \begin_layout Plain Layout
  4988. 1148
  4989. \end_layout
  4990. \end_inset
  4991. </cell>
  4992. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  4993. \begin_inset Text
  4994. \begin_layout Plain Layout
  4995. 4141
  4996. \end_layout
  4997. \end_inset
  4998. </cell>
  4999. </row>
  5000. <row>
  5001. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  5002. \begin_inset Text
  5003. \begin_layout Plain Layout
  5004. Day 14
  5005. \end_layout
  5006. \end_inset
  5007. </cell>
  5008. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  5009. \begin_inset Text
  5010. \begin_layout Plain Layout
  5011. 0
  5012. \end_layout
  5013. \end_inset
  5014. </cell>
  5015. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  5016. \begin_inset Text
  5017. \begin_layout Plain Layout
  5018. 0
  5019. \end_layout
  5020. \end_inset
  5021. </cell>
  5022. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  5023. \begin_inset Text
  5024. \begin_layout Plain Layout
  5025. 0
  5026. \end_layout
  5027. \end_inset
  5028. </cell>
  5029. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  5030. \begin_inset Text
  5031. \begin_layout Plain Layout
  5032. 0
  5033. \end_layout
  5034. \end_inset
  5035. </cell>
  5036. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  5037. \begin_inset Text
  5038. \begin_layout Plain Layout
  5039. 0
  5040. \end_layout
  5041. \end_inset
  5042. </cell>
  5043. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  5044. \begin_inset Text
  5045. \begin_layout Plain Layout
  5046. 0
  5047. \end_layout
  5048. \end_inset
  5049. </cell>
  5050. </row>
  5051. </lyxtabular>
  5052. \end_inset
  5053. \end_layout
  5054. \begin_layout Plain Layout
  5055. \begin_inset Caption Standard
  5056. \begin_layout Plain Layout
  5057. \series bold
  5058. \begin_inset CommandInset label
  5059. LatexCommand label
  5060. name "tab:Number-signif-promoters"
  5061. \end_inset
  5062. Number of differentially modified promoters between naïve and memory cells
  5063. at each time point after activation.
  5064. \series default
  5065. This table shows both the number of differentially modified promoters detected
  5066. at a 10% FDR threshold (left half), and the total number of differentially
  5067. modified promoters as estimated using the method of
  5068. \begin_inset CommandInset citation
  5069. LatexCommand cite
  5070. key "Phipson2013"
  5071. literal "false"
  5072. \end_inset
  5073. (right half).
  5074. \end_layout
  5075. \end_inset
  5076. \end_layout
  5077. \end_inset
  5078. \end_layout
  5079. \begin_layout Standard
  5080. \begin_inset ERT
  5081. status open
  5082. \begin_layout Plain Layout
  5083. \backslash
  5084. end{landscape}
  5085. \end_layout
  5086. \begin_layout Plain Layout
  5087. }
  5088. \end_layout
  5089. \end_inset
  5090. \end_layout
  5091. \begin_layout Standard
  5092. \begin_inset Float figure
  5093. placement p
  5094. wide false
  5095. sideways false
  5096. status open
  5097. \begin_layout Plain Layout
  5098. \align center
  5099. \begin_inset Float figure
  5100. wide false
  5101. sideways false
  5102. status open
  5103. \begin_layout Plain Layout
  5104. \align center
  5105. \begin_inset Graphics
  5106. filename graphics/CD4-csaw/ChIP-seq/H3K4me2-promoter-PCA-group-CROP.png
  5107. lyxscale 25
  5108. width 45col%
  5109. groupId pcoa-prom-subfig
  5110. \end_inset
  5111. \end_layout
  5112. \begin_layout Plain Layout
  5113. \begin_inset Caption Standard
  5114. \begin_layout Plain Layout
  5115. \series bold
  5116. \begin_inset CommandInset label
  5117. LatexCommand label
  5118. name "fig:PCoA-H3K4me2-prom"
  5119. \end_inset
  5120. PCoA plot of H3K4me2 promoters, after subtracting surrogate variables
  5121. \end_layout
  5122. \end_inset
  5123. \end_layout
  5124. \end_inset
  5125. \begin_inset space \hfill{}
  5126. \end_inset
  5127. \begin_inset Float figure
  5128. wide false
  5129. sideways false
  5130. status open
  5131. \begin_layout Plain Layout
  5132. \align center
  5133. \begin_inset Graphics
  5134. filename graphics/CD4-csaw/ChIP-seq/H3K4me3-promoter-PCA-group-CROP.png
  5135. lyxscale 25
  5136. width 45col%
  5137. groupId pcoa-prom-subfig
  5138. \end_inset
  5139. \end_layout
  5140. \begin_layout Plain Layout
  5141. \begin_inset Caption Standard
  5142. \begin_layout Plain Layout
  5143. \series bold
  5144. \begin_inset CommandInset label
  5145. LatexCommand label
  5146. name "fig:PCoA-H3K4me3-prom"
  5147. \end_inset
  5148. PCoA plot of H3K4me3 promoters, after subtracting surrogate variables
  5149. \end_layout
  5150. \end_inset
  5151. \end_layout
  5152. \end_inset
  5153. \end_layout
  5154. \begin_layout Plain Layout
  5155. \align center
  5156. \begin_inset Float figure
  5157. wide false
  5158. sideways false
  5159. status collapsed
  5160. \begin_layout Plain Layout
  5161. \align center
  5162. \begin_inset Graphics
  5163. filename graphics/CD4-csaw/ChIP-seq/H3K27me3-promoter-PCA-group-CROP.png
  5164. lyxscale 25
  5165. width 45col%
  5166. groupId pcoa-prom-subfig
  5167. \end_inset
  5168. \end_layout
  5169. \begin_layout Plain Layout
  5170. \begin_inset Caption Standard
  5171. \begin_layout Plain Layout
  5172. \series bold
  5173. \begin_inset CommandInset label
  5174. LatexCommand label
  5175. name "fig:PCoA-H3K27me3-prom"
  5176. \end_inset
  5177. PCoA plot of H3K27me3 promoters, after subtracting surrogate variables
  5178. \end_layout
  5179. \end_inset
  5180. \end_layout
  5181. \end_inset
  5182. \begin_inset space \hfill{}
  5183. \end_inset
  5184. \begin_inset Float figure
  5185. wide false
  5186. sideways false
  5187. status open
  5188. \begin_layout Plain Layout
  5189. \align center
  5190. \begin_inset Graphics
  5191. filename graphics/CD4-csaw/RNA-seq/PCA-final-23-CROP.png
  5192. lyxscale 25
  5193. width 45col%
  5194. groupId pcoa-prom-subfig
  5195. \end_inset
  5196. \end_layout
  5197. \begin_layout Plain Layout
  5198. \begin_inset Caption Standard
  5199. \begin_layout Plain Layout
  5200. \series bold
  5201. \begin_inset CommandInset label
  5202. LatexCommand label
  5203. name "fig:RNA-PCA-group"
  5204. \end_inset
  5205. RNA-seq PCoA showing principal coordinates 2 and 3.
  5206. \end_layout
  5207. \end_inset
  5208. \end_layout
  5209. \end_inset
  5210. \end_layout
  5211. \begin_layout Plain Layout
  5212. \begin_inset Caption Standard
  5213. \begin_layout Plain Layout
  5214. \series bold
  5215. \begin_inset CommandInset label
  5216. LatexCommand label
  5217. name "fig:PCoA-promoters"
  5218. \end_inset
  5219. PCoA plots for promoter ChIP-seq and expression RNA-seq data
  5220. \end_layout
  5221. \end_inset
  5222. \end_layout
  5223. \end_inset
  5224. \end_layout
  5225. \begin_layout Standard
  5226. \begin_inset Flex TODO Note (inline)
  5227. status open
  5228. \begin_layout Plain Layout
  5229. Check up on figure refs in this paragraph
  5230. \end_layout
  5231. \end_inset
  5232. \end_layout
  5233. \begin_layout Standard
  5234. We hypothesized that if naïve cells had differentiated into memory cells
  5235. by Day 14, then their patterns of expression and histone modification should
  5236. converge with those of memory cells at Day 14.
  5237. Figure
  5238. \begin_inset CommandInset ref
  5239. LatexCommand ref
  5240. reference "fig:PCoA-promoters"
  5241. plural "false"
  5242. caps "false"
  5243. noprefix "false"
  5244. \end_inset
  5245. shows the patterns of variation in all 3 histone marks in the promoter
  5246. regions of the genome using
  5247. \begin_inset Flex Glossary Term
  5248. status open
  5249. \begin_layout Plain Layout
  5250. PCoA
  5251. \end_layout
  5252. \end_inset
  5253. \begin_inset CommandInset nomenclature
  5254. LatexCommand nomenclature
  5255. symbol "PCoA"
  5256. description "principal coordinate analysis"
  5257. literal "false"
  5258. \end_inset
  5259. .
  5260. All 3 marks show a noticeable convergence between the naïve and memory
  5261. samples at day 14, visible as an overlapping of the day 14 groups on each
  5262. plot.
  5263. This is consistent with the counts of significantly differentially modified
  5264. promoters and estimates of the total numbers of differentially modified
  5265. promoters shown in Table
  5266. \begin_inset CommandInset ref
  5267. LatexCommand ref
  5268. reference "tab:Number-signif-promoters"
  5269. plural "false"
  5270. caps "false"
  5271. noprefix "false"
  5272. \end_inset
  5273. .
  5274. For all histone marks, evidence of differential modification between naïve
  5275. and memory samples was detected at every time point except day 14.
  5276. The day 14 convergence pattern is also present in the
  5277. \begin_inset Flex Glossary Term
  5278. status open
  5279. \begin_layout Plain Layout
  5280. RNA-seq
  5281. \end_layout
  5282. \end_inset
  5283. data (Figure
  5284. \begin_inset CommandInset ref
  5285. LatexCommand ref
  5286. reference "fig:RNA-PCA-group"
  5287. plural "false"
  5288. caps "false"
  5289. noprefix "false"
  5290. \end_inset
  5291. ), albeit in the 2nd and 3rd principal coordinates, indicating that it is
  5292. not the most dominant pattern driving gene expression.
  5293. Taken together, the data show that promoter histone methylation for these
  5294. 3 histone marks and RNA expression for naïve and memory cells are most
  5295. similar at day 14, the furthest time point after activation.
  5296. \begin_inset Flex Glossary Term
  5297. status open
  5298. \begin_layout Plain Layout
  5299. MOFA
  5300. \end_layout
  5301. \end_inset
  5302. was also able to capture this day 14 convergence pattern in
  5303. \begin_inset Flex Glossary Term
  5304. status open
  5305. \begin_layout Plain Layout
  5306. LF
  5307. \end_layout
  5308. \end_inset
  5309. 5 (Figure
  5310. \begin_inset CommandInset ref
  5311. LatexCommand ref
  5312. reference "fig:mofa-lf-scatter"
  5313. plural "false"
  5314. caps "false"
  5315. noprefix "false"
  5316. \end_inset
  5317. ), which accounts for shared variation across all 3 histone marks and the
  5318. \begin_inset Flex Glossary Term
  5319. status open
  5320. \begin_layout Plain Layout
  5321. RNA-seq
  5322. \end_layout
  5323. \end_inset
  5324. data, confirming that this convergence is a coordinated pattern across
  5325. all 4 data sets.
  5326. While this observation does not prove that the naïve cells have differentiated
  5327. into memory cells at Day 14, it is consistent with that hypothesis.
  5328. \end_layout
  5329. \begin_layout Subsection
  5330. Effect of H3K4me2 and H3K4me3 promoter coverage upstream vs downstream of
  5331. TSS
  5332. \end_layout
  5333. \begin_layout Standard
  5334. \begin_inset Flex TODO Note (inline)
  5335. status open
  5336. \begin_layout Plain Layout
  5337. Need a better section title, for this and the next one.
  5338. \end_layout
  5339. \end_inset
  5340. \end_layout
  5341. \begin_layout Standard
  5342. \begin_inset Flex TODO Note (inline)
  5343. status open
  5344. \begin_layout Plain Layout
  5345. Make sure use of coverage/abundance/whatever is consistent.
  5346. \end_layout
  5347. \end_inset
  5348. \end_layout
  5349. \begin_layout Standard
  5350. \begin_inset Flex TODO Note (inline)
  5351. status open
  5352. \begin_layout Plain Layout
  5353. For the figures in this section and the next, the group labels are arbitrary,
  5354. so if time allows, it would be good to manually reorder them in a logical
  5355. way, e.g.
  5356. most upstream to most downstream.
  5357. If this is done, make sure to update the text with the correct group labels.
  5358. \end_layout
  5359. \end_inset
  5360. \end_layout
  5361. \begin_layout Standard
  5362. \begin_inset ERT
  5363. status open
  5364. \begin_layout Plain Layout
  5365. \backslash
  5366. afterpage{
  5367. \end_layout
  5368. \begin_layout Plain Layout
  5369. \backslash
  5370. begin{landscape}
  5371. \end_layout
  5372. \end_inset
  5373. \end_layout
  5374. \begin_layout Standard
  5375. \begin_inset Float figure
  5376. wide false
  5377. sideways false
  5378. status open
  5379. \begin_layout Plain Layout
  5380. \align center
  5381. \begin_inset Float figure
  5382. wide false
  5383. sideways false
  5384. status open
  5385. \begin_layout Plain Layout
  5386. \align center
  5387. \begin_inset Graphics
  5388. filename graphics/CD4-csaw/ChIP-seq/H3K4me2-neighborhood-clusters-CROP.png
  5389. lyxscale 25
  5390. width 30col%
  5391. groupId covprof-subfig
  5392. \end_inset
  5393. \end_layout
  5394. \begin_layout Plain Layout
  5395. \begin_inset Caption Standard
  5396. \begin_layout Plain Layout
  5397. \series bold
  5398. \begin_inset CommandInset label
  5399. LatexCommand label
  5400. name "fig:H3K4me2-neighborhood-clusters"
  5401. \end_inset
  5402. Average relative coverage for each bin in each cluster
  5403. \end_layout
  5404. \end_inset
  5405. \end_layout
  5406. \end_inset
  5407. \begin_inset space \hfill{}
  5408. \end_inset
  5409. \begin_inset Float figure
  5410. wide false
  5411. sideways false
  5412. status open
  5413. \begin_layout Plain Layout
  5414. \align center
  5415. \begin_inset Graphics
  5416. filename graphics/CD4-csaw/ChIP-seq/H3K4me2-neighborhood-PCA-CROP.png
  5417. lyxscale 25
  5418. width 30col%
  5419. groupId covprof-subfig
  5420. \end_inset
  5421. \end_layout
  5422. \begin_layout Plain Layout
  5423. \begin_inset Caption Standard
  5424. \begin_layout Plain Layout
  5425. \series bold
  5426. \begin_inset CommandInset label
  5427. LatexCommand label
  5428. name "fig:H3K4me2-neighborhood-pca"
  5429. \end_inset
  5430. PCA of relative coverage depth, colored by K-means cluster membership.
  5431. \end_layout
  5432. \end_inset
  5433. \end_layout
  5434. \end_inset
  5435. \begin_inset space \hfill{}
  5436. \end_inset
  5437. \begin_inset Float figure
  5438. wide false
  5439. sideways false
  5440. status open
  5441. \begin_layout Plain Layout
  5442. \align center
  5443. \begin_inset Graphics
  5444. filename graphics/CD4-csaw/ChIP-seq/H3K4me2-neighborhood-expression-CROP.png
  5445. lyxscale 25
  5446. width 30col%
  5447. groupId covprof-subfig
  5448. \end_inset
  5449. \end_layout
  5450. \begin_layout Plain Layout
  5451. \begin_inset Caption Standard
  5452. \begin_layout Plain Layout
  5453. \series bold
  5454. \begin_inset CommandInset label
  5455. LatexCommand label
  5456. name "fig:H3K4me2-neighborhood-expression"
  5457. \end_inset
  5458. Gene expression grouped by promoter coverage clusters.
  5459. \end_layout
  5460. \end_inset
  5461. \end_layout
  5462. \end_inset
  5463. \end_layout
  5464. \begin_layout Plain Layout
  5465. \begin_inset Caption Standard
  5466. \begin_layout Plain Layout
  5467. \series bold
  5468. \begin_inset CommandInset label
  5469. LatexCommand label
  5470. name "fig:H3K4me2-neighborhood"
  5471. \end_inset
  5472. K-means clustering of promoter H3K4me2 relative coverage depth in naïve
  5473. day 0 samples.
  5474. \series default
  5475. H3K4me2 ChIP-seq reads were binned into 500-bp windows tiled across each
  5476. promoter from 5
  5477. \begin_inset space ~
  5478. \end_inset
  5479. kbp upstream to 5
  5480. \begin_inset space ~
  5481. \end_inset
  5482. kbp downstream, and the logCPM values were normalized within each promoter
  5483. to an average of 0, yielding relative coverage depths.
  5484. These were then grouped using K-means clustering with
  5485. \begin_inset Formula $K=6$
  5486. \end_inset
  5487. ,
  5488. \series bold
  5489. \series default
  5490. and the average bin values were plotted for each cluster (a).
  5491. The
  5492. \begin_inset Formula $x$
  5493. \end_inset
  5494. -axis is the genomic coordinate of each bin relative to the the transcription
  5495. start site, and the
  5496. \begin_inset Formula $y$
  5497. \end_inset
  5498. -axis is the mean relative coverage depth of that bin across all promoters
  5499. in the cluster.
  5500. Each line represents the average
  5501. \begin_inset Quotes eld
  5502. \end_inset
  5503. shape
  5504. \begin_inset Quotes erd
  5505. \end_inset
  5506. of the promoter coverage for promoters in that cluster.
  5507. PCA was performed on the same data, and the first two PCs were plotted,
  5508. coloring each point by its K-means cluster identity (b).
  5509. For each cluster, the distribution of gene expression values was plotted
  5510. (c).
  5511. \end_layout
  5512. \end_inset
  5513. \end_layout
  5514. \end_inset
  5515. \end_layout
  5516. \begin_layout Standard
  5517. \begin_inset ERT
  5518. status open
  5519. \begin_layout Plain Layout
  5520. \backslash
  5521. end{landscape}
  5522. \end_layout
  5523. \begin_layout Plain Layout
  5524. }
  5525. \end_layout
  5526. \end_inset
  5527. \end_layout
  5528. \begin_layout Standard
  5529. To test whether the position of a histone mark relative to a gene's
  5530. \begin_inset Flex Glossary Term
  5531. status open
  5532. \begin_layout Plain Layout
  5533. TSS
  5534. \end_layout
  5535. \end_inset
  5536. was important, we looked at the
  5537. \begin_inset Quotes eld
  5538. \end_inset
  5539. landscape
  5540. \begin_inset Quotes erd
  5541. \end_inset
  5542. of
  5543. \begin_inset Flex Glossary Term
  5544. status open
  5545. \begin_layout Plain Layout
  5546. ChIP-seq
  5547. \end_layout
  5548. \end_inset
  5549. read coverage in naïve Day 0 samples within 5 kb of each gene's
  5550. \begin_inset Flex Glossary Term
  5551. status open
  5552. \begin_layout Plain Layout
  5553. TSS
  5554. \end_layout
  5555. \end_inset
  5556. by binning reads into 500-bp windows tiled across each promoter
  5557. \begin_inset Flex Glossary Term
  5558. status open
  5559. \begin_layout Plain Layout
  5560. logCPM
  5561. \end_layout
  5562. \end_inset
  5563. values were calculated for the bins in each promoter and then the average
  5564. \begin_inset Flex Glossary Term
  5565. status open
  5566. \begin_layout Plain Layout
  5567. logCPM
  5568. \end_layout
  5569. \end_inset
  5570. for each promoter's bins was normalized to zero, such that the values represent
  5571. coverage relative to other regions of the same promoter rather than being
  5572. proportional to absolute read count.
  5573. The promoters were then clustered based on the normalized bin abundances
  5574. using
  5575. \begin_inset Formula $k$
  5576. \end_inset
  5577. -means clustering with
  5578. \begin_inset Formula $K=6$
  5579. \end_inset
  5580. .
  5581. Different values of
  5582. \begin_inset Formula $K$
  5583. \end_inset
  5584. were also tested, but did not substantially change the interpretation of
  5585. the data.
  5586. \end_layout
  5587. \begin_layout Standard
  5588. For H3K4me2, plotting the average bin abundances for each cluster reveals
  5589. a simple pattern (Figure
  5590. \begin_inset CommandInset ref
  5591. LatexCommand ref
  5592. reference "fig:H3K4me2-neighborhood-clusters"
  5593. plural "false"
  5594. caps "false"
  5595. noprefix "false"
  5596. \end_inset
  5597. ): Cluster 5 represents a completely flat promoter coverage profile, likely
  5598. consisting of genes with no H3K4me2 methylation in the promoter.
  5599. All the other clusters represent a continuum of peak positions relative
  5600. to the
  5601. \begin_inset Flex Glossary Term
  5602. status open
  5603. \begin_layout Plain Layout
  5604. TSS
  5605. \end_layout
  5606. \end_inset
  5607. .
  5608. In order from must upstream to most downstream, they are Clusters 6, 4,
  5609. 3, 1, and 2.
  5610. There do not appear to be any clusters representing coverage patterns other
  5611. than lone peaks, such as coverage troughs or double peaks.
  5612. Next, all promoters were plotted in a
  5613. \begin_inset Flex Glossary Term
  5614. status open
  5615. \begin_layout Plain Layout
  5616. PCA
  5617. \end_layout
  5618. \end_inset
  5619. \begin_inset CommandInset nomenclature
  5620. LatexCommand nomenclature
  5621. symbol "PCA"
  5622. description "principal component analysis"
  5623. literal "false"
  5624. \end_inset
  5625. plot based on the same relative bin abundance data, and colored based on
  5626. cluster membership (Figure
  5627. \begin_inset CommandInset ref
  5628. LatexCommand ref
  5629. reference "fig:H3K4me2-neighborhood-pca"
  5630. plural "false"
  5631. caps "false"
  5632. noprefix "false"
  5633. \end_inset
  5634. ).
  5635. The
  5636. \begin_inset Flex Glossary Term
  5637. status open
  5638. \begin_layout Plain Layout
  5639. PCA
  5640. \end_layout
  5641. \end_inset
  5642. plot shows Cluster 5 (the
  5643. \begin_inset Quotes eld
  5644. \end_inset
  5645. no peak
  5646. \begin_inset Quotes erd
  5647. \end_inset
  5648. cluster) at the center, with the other clusters arranged in a counter-clockwise
  5649. arc around it in the order noted above, from most upstream peak to most
  5650. downstream.
  5651. Notably, the
  5652. \begin_inset Quotes eld
  5653. \end_inset
  5654. clusters
  5655. \begin_inset Quotes erd
  5656. \end_inset
  5657. form a single large
  5658. \begin_inset Quotes eld
  5659. \end_inset
  5660. cloud
  5661. \begin_inset Quotes erd
  5662. \end_inset
  5663. with no apparent separation between them, further supporting the conclusion
  5664. that these clusters represent an arbitrary partitioning of a continuous
  5665. distribution of promoter coverage landscapes.
  5666. While the clusters are a useful abstraction that aids in visualization,
  5667. they are ultimately not an accurate representation of the data.
  5668. A better representation might be something like a polar coordinate system
  5669. with the origin at the center of Cluster 5, where the radius represents
  5670. the peak height above the background and the angle represents the peak's
  5671. position upstream or downstream of the
  5672. \begin_inset Flex Glossary Term
  5673. status open
  5674. \begin_layout Plain Layout
  5675. TSS
  5676. \end_layout
  5677. \end_inset
  5678. .
  5679. The continuous nature of the distribution also explains why different values
  5680. of
  5681. \begin_inset Formula $K$
  5682. \end_inset
  5683. led to similar conclusions.
  5684. \end_layout
  5685. \begin_layout Standard
  5686. \begin_inset Flex TODO Note (inline)
  5687. status open
  5688. \begin_layout Plain Layout
  5689. RNA-seq values in the plots use logCPM but should really use logFPKM or
  5690. logTPM.
  5691. Fix if time allows.
  5692. \end_layout
  5693. \end_inset
  5694. \end_layout
  5695. \begin_layout Standard
  5696. \begin_inset Flex TODO Note (inline)
  5697. status open
  5698. \begin_layout Plain Layout
  5699. Should have a table of p-values on difference of means between Cluster 5
  5700. and the others.
  5701. \end_layout
  5702. \end_inset
  5703. \end_layout
  5704. \begin_layout Standard
  5705. To investigate the association between relative peak position and gene expressio
  5706. n, we plotted the Naïve Day 0 expression for the genes in each cluster (Figure
  5707. \begin_inset CommandInset ref
  5708. LatexCommand ref
  5709. reference "fig:H3K4me2-neighborhood-expression"
  5710. plural "false"
  5711. caps "false"
  5712. noprefix "false"
  5713. \end_inset
  5714. ).
  5715. Most genes in Cluster 5, the
  5716. \begin_inset Quotes eld
  5717. \end_inset
  5718. no peak
  5719. \begin_inset Quotes erd
  5720. \end_inset
  5721. cluster, have low expression values.
  5722. Taking this as the
  5723. \begin_inset Quotes eld
  5724. \end_inset
  5725. baseline
  5726. \begin_inset Quotes erd
  5727. \end_inset
  5728. distribution when no H3K4me2 methylation is present, we can compare the
  5729. other clusters' distributions to determine which peak positions are associated
  5730. with elevated expression.
  5731. As might be expected, the 3 clusters representing peaks closest to the
  5732. \begin_inset Flex Glossary Term
  5733. status open
  5734. \begin_layout Plain Layout
  5735. TSS
  5736. \end_layout
  5737. \end_inset
  5738. , Clusters 1, 3, and 4, show the highest average expression distributions.
  5739. Specifically, these clusters all have their highest
  5740. \begin_inset Flex Glossary Term
  5741. status open
  5742. \begin_layout Plain Layout
  5743. ChIP-seq
  5744. \end_layout
  5745. \end_inset
  5746. abundance within 1kb of the
  5747. \begin_inset Flex Glossary Term
  5748. status open
  5749. \begin_layout Plain Layout
  5750. TSS
  5751. \end_layout
  5752. \end_inset
  5753. , consistent with the previously determined promoter radius.
  5754. In contrast, cluster 6, which represents peaks several kb upstream of the
  5755. \begin_inset Flex Glossary Term
  5756. status open
  5757. \begin_layout Plain Layout
  5758. TSS
  5759. \end_layout
  5760. \end_inset
  5761. , shows a slightly higher average expression than baseline, while Cluster
  5762. 2, which represents peaks several kb downstream, doesn't appear to show
  5763. any appreciable difference.
  5764. Interestingly, the cluster with the highest average expression is Cluster
  5765. 1, which represents peaks about 1 kb downstream of the
  5766. \begin_inset Flex Glossary Term
  5767. status open
  5768. \begin_layout Plain Layout
  5769. TSS
  5770. \end_layout
  5771. \end_inset
  5772. , rather than Cluster 3, which represents peaks centered directly at the
  5773. \begin_inset Flex Glossary Term
  5774. status open
  5775. \begin_layout Plain Layout
  5776. TSS
  5777. \end_layout
  5778. \end_inset
  5779. .
  5780. This suggests that conceptualizing the promoter as a region centered on
  5781. the
  5782. \begin_inset Flex Glossary Term
  5783. status open
  5784. \begin_layout Plain Layout
  5785. TSS
  5786. \end_layout
  5787. \end_inset
  5788. with a certain
  5789. \begin_inset Quotes eld
  5790. \end_inset
  5791. radius
  5792. \begin_inset Quotes erd
  5793. \end_inset
  5794. may be an oversimplification – a peak that is a specific distance from
  5795. the
  5796. \begin_inset Flex Glossary Term
  5797. status open
  5798. \begin_layout Plain Layout
  5799. TSS
  5800. \end_layout
  5801. \end_inset
  5802. may have a different degree of influence depending on whether it is upstream
  5803. or downstream of the
  5804. \begin_inset Flex Glossary Term
  5805. status open
  5806. \begin_layout Plain Layout
  5807. TSS
  5808. \end_layout
  5809. \end_inset
  5810. .
  5811. \end_layout
  5812. \begin_layout Standard
  5813. \begin_inset ERT
  5814. status open
  5815. \begin_layout Plain Layout
  5816. \backslash
  5817. afterpage{
  5818. \end_layout
  5819. \begin_layout Plain Layout
  5820. \backslash
  5821. begin{landscape}
  5822. \end_layout
  5823. \end_inset
  5824. \end_layout
  5825. \begin_layout Standard
  5826. \begin_inset Float figure
  5827. wide false
  5828. sideways false
  5829. status open
  5830. \begin_layout Plain Layout
  5831. \align center
  5832. \begin_inset Float figure
  5833. wide false
  5834. sideways false
  5835. status open
  5836. \begin_layout Plain Layout
  5837. \align center
  5838. \begin_inset Graphics
  5839. filename graphics/CD4-csaw/ChIP-seq/H3K4me3-neighborhood-clusters-CROP.png
  5840. lyxscale 25
  5841. width 30col%
  5842. groupId covprof-subfig
  5843. \end_inset
  5844. \end_layout
  5845. \begin_layout Plain Layout
  5846. \begin_inset Caption Standard
  5847. \begin_layout Plain Layout
  5848. \series bold
  5849. \begin_inset CommandInset label
  5850. LatexCommand label
  5851. name "fig:H3K4me3-neighborhood-clusters"
  5852. \end_inset
  5853. Average relative coverage for each bin in each cluster
  5854. \end_layout
  5855. \end_inset
  5856. \end_layout
  5857. \end_inset
  5858. \begin_inset space \hfill{}
  5859. \end_inset
  5860. \begin_inset Float figure
  5861. wide false
  5862. sideways false
  5863. status open
  5864. \begin_layout Plain Layout
  5865. \align center
  5866. \begin_inset Graphics
  5867. filename graphics/CD4-csaw/ChIP-seq/H3K4me3-neighborhood-PCA-CROP.png
  5868. lyxscale 25
  5869. width 30col%
  5870. groupId covprof-subfig
  5871. \end_inset
  5872. \end_layout
  5873. \begin_layout Plain Layout
  5874. \begin_inset Caption Standard
  5875. \begin_layout Plain Layout
  5876. \series bold
  5877. \begin_inset CommandInset label
  5878. LatexCommand label
  5879. name "fig:H3K4me3-neighborhood-pca"
  5880. \end_inset
  5881. PCA of relative coverage depth, colored by K-means cluster membership.
  5882. \end_layout
  5883. \end_inset
  5884. \end_layout
  5885. \end_inset
  5886. \begin_inset space \hfill{}
  5887. \end_inset
  5888. \begin_inset Float figure
  5889. wide false
  5890. sideways false
  5891. status open
  5892. \begin_layout Plain Layout
  5893. \align center
  5894. \begin_inset Graphics
  5895. filename graphics/CD4-csaw/ChIP-seq/H3K4me3-neighborhood-expression-CROP.png
  5896. lyxscale 25
  5897. width 30col%
  5898. groupId covprof-subfig
  5899. \end_inset
  5900. \end_layout
  5901. \begin_layout Plain Layout
  5902. \begin_inset Caption Standard
  5903. \begin_layout Plain Layout
  5904. \series bold
  5905. \begin_inset CommandInset label
  5906. LatexCommand label
  5907. name "fig:H3K4me3-neighborhood-expression"
  5908. \end_inset
  5909. Gene expression grouped by promoter coverage clusters.
  5910. \end_layout
  5911. \end_inset
  5912. \end_layout
  5913. \end_inset
  5914. \end_layout
  5915. \begin_layout Plain Layout
  5916. \begin_inset Caption Standard
  5917. \begin_layout Plain Layout
  5918. \series bold
  5919. \begin_inset CommandInset label
  5920. LatexCommand label
  5921. name "fig:H3K4me3-neighborhood"
  5922. \end_inset
  5923. K-means clustering of promoter H3K4me3 relative coverage depth in naïve
  5924. day 0 samples.
  5925. \series default
  5926. H3K4me2 ChIP-seq reads were binned into 500-bp windows tiled across each
  5927. promoter from 5
  5928. \begin_inset space ~
  5929. \end_inset
  5930. kbp upstream to 5
  5931. \begin_inset space ~
  5932. \end_inset
  5933. kbp downstream, and the logCPM values were normalized within each promoter
  5934. to an average of 0, yielding relative coverage depths.
  5935. These were then grouped using K-means clustering with
  5936. \begin_inset Formula $K=6$
  5937. \end_inset
  5938. ,
  5939. \series bold
  5940. \series default
  5941. and the average bin values were plotted for each cluster (a).
  5942. The
  5943. \begin_inset Formula $x$
  5944. \end_inset
  5945. -axis is the genomic coordinate of each bin relative to the the transcription
  5946. start site, and the
  5947. \begin_inset Formula $y$
  5948. \end_inset
  5949. -axis is the mean relative coverage depth of that bin across all promoters
  5950. in the cluster.
  5951. Each line represents the average
  5952. \begin_inset Quotes eld
  5953. \end_inset
  5954. shape
  5955. \begin_inset Quotes erd
  5956. \end_inset
  5957. of the promoter coverage for promoters in that cluster.
  5958. PCA was performed on the same data, and the first two PCs were plotted,
  5959. coloring each point by its K-means cluster identity (b).
  5960. For each cluster, the distribution of gene expression values was plotted
  5961. (c).
  5962. \end_layout
  5963. \end_inset
  5964. \end_layout
  5965. \end_inset
  5966. \end_layout
  5967. \begin_layout Standard
  5968. \begin_inset ERT
  5969. status open
  5970. \begin_layout Plain Layout
  5971. \backslash
  5972. end{landscape}
  5973. \end_layout
  5974. \begin_layout Plain Layout
  5975. }
  5976. \end_layout
  5977. \end_inset
  5978. \end_layout
  5979. \begin_layout Standard
  5980. \begin_inset Flex TODO Note (inline)
  5981. status open
  5982. \begin_layout Plain Layout
  5983. Is there more to say here?
  5984. \end_layout
  5985. \end_inset
  5986. \end_layout
  5987. \begin_layout Standard
  5988. All observations described above for H3K4me2
  5989. \begin_inset Flex Glossary Term
  5990. status open
  5991. \begin_layout Plain Layout
  5992. ChIP-seq
  5993. \end_layout
  5994. \end_inset
  5995. also appear to hold for H3K4me3 as well (Figure
  5996. \begin_inset CommandInset ref
  5997. LatexCommand ref
  5998. reference "fig:H3K4me3-neighborhood"
  5999. plural "false"
  6000. caps "false"
  6001. noprefix "false"
  6002. \end_inset
  6003. ).
  6004. This is expected, since there is a high correlation between the positions
  6005. where both histone marks occur.
  6006. \end_layout
  6007. \begin_layout Subsection
  6008. Promoter coverage H3K27me3
  6009. \end_layout
  6010. \begin_layout Standard
  6011. \begin_inset ERT
  6012. status open
  6013. \begin_layout Plain Layout
  6014. \backslash
  6015. afterpage{
  6016. \end_layout
  6017. \begin_layout Plain Layout
  6018. \backslash
  6019. begin{landscape}
  6020. \end_layout
  6021. \end_inset
  6022. \end_layout
  6023. \begin_layout Standard
  6024. \begin_inset Float figure
  6025. wide false
  6026. sideways false
  6027. status collapsed
  6028. \begin_layout Plain Layout
  6029. \align center
  6030. \begin_inset Float figure
  6031. wide false
  6032. sideways false
  6033. status open
  6034. \begin_layout Plain Layout
  6035. \align center
  6036. \begin_inset Graphics
  6037. filename graphics/CD4-csaw/ChIP-seq/H3K27me3-neighborhood-clusters-CROP.png
  6038. lyxscale 25
  6039. width 30col%
  6040. groupId covprof-subfig
  6041. \end_inset
  6042. \end_layout
  6043. \begin_layout Plain Layout
  6044. \begin_inset Caption Standard
  6045. \begin_layout Plain Layout
  6046. \series bold
  6047. \begin_inset CommandInset label
  6048. LatexCommand label
  6049. name "fig:H3K27me3-neighborhood-clusters"
  6050. \end_inset
  6051. Average relative coverage for each bin in each cluster
  6052. \end_layout
  6053. \end_inset
  6054. \end_layout
  6055. \end_inset
  6056. \begin_inset space \hfill{}
  6057. \end_inset
  6058. \begin_inset Float figure
  6059. wide false
  6060. sideways false
  6061. status open
  6062. \begin_layout Plain Layout
  6063. \align center
  6064. \begin_inset Graphics
  6065. filename graphics/CD4-csaw/ChIP-seq/H3K27me3-neighborhood-PCA-CROP.png
  6066. lyxscale 25
  6067. width 30col%
  6068. groupId covprof-subfig
  6069. \end_inset
  6070. \end_layout
  6071. \begin_layout Plain Layout
  6072. \begin_inset Caption Standard
  6073. \begin_layout Plain Layout
  6074. \series bold
  6075. \begin_inset CommandInset label
  6076. LatexCommand label
  6077. name "fig:H3K27me3-neighborhood-pca"
  6078. \end_inset
  6079. PCA of relative coverage depth, colored by K-means cluster membership.
  6080. \series default
  6081. Note that Cluster 6 is hidden behind all the other clusters.
  6082. \end_layout
  6083. \end_inset
  6084. \end_layout
  6085. \end_inset
  6086. \begin_inset space \hfill{}
  6087. \end_inset
  6088. \begin_inset Float figure
  6089. wide false
  6090. sideways false
  6091. status open
  6092. \begin_layout Plain Layout
  6093. \align center
  6094. \begin_inset Graphics
  6095. filename graphics/CD4-csaw/ChIP-seq/H3K27me3-neighborhood-expression-CROP.png
  6096. lyxscale 25
  6097. width 30col%
  6098. groupId covprof-subfig
  6099. \end_inset
  6100. \end_layout
  6101. \begin_layout Plain Layout
  6102. \begin_inset Caption Standard
  6103. \begin_layout Plain Layout
  6104. \series bold
  6105. \begin_inset CommandInset label
  6106. LatexCommand label
  6107. name "fig:H3K27me3-neighborhood-expression"
  6108. \end_inset
  6109. Gene expression grouped by promoter coverage clusters.
  6110. \end_layout
  6111. \end_inset
  6112. \end_layout
  6113. \end_inset
  6114. \end_layout
  6115. \begin_layout Plain Layout
  6116. \begin_inset Flex TODO Note (inline)
  6117. status open
  6118. \begin_layout Plain Layout
  6119. Repeated figure legends are kind of an issue here.
  6120. What to do?
  6121. \end_layout
  6122. \end_inset
  6123. \end_layout
  6124. \begin_layout Plain Layout
  6125. \begin_inset Caption Standard
  6126. \begin_layout Plain Layout
  6127. \series bold
  6128. \begin_inset CommandInset label
  6129. LatexCommand label
  6130. name "fig:H3K27me3-neighborhood"
  6131. \end_inset
  6132. K-means clustering of promoter H3K27me3 relative coverage depth in naïve
  6133. day 0 samples.
  6134. \series default
  6135. H3K27me3 ChIP-seq reads were binned into 500-bp windows tiled across each
  6136. promoter from 5
  6137. \begin_inset space ~
  6138. \end_inset
  6139. kbp upstream to 5
  6140. \begin_inset space ~
  6141. \end_inset
  6142. kbp downstream, and the logCPM values were normalized within each promoter
  6143. to an average of 0, yielding relative coverage depths.
  6144. These were then grouped using
  6145. \begin_inset Formula $k$
  6146. \end_inset
  6147. -means clustering with
  6148. \begin_inset Formula $K=6$
  6149. \end_inset
  6150. ,
  6151. \series bold
  6152. \series default
  6153. and the average bin values were plotted for each cluster (a).
  6154. The
  6155. \begin_inset Formula $x$
  6156. \end_inset
  6157. -axis is the genomic coordinate of each bin relative to the the transcription
  6158. start site, and the
  6159. \begin_inset Formula $y$
  6160. \end_inset
  6161. -axis is the mean relative coverage depth of that bin across all promoters
  6162. in the cluster.
  6163. Each line represents the average
  6164. \begin_inset Quotes eld
  6165. \end_inset
  6166. shape
  6167. \begin_inset Quotes erd
  6168. \end_inset
  6169. of the promoter coverage for promoters in that cluster.
  6170. PCA was performed on the same data, and the first two PCs were plotted,
  6171. coloring each point by its K-means cluster identity (b).
  6172. For each cluster, the distribution of gene expression values was plotted
  6173. (c).
  6174. \end_layout
  6175. \end_inset
  6176. \end_layout
  6177. \end_inset
  6178. \end_layout
  6179. \begin_layout Standard
  6180. \begin_inset ERT
  6181. status open
  6182. \begin_layout Plain Layout
  6183. \backslash
  6184. end{landscape}
  6185. \end_layout
  6186. \begin_layout Plain Layout
  6187. }
  6188. \end_layout
  6189. \end_inset
  6190. \end_layout
  6191. \begin_layout Standard
  6192. \begin_inset Flex TODO Note (inline)
  6193. status open
  6194. \begin_layout Plain Layout
  6195. Should maybe re-explain what was done or refer back to the previous section.
  6196. \end_layout
  6197. \end_inset
  6198. \end_layout
  6199. \begin_layout Standard
  6200. Unlike both H3K4 marks, whose main patterns of variation appear directly
  6201. related to the size and position of a single peak within the promoter,
  6202. the patterns of H3K27me3 methylation in promoters are more complex (Figure
  6203. \begin_inset CommandInset ref
  6204. LatexCommand ref
  6205. reference "fig:H3K27me3-neighborhood"
  6206. plural "false"
  6207. caps "false"
  6208. noprefix "false"
  6209. \end_inset
  6210. ).
  6211. Once again looking at the relative coverage in a 500-bp wide bins in a
  6212. 5kb radius around each
  6213. \begin_inset Flex Glossary Term
  6214. status open
  6215. \begin_layout Plain Layout
  6216. TSS
  6217. \end_layout
  6218. \end_inset
  6219. , promoters were clustered based on the normalized relative coverage values
  6220. in each bin using
  6221. \begin_inset Formula $k$
  6222. \end_inset
  6223. -means clustering with
  6224. \begin_inset Formula $K=6$
  6225. \end_inset
  6226. (Figure
  6227. \begin_inset CommandInset ref
  6228. LatexCommand ref
  6229. reference "fig:H3K27me3-neighborhood-clusters"
  6230. plural "false"
  6231. caps "false"
  6232. noprefix "false"
  6233. \end_inset
  6234. ).
  6235. This time, 3
  6236. \begin_inset Quotes eld
  6237. \end_inset
  6238. axes
  6239. \begin_inset Quotes erd
  6240. \end_inset
  6241. of variation can be observed, each represented by 2 clusters with opposing
  6242. patterns.
  6243. The first axis is greater upstream coverage (Cluster 1) vs.
  6244. greater downstream coverage (Cluster 3); the second axis is the coverage
  6245. at the
  6246. \begin_inset Flex Glossary Term
  6247. status open
  6248. \begin_layout Plain Layout
  6249. TSS
  6250. \end_layout
  6251. \end_inset
  6252. itself: peak (Cluster 4) or trough (Cluster 2); lastly, the third axis
  6253. represents a trough upstream of the
  6254. \begin_inset Flex Glossary Term
  6255. status open
  6256. \begin_layout Plain Layout
  6257. TSS
  6258. \end_layout
  6259. \end_inset
  6260. (Cluster 5) vs.
  6261. downstream of the
  6262. \begin_inset Flex Glossary Term
  6263. status open
  6264. \begin_layout Plain Layout
  6265. TSS
  6266. \end_layout
  6267. \end_inset
  6268. (Cluster 6).
  6269. Referring to these opposing pairs of clusters as axes of variation is justified
  6270. , because they correspond precisely to the first 3
  6271. \begin_inset ERT
  6272. status collapsed
  6273. \begin_layout Plain Layout
  6274. \backslash
  6275. glspl*{PC}
  6276. \end_layout
  6277. \end_inset
  6278. in the
  6279. \begin_inset Flex Glossary Term
  6280. status open
  6281. \begin_layout Plain Layout
  6282. PCA
  6283. \end_layout
  6284. \end_inset
  6285. plot of the relative coverage values (Figure
  6286. \begin_inset CommandInset ref
  6287. LatexCommand ref
  6288. reference "fig:H3K27me3-neighborhood-pca"
  6289. plural "false"
  6290. caps "false"
  6291. noprefix "false"
  6292. \end_inset
  6293. ).
  6294. The
  6295. \begin_inset Flex Glossary Term
  6296. status open
  6297. \begin_layout Plain Layout
  6298. PCA
  6299. \end_layout
  6300. \end_inset
  6301. plot reveals that as in the case of H3K4me2, all the
  6302. \begin_inset Quotes eld
  6303. \end_inset
  6304. clusters
  6305. \begin_inset Quotes erd
  6306. \end_inset
  6307. are really just sections of a single connected cloud rather than discrete
  6308. clusters.
  6309. The cloud is approximately ellipsoid-shaped, with each PC being an axis
  6310. of the ellipse, and each cluster consisting of a pyramidal section of the
  6311. ellipsoid.
  6312. \end_layout
  6313. \begin_layout Standard
  6314. In Figure
  6315. \begin_inset CommandInset ref
  6316. LatexCommand ref
  6317. reference "fig:H3K27me3-neighborhood-expression"
  6318. plural "false"
  6319. caps "false"
  6320. noprefix "false"
  6321. \end_inset
  6322. , we can see that Clusters 1 and 2 are the only clusters with higher gene
  6323. expression than the others.
  6324. For Cluster 2, this is expected, since this cluster represents genes with
  6325. depletion of H3K27me3 near the promoter.
  6326. Hence, elevated expression in cluster 2 is consistent with the conventional
  6327. view of H3K27me3 as a deactivating mark.
  6328. However, Cluster 1, the cluster with the most elevated gene expression,
  6329. represents genes with elevated coverage upstream of the
  6330. \begin_inset Flex Glossary Term
  6331. status open
  6332. \begin_layout Plain Layout
  6333. TSS
  6334. \end_layout
  6335. \end_inset
  6336. , or equivalently, decreased coverage downstream, inside the gene body.
  6337. The opposite pattern, in which H3K27me3 is more abundant within the gene
  6338. body and less abundance in the upstream promoter region, does not show
  6339. any elevation in gene expression.
  6340. As with H3K4me2, this shows that the location of H3K27 trimethylation relative
  6341. to the
  6342. \begin_inset Flex Glossary Term
  6343. status open
  6344. \begin_layout Plain Layout
  6345. TSS
  6346. \end_layout
  6347. \end_inset
  6348. is potentially an important factor beyond simple proximity.
  6349. \end_layout
  6350. \begin_layout Standard
  6351. \begin_inset Flex TODO Note (inline)
  6352. status open
  6353. \begin_layout Plain Layout
  6354. Show the figures where the negative result ended this line of inquiry.
  6355. I need to debug some errors resulting from an R upgrade to do this.
  6356. \end_layout
  6357. \end_inset
  6358. \end_layout
  6359. \begin_layout Subsection
  6360. Defined pattern analysis
  6361. \end_layout
  6362. \begin_layout Standard
  6363. \begin_inset Flex TODO Note (inline)
  6364. status open
  6365. \begin_layout Plain Layout
  6366. This was where I defined interesting expression patterns and then looked
  6367. at initial relative promoter coverage for each expression pattern.
  6368. Negative result.
  6369. I forgot about this until recently.
  6370. Worth including? Remember to also write methods.
  6371. \end_layout
  6372. \end_inset
  6373. \end_layout
  6374. \begin_layout Subsection
  6375. Promoter CpG islands?
  6376. \end_layout
  6377. \begin_layout Standard
  6378. \begin_inset Flex TODO Note (inline)
  6379. status collapsed
  6380. \begin_layout Plain Layout
  6381. I forgot until recently about the work I did on this.
  6382. Worth including? Remember to also write methods.
  6383. \end_layout
  6384. \end_inset
  6385. \end_layout
  6386. \begin_layout Section
  6387. Discussion
  6388. \end_layout
  6389. \begin_layout Standard
  6390. \begin_inset Flex TODO Note (inline)
  6391. status open
  6392. \begin_layout Plain Layout
  6393. Write better section headers
  6394. \end_layout
  6395. \end_inset
  6396. \end_layout
  6397. \begin_layout Subsection
  6398. Effective promoter radius
  6399. \end_layout
  6400. \begin_layout Standard
  6401. Figure
  6402. \begin_inset CommandInset ref
  6403. LatexCommand ref
  6404. reference "fig:near-promoter-peak-enrich"
  6405. plural "false"
  6406. caps "false"
  6407. noprefix "false"
  6408. \end_inset
  6409. shows that H3K4me2, H3K4me3, and H3K27me3 are all enriched near promoters,
  6410. relative to the rest of the genome, consistent with their conventionally
  6411. understood role in regulating gene transcription.
  6412. Interestingly, the radius within this enrichment occurs is not the same
  6413. for each histone mark.
  6414. H3K4me2 and H3K4me3 are enriched within a 1
  6415. \begin_inset space \thinspace{}
  6416. \end_inset
  6417. kb radius, while H3K27me3 is enriched within 2.5
  6418. \begin_inset space \thinspace{}
  6419. \end_inset
  6420. kb.
  6421. Notably, the determined promoter radius was consistent across all experimental
  6422. conditions, varying only between different histone marks.
  6423. This suggests that the conventional
  6424. \begin_inset Quotes eld
  6425. \end_inset
  6426. one size fits all
  6427. \begin_inset Quotes erd
  6428. \end_inset
  6429. approach of defining a single promoter region for each gene (or each
  6430. \begin_inset Flex Glossary Term
  6431. status open
  6432. \begin_layout Plain Layout
  6433. TSS
  6434. \end_layout
  6435. \end_inset
  6436. ) and using that same promoter region for analyzing all types of genomic
  6437. data within an experiment may not be appropriate, and a better approach
  6438. may be to use a separate promoter radius for each kind of data, with each
  6439. radius being derived from the data itself.
  6440. Furthermore, the apparent asymmetry of upstream and downstream promoter
  6441. histone modification with respect to gene expression, seen in Figures
  6442. \begin_inset CommandInset ref
  6443. LatexCommand ref
  6444. reference "fig:H3K4me2-neighborhood"
  6445. plural "false"
  6446. caps "false"
  6447. noprefix "false"
  6448. \end_inset
  6449. ,
  6450. \begin_inset CommandInset ref
  6451. LatexCommand ref
  6452. reference "fig:H3K4me3-neighborhood"
  6453. plural "false"
  6454. caps "false"
  6455. noprefix "false"
  6456. \end_inset
  6457. , and
  6458. \begin_inset CommandInset ref
  6459. LatexCommand ref
  6460. reference "fig:H3K27me3-neighborhood"
  6461. plural "false"
  6462. caps "false"
  6463. noprefix "false"
  6464. \end_inset
  6465. , shows that even the concept of a promoter
  6466. \begin_inset Quotes eld
  6467. \end_inset
  6468. radius
  6469. \begin_inset Quotes erd
  6470. \end_inset
  6471. is likely an oversimplification.
  6472. At a minimum, nearby enrichment of peaks should be evaluated separately
  6473. for both upstream and downstream peaks, and an appropriate
  6474. \begin_inset Quotes eld
  6475. \end_inset
  6476. radius
  6477. \begin_inset Quotes erd
  6478. \end_inset
  6479. should be selected for each direction.
  6480. \end_layout
  6481. \begin_layout Standard
  6482. Figures
  6483. \begin_inset CommandInset ref
  6484. LatexCommand ref
  6485. reference "fig:H3K4me2-neighborhood"
  6486. plural "false"
  6487. caps "false"
  6488. noprefix "false"
  6489. \end_inset
  6490. and
  6491. \begin_inset CommandInset ref
  6492. LatexCommand ref
  6493. reference "fig:H3K4me3-neighborhood"
  6494. plural "false"
  6495. caps "false"
  6496. noprefix "false"
  6497. \end_inset
  6498. show that the determined promoter radius of 1
  6499. \begin_inset space ~
  6500. \end_inset
  6501. kb is approximately consistent with the distance from the
  6502. \begin_inset Flex Glossary Term
  6503. status open
  6504. \begin_layout Plain Layout
  6505. TSS
  6506. \end_layout
  6507. \end_inset
  6508. at which enrichment of H3K4 methylation correlates with increased expression,
  6509. showing that this radius, which was determined by a simple analysis of
  6510. measuring the distance from each
  6511. \begin_inset Flex Glossary Term
  6512. status open
  6513. \begin_layout Plain Layout
  6514. TSS
  6515. \end_layout
  6516. \end_inset
  6517. to the nearest peak, also has functional significance.
  6518. For H3K27me3, the correlation between histone modification near the promoter
  6519. and gene expression is more complex, involving non-peak variations such
  6520. as troughs in coverage at the
  6521. \begin_inset Flex Glossary Term
  6522. status open
  6523. \begin_layout Plain Layout
  6524. TSS
  6525. \end_layout
  6526. \end_inset
  6527. and asymmetric coverage upstream and downstream, so it is difficult in
  6528. this case to evaluate whether the 2.5
  6529. \begin_inset space ~
  6530. \end_inset
  6531. kb radius determined from TSS-to-peak distances is functionally significant.
  6532. However, the two patterns of coverage associated with elevated expression
  6533. levels both have interesting features within this radius.
  6534. \end_layout
  6535. \begin_layout Standard
  6536. \begin_inset Flex TODO Note (inline)
  6537. status open
  6538. \begin_layout Plain Layout
  6539. My instinct is to say
  6540. \begin_inset Quotes eld
  6541. \end_inset
  6542. further study is needed
  6543. \begin_inset Quotes erd
  6544. \end_inset
  6545. here, but that goes in Chapter 5, right?
  6546. \end_layout
  6547. \end_inset
  6548. \end_layout
  6549. \begin_layout Subsection
  6550. Convergence
  6551. \end_layout
  6552. \begin_layout Standard
  6553. \begin_inset Flex TODO Note (inline)
  6554. status open
  6555. \begin_layout Plain Layout
  6556. Look up some more references for these histone marks being involved in memory
  6557. differentiation.
  6558. (Ask Sarah)
  6559. \end_layout
  6560. \end_inset
  6561. \end_layout
  6562. \begin_layout Standard
  6563. We have observed that all 3 histone marks and the gene expression data all
  6564. exhibit evidence of convergence in abundance between naïve and memory cells
  6565. by day 14 after activation (Figure
  6566. \begin_inset CommandInset ref
  6567. LatexCommand ref
  6568. reference "fig:PCoA-promoters"
  6569. plural "false"
  6570. caps "false"
  6571. noprefix "false"
  6572. \end_inset
  6573. , Table
  6574. \begin_inset CommandInset ref
  6575. LatexCommand ref
  6576. reference "tab:Number-signif-promoters"
  6577. plural "false"
  6578. caps "false"
  6579. noprefix "false"
  6580. \end_inset
  6581. ).
  6582. The
  6583. \begin_inset Flex Glossary Term
  6584. status open
  6585. \begin_layout Plain Layout
  6586. MOFA
  6587. \end_layout
  6588. \end_inset
  6589. \begin_inset Flex Glossary Term
  6590. status open
  6591. \begin_layout Plain Layout
  6592. LF
  6593. \end_layout
  6594. \end_inset
  6595. scatter plots (Figure
  6596. \begin_inset CommandInset ref
  6597. LatexCommand ref
  6598. reference "fig:mofa-lf-scatter"
  6599. plural "false"
  6600. caps "false"
  6601. noprefix "false"
  6602. \end_inset
  6603. ) show that this pattern of convergence is captured in
  6604. \begin_inset Flex Glossary Term
  6605. status open
  6606. \begin_layout Plain Layout
  6607. LF
  6608. \end_layout
  6609. \end_inset
  6610. 5.
  6611. Like all the
  6612. \begin_inset ERT
  6613. status open
  6614. \begin_layout Plain Layout
  6615. \backslash
  6616. glspl*{LF}
  6617. \end_layout
  6618. \end_inset
  6619. in this plot, this factor explains a substantial portion of the variance
  6620. in all 4 data sets, indicating a coordinated pattern of variation shared
  6621. across all histone marks and gene expression.
  6622. This, of course, is consistent with the expectation that any naïve CD4
  6623. T-cells remaining at day 14 should have differentiated into memory cells
  6624. by that time, and should therefore have a genomic state similar to memory
  6625. cells.
  6626. This convergence is evidence that these histone marks all play an important
  6627. role in the naïve-to-memory differentiation process.
  6628. A histone mark that was not involved in naïve-to-memory differentiation
  6629. would not be expected to converge in this way after activation.
  6630. \end_layout
  6631. \begin_layout Standard
  6632. \begin_inset Float figure
  6633. wide false
  6634. sideways false
  6635. status collapsed
  6636. \begin_layout Plain Layout
  6637. \align center
  6638. \begin_inset Graphics
  6639. filename graphics/CD4-csaw/LaMere2016_fig8.pdf
  6640. lyxscale 50
  6641. width 60col%
  6642. groupId colwidth
  6643. \end_inset
  6644. \end_layout
  6645. \begin_layout Plain Layout
  6646. \begin_inset Caption Standard
  6647. \begin_layout Plain Layout
  6648. \series bold
  6649. \begin_inset CommandInset label
  6650. LatexCommand label
  6651. name "fig:Lamere2016-Fig8"
  6652. \end_inset
  6653. Lamere 2016 Figure 8
  6654. \begin_inset CommandInset citation
  6655. LatexCommand cite
  6656. key "LaMere2016"
  6657. literal "false"
  6658. \end_inset
  6659. ,
  6660. \begin_inset Quotes eld
  6661. \end_inset
  6662. Model for the role of H3K4 methylation during CD4 T-cell activation.
  6663. \begin_inset Quotes erd
  6664. \end_inset
  6665. \series default
  6666. Reproduced with permission.
  6667. \end_layout
  6668. \end_inset
  6669. \end_layout
  6670. \end_inset
  6671. \end_layout
  6672. \begin_layout Standard
  6673. In H3K4me2, H3K4me3, and
  6674. \begin_inset Flex Glossary Term
  6675. status open
  6676. \begin_layout Plain Layout
  6677. RNA-seq
  6678. \end_layout
  6679. \end_inset
  6680. , this convergence appears to be in progress already by Day 5, shown by
  6681. the smaller distance between naïve and memory cells at day 5 along the
  6682. \begin_inset Formula $y$
  6683. \end_inset
  6684. -axes in Figures
  6685. \begin_inset CommandInset ref
  6686. LatexCommand ref
  6687. reference "fig:PCoA-H3K4me2-prom"
  6688. plural "false"
  6689. caps "false"
  6690. noprefix "false"
  6691. \end_inset
  6692. ,
  6693. \begin_inset CommandInset ref
  6694. LatexCommand ref
  6695. reference "fig:PCoA-H3K4me3-prom"
  6696. plural "false"
  6697. caps "false"
  6698. noprefix "false"
  6699. \end_inset
  6700. , and
  6701. \begin_inset CommandInset ref
  6702. LatexCommand ref
  6703. reference "fig:RNA-PCA-group"
  6704. plural "false"
  6705. caps "false"
  6706. noprefix "false"
  6707. \end_inset
  6708. .
  6709. This agrees with the model proposed by Sarah Lamere based on an prior analysis
  6710. of the same data, shown in Figure
  6711. \begin_inset CommandInset ref
  6712. LatexCommand ref
  6713. reference "fig:Lamere2016-Fig8"
  6714. plural "false"
  6715. caps "false"
  6716. noprefix "false"
  6717. \end_inset
  6718. , which shows the pattern of H3K4 methylation and expression for naïve cells
  6719. and memory cells converging at day 5.
  6720. This model was developed without the benefit of the
  6721. \begin_inset Flex Glossary Term
  6722. status open
  6723. \begin_layout Plain Layout
  6724. PCoA
  6725. \end_layout
  6726. \end_inset
  6727. plots in Figure
  6728. \begin_inset CommandInset ref
  6729. LatexCommand ref
  6730. reference "fig:PCoA-promoters"
  6731. plural "false"
  6732. caps "false"
  6733. noprefix "false"
  6734. \end_inset
  6735. , which have been corrected for confounding factors by ComBat and
  6736. \begin_inset Flex Glossary Term
  6737. status open
  6738. \begin_layout Plain Layout
  6739. SVA
  6740. \end_layout
  6741. \end_inset
  6742. .
  6743. This shows that proper batch correction assists in extracting meaningful
  6744. patterns in the data while eliminating systematic sources of irrelevant
  6745. variation in the data, allowing simple automated procedures like
  6746. \begin_inset Flex Glossary Term
  6747. status open
  6748. \begin_layout Plain Layout
  6749. PCoA
  6750. \end_layout
  6751. \end_inset
  6752. to reveal interesting behaviors in the data that were previously only detectabl
  6753. e by a detailed manual analysis.
  6754. \end_layout
  6755. \begin_layout Standard
  6756. While the ideal comparison to demonstrate this convergence would be naïve
  6757. cells at day 14 to memory cells at day 0, this is not feasible in this
  6758. experimental system, since neither naïve nor memory cells are able to fully
  6759. return to their pre-activation state, as shown by the lack of overlap between
  6760. days 0 and 14 for either naïve or memory cells in Figure
  6761. \begin_inset CommandInset ref
  6762. LatexCommand ref
  6763. reference "fig:PCoA-promoters"
  6764. plural "false"
  6765. caps "false"
  6766. noprefix "false"
  6767. \end_inset
  6768. .
  6769. \end_layout
  6770. \begin_layout Subsection
  6771. Positional
  6772. \end_layout
  6773. \begin_layout Standard
  6774. When looking at patterns in the relative coverage of each histone mark near
  6775. the
  6776. \begin_inset Flex Glossary Term
  6777. status open
  6778. \begin_layout Plain Layout
  6779. TSS
  6780. \end_layout
  6781. \end_inset
  6782. of each gene, several interesting patterns were apparent.
  6783. For H3K4me2 and H3K4me3, the pattern was straightforward: the consistent
  6784. pattern across all promoters was a single peak a few kb wide, with the
  6785. main axis of variation being the position of this peak relative to the
  6786. \begin_inset Flex Glossary Term
  6787. status open
  6788. \begin_layout Plain Layout
  6789. TSS
  6790. \end_layout
  6791. \end_inset
  6792. (Figures
  6793. \begin_inset CommandInset ref
  6794. LatexCommand ref
  6795. reference "fig:H3K4me2-neighborhood"
  6796. plural "false"
  6797. caps "false"
  6798. noprefix "false"
  6799. \end_inset
  6800. &
  6801. \begin_inset CommandInset ref
  6802. LatexCommand ref
  6803. reference "fig:H3K4me3-neighborhood"
  6804. plural "false"
  6805. caps "false"
  6806. noprefix "false"
  6807. \end_inset
  6808. ).
  6809. There were no obvious
  6810. \begin_inset Quotes eld
  6811. \end_inset
  6812. preferred
  6813. \begin_inset Quotes erd
  6814. \end_inset
  6815. positions, but rather a continuous distribution of relative positions ranging
  6816. all across the promoter region.
  6817. The association with gene expression was also straightforward: peaks closer
  6818. to the
  6819. \begin_inset Flex Glossary Term
  6820. status open
  6821. \begin_layout Plain Layout
  6822. TSS
  6823. \end_layout
  6824. \end_inset
  6825. were more strongly associated with elevated gene expression.
  6826. Coverage downstream of the
  6827. \begin_inset Flex Glossary Term
  6828. status open
  6829. \begin_layout Plain Layout
  6830. TSS
  6831. \end_layout
  6832. \end_inset
  6833. appears to be more strongly associated with elevated expression than coverage
  6834. the same distance upstream, indicating that the
  6835. \begin_inset Quotes eld
  6836. \end_inset
  6837. effective promoter region
  6838. \begin_inset Quotes erd
  6839. \end_inset
  6840. for H3K4me2 and H3K4me3 may be centered downstream of the
  6841. \begin_inset Flex Glossary Term
  6842. status open
  6843. \begin_layout Plain Layout
  6844. TSS
  6845. \end_layout
  6846. \end_inset
  6847. .
  6848. \end_layout
  6849. \begin_layout Standard
  6850. The relative promoter coverage for H3K27me3 had a more complex pattern,
  6851. with two specific patterns of promoter coverage associated with elevated
  6852. expression: a sharp depletion of H3K27me3 around the
  6853. \begin_inset Flex Glossary Term
  6854. status open
  6855. \begin_layout Plain Layout
  6856. TSS
  6857. \end_layout
  6858. \end_inset
  6859. relative to the surrounding area, and a depletion of H3K27me3 downstream
  6860. of the
  6861. \begin_inset Flex Glossary Term
  6862. status open
  6863. \begin_layout Plain Layout
  6864. TSS
  6865. \end_layout
  6866. \end_inset
  6867. relative to upstream (Figure
  6868. \begin_inset CommandInset ref
  6869. LatexCommand ref
  6870. reference "fig:H3K27me3-neighborhood"
  6871. plural "false"
  6872. caps "false"
  6873. noprefix "false"
  6874. \end_inset
  6875. ).
  6876. A previous study found that H3K27me3 depletion within the gene body was
  6877. associated with elevated gene expression in 4 different cell types in mice
  6878. \begin_inset CommandInset citation
  6879. LatexCommand cite
  6880. key "Young2011"
  6881. literal "false"
  6882. \end_inset
  6883. .
  6884. This is consistent with the second pattern described here.
  6885. This study also reported that a spike in coverage at the
  6886. \begin_inset Flex Glossary Term
  6887. status open
  6888. \begin_layout Plain Layout
  6889. TSS
  6890. \end_layout
  6891. \end_inset
  6892. was associated with
  6893. \emph on
  6894. lower
  6895. \emph default
  6896. expression, which is indirectly consistent with the first pattern described
  6897. here, in the sense that it associates lower H3K27me3 levels near the
  6898. \begin_inset Flex Glossary Term
  6899. status open
  6900. \begin_layout Plain Layout
  6901. TSS
  6902. \end_layout
  6903. \end_inset
  6904. with higher expression.
  6905. \end_layout
  6906. \begin_layout Subsection
  6907. Workflow
  6908. \end_layout
  6909. \begin_layout Standard
  6910. \begin_inset ERT
  6911. status open
  6912. \begin_layout Plain Layout
  6913. \backslash
  6914. afterpage{
  6915. \end_layout
  6916. \begin_layout Plain Layout
  6917. \backslash
  6918. begin{landscape}
  6919. \end_layout
  6920. \end_inset
  6921. \end_layout
  6922. \begin_layout Standard
  6923. \begin_inset Float figure
  6924. wide false
  6925. sideways false
  6926. status open
  6927. \begin_layout Plain Layout
  6928. \align center
  6929. \begin_inset Graphics
  6930. filename graphics/CD4-csaw/rulegraphs/rulegraph-all.pdf
  6931. lyxscale 50
  6932. width 100col%
  6933. height 95theight%
  6934. \end_inset
  6935. \end_layout
  6936. \begin_layout Plain Layout
  6937. \begin_inset Caption Standard
  6938. \begin_layout Plain Layout
  6939. \begin_inset CommandInset label
  6940. LatexCommand label
  6941. name "fig:rulegraph"
  6942. \end_inset
  6943. \series bold
  6944. Dependency graph of steps in reproducible workflow.
  6945. \end_layout
  6946. \end_inset
  6947. \end_layout
  6948. \end_inset
  6949. \end_layout
  6950. \begin_layout Standard
  6951. \begin_inset ERT
  6952. status open
  6953. \begin_layout Plain Layout
  6954. \backslash
  6955. end{landscape}
  6956. \end_layout
  6957. \begin_layout Plain Layout
  6958. }
  6959. \end_layout
  6960. \end_inset
  6961. \end_layout
  6962. \begin_layout Standard
  6963. The analyses described in this chapter were organized into a reproducible
  6964. workflow using the Snakemake workflow management system
  6965. \begin_inset CommandInset citation
  6966. LatexCommand cite
  6967. key "Koster2012"
  6968. literal "false"
  6969. \end_inset
  6970. .
  6971. As shown in Figure
  6972. \begin_inset CommandInset ref
  6973. LatexCommand ref
  6974. reference "fig:rulegraph"
  6975. plural "false"
  6976. caps "false"
  6977. noprefix "false"
  6978. \end_inset
  6979. , the workflow includes many steps with complex dependencies between them.
  6980. For example, the step that counts the number of
  6981. \begin_inset Flex Glossary Term
  6982. status open
  6983. \begin_layout Plain Layout
  6984. ChIP-seq
  6985. \end_layout
  6986. \end_inset
  6987. reads in 500
  6988. \begin_inset space ~
  6989. \end_inset
  6990. bp windows in each promoter (the starting point for Figures
  6991. \begin_inset CommandInset ref
  6992. LatexCommand ref
  6993. reference "fig:H3K4me2-neighborhood"
  6994. plural "false"
  6995. caps "false"
  6996. noprefix "false"
  6997. \end_inset
  6998. ,
  6999. \begin_inset CommandInset ref
  7000. LatexCommand ref
  7001. reference "fig:H3K4me3-neighborhood"
  7002. plural "false"
  7003. caps "false"
  7004. noprefix "false"
  7005. \end_inset
  7006. , and
  7007. \begin_inset CommandInset ref
  7008. LatexCommand ref
  7009. reference "fig:H3K27me3-neighborhood"
  7010. plural "false"
  7011. caps "false"
  7012. noprefix "false"
  7013. \end_inset
  7014. ), named
  7015. \begin_inset Flex Code
  7016. status open
  7017. \begin_layout Plain Layout
  7018. chipseq_count_tss_neighborhoods
  7019. \end_layout
  7020. \end_inset
  7021. , depends on the
  7022. \begin_inset Flex Glossary Term
  7023. status open
  7024. \begin_layout Plain Layout
  7025. RNA-seq
  7026. \end_layout
  7027. \end_inset
  7028. abundance estimates in order to select the most-used
  7029. \begin_inset Flex Glossary Term
  7030. status open
  7031. \begin_layout Plain Layout
  7032. TSS
  7033. \end_layout
  7034. \end_inset
  7035. for each gene, the aligned
  7036. \begin_inset Flex Glossary Term
  7037. status open
  7038. \begin_layout Plain Layout
  7039. ChIP-seq
  7040. \end_layout
  7041. \end_inset
  7042. reads, the index for those reads, and the blacklist of regions to be excluded
  7043. from
  7044. \begin_inset Flex Glossary Term
  7045. status open
  7046. \begin_layout Plain Layout
  7047. ChIP-seq
  7048. \end_layout
  7049. \end_inset
  7050. analysis.
  7051. Each step declares its inputs and outputs, and Snakemake uses these to
  7052. determine the dependencies between steps.
  7053. Each step is marked as depending on all the steps whose outputs match its
  7054. inputs, generating the workflow graph in Figure
  7055. \begin_inset CommandInset ref
  7056. LatexCommand ref
  7057. reference "fig:rulegraph"
  7058. plural "false"
  7059. caps "false"
  7060. noprefix "false"
  7061. \end_inset
  7062. , which Snakemake uses to determine order in which to execute each step
  7063. so that each step is executed only after all of the steps it depends on
  7064. have completed, thereby automating the entire workflow from start to finish.
  7065. \end_layout
  7066. \begin_layout Standard
  7067. In addition to simply making it easier to organize the steps in the analysis,
  7068. structuring the analysis as a workflow allowed for some analysis strategies
  7069. that would not have been practical otherwise.
  7070. For example, 5 different
  7071. \begin_inset Flex Glossary Term
  7072. status open
  7073. \begin_layout Plain Layout
  7074. RNA-seq
  7075. \end_layout
  7076. \end_inset
  7077. quantification methods were tested against two different reference transcriptom
  7078. e annotations for a total of 10 different quantifications of the same
  7079. \begin_inset Flex Glossary Term
  7080. status open
  7081. \begin_layout Plain Layout
  7082. RNA-seq
  7083. \end_layout
  7084. \end_inset
  7085. data.
  7086. These were then compared against each other in the exploratory data analysis
  7087. step, to determine that the results were not very sensitive to either the
  7088. choice of quantification method or the choice of annotation.
  7089. This was possible with a single script for the exploratory data analysis,
  7090. because Snakemake was able to automate running this script for every combinatio
  7091. n of method and reference.
  7092. In a similar manner, two different peak calling methods were tested against
  7093. each other, and in this case it was determined that
  7094. \begin_inset Flex Glossary Term
  7095. status open
  7096. \begin_layout Plain Layout
  7097. SICER
  7098. \end_layout
  7099. \end_inset
  7100. was unambiguously superior to
  7101. \begin_inset Flex Glossary Term
  7102. status open
  7103. \begin_layout Plain Layout
  7104. MACS
  7105. \end_layout
  7106. \end_inset
  7107. for all histone marks studied.
  7108. By enabling these types of comparisons, structuring the analysis as an
  7109. automated workflow allowed important analysis decisions to be made in a
  7110. data-driven way, by running every reasonable option through the downstream
  7111. steps, seeing the consequences of choosing each option, and deciding accordingl
  7112. y.
  7113. \end_layout
  7114. \begin_layout Subsection
  7115. Data quality issues limit conclusions
  7116. \end_layout
  7117. \begin_layout Standard
  7118. \begin_inset Flex TODO Note (inline)
  7119. status open
  7120. \begin_layout Plain Layout
  7121. Is this needed?
  7122. \end_layout
  7123. \end_inset
  7124. \end_layout
  7125. \begin_layout Section
  7126. Future Directions
  7127. \end_layout
  7128. \begin_layout Standard
  7129. The analysis of
  7130. \begin_inset Flex Glossary Term
  7131. status open
  7132. \begin_layout Plain Layout
  7133. RNA-seq
  7134. \end_layout
  7135. \end_inset
  7136. and
  7137. \begin_inset Flex Glossary Term
  7138. status open
  7139. \begin_layout Plain Layout
  7140. ChIP-seq
  7141. \end_layout
  7142. \end_inset
  7143. in CD4 T-cells in Chapter 2 is in many ways a preliminary study that suggests
  7144. a multitude of new avenues of investigation.
  7145. Here we consider a selection of such avenues.
  7146. \end_layout
  7147. \begin_layout Subsection
  7148. Negative results
  7149. \end_layout
  7150. \begin_layout Standard
  7151. Two additional analyses were conducted beyond those reported in the results.
  7152. First, we searched for evidence that the presence or absence of a
  7153. \begin_inset Flex Glossary Term
  7154. status open
  7155. \begin_layout Plain Layout
  7156. CpGi
  7157. \end_layout
  7158. \end_inset
  7159. \begin_inset CommandInset nomenclature
  7160. LatexCommand nomenclature
  7161. symbol "CpGi"
  7162. description "CpG island"
  7163. literal "false"
  7164. \end_inset
  7165. in the promoter was correlated with increases or decreases in gene expression
  7166. or any histone mark in any of the tested contrasts.
  7167. Second, we searched for evidence that the relative
  7168. \begin_inset Flex Glossary Term
  7169. status open
  7170. \begin_layout Plain Layout
  7171. ChIP-seq
  7172. \end_layout
  7173. \end_inset
  7174. coverage profiles prior to activations could predict the change in expression
  7175. of a gene after activation.
  7176. Neither analysis turned up any clear positive results.
  7177. \end_layout
  7178. \begin_layout Subsection
  7179. Improve on the idea of an effective promoter radius
  7180. \end_layout
  7181. \begin_layout Standard
  7182. This study introduced the concept of an
  7183. \begin_inset Quotes eld
  7184. \end_inset
  7185. effective promoter radius
  7186. \begin_inset Quotes erd
  7187. \end_inset
  7188. specific to each histone mark based on distance from the
  7189. \begin_inset Flex Glossary Term
  7190. status open
  7191. \begin_layout Plain Layout
  7192. TSS
  7193. \end_layout
  7194. \end_inset
  7195. within which an excess of peaks was called for that mark.
  7196. This concept was then used to guide further analyses throughout the study.
  7197. However, while the effective promoter radius was useful in those analyses,
  7198. it is both limited in theory and shown in practice to be a possible oversimplif
  7199. ication.
  7200. First, the effective promoter radii used in this study were chosen based
  7201. on manual inspection of the TSS-to-peak distance distributions in Figure
  7202. \begin_inset CommandInset ref
  7203. LatexCommand ref
  7204. reference "fig:near-promoter-peak-enrich"
  7205. plural "false"
  7206. caps "false"
  7207. noprefix "false"
  7208. \end_inset
  7209. , selecting round numbers of analyst convenience (Table
  7210. \begin_inset CommandInset ref
  7211. LatexCommand ref
  7212. reference "tab:effective-promoter-radius"
  7213. plural "false"
  7214. caps "false"
  7215. noprefix "false"
  7216. \end_inset
  7217. ).
  7218. It would be better to define an algorithm that selects a more precise radius
  7219. based on the features of the graph.
  7220. One possible way to do this would be to randomly rearrange the called peaks
  7221. throughout the genome many (while preserving the distribution of peak widths)
  7222. and re-generate the same plot as in Figure
  7223. \begin_inset CommandInset ref
  7224. LatexCommand ref
  7225. reference "fig:near-promoter-peak-enrich"
  7226. plural "false"
  7227. caps "false"
  7228. noprefix "false"
  7229. \end_inset
  7230. .
  7231. This would yield a better
  7232. \begin_inset Quotes eld
  7233. \end_inset
  7234. background
  7235. \begin_inset Quotes erd
  7236. \end_inset
  7237. distribution that demonstrates the degree of near-TSS enrichment that would
  7238. be expected by random chance.
  7239. The effective promoter radius could be defined as the point where the true
  7240. distribution diverges from the randomized background distribution.
  7241. \end_layout
  7242. \begin_layout Standard
  7243. Furthermore, the above definition of effective promoter radius has the significa
  7244. nt limitation of being based on the peak calling method.
  7245. It is thus very sensitive to the choice of peak caller and significance
  7246. threshold for calling peaks, as well as the degree of saturation in the
  7247. sequencing.
  7248. Calling peaks from
  7249. \begin_inset Flex Glossary Term
  7250. status open
  7251. \begin_layout Plain Layout
  7252. ChIP-seq
  7253. \end_layout
  7254. \end_inset
  7255. samples with insufficient coverage depth, with the wrong peak caller, or
  7256. with a different significance threshold could give a drastically different
  7257. number of called peaks, and hence a drastically different distribution
  7258. of peak-to-TSS distances.
  7259. To address this, it is desirable to develop a better method of determining
  7260. the effective promoter radius that relies only on the distribution of read
  7261. coverage around the
  7262. \begin_inset Flex Glossary Term
  7263. status open
  7264. \begin_layout Plain Layout
  7265. TSS
  7266. \end_layout
  7267. \end_inset
  7268. , independent of the peak calling.
  7269. Furthermore, as demonstrated by the upstream-downstream asymmetries observed
  7270. in Figures
  7271. \begin_inset CommandInset ref
  7272. LatexCommand ref
  7273. reference "fig:H3K4me2-neighborhood"
  7274. plural "false"
  7275. caps "false"
  7276. noprefix "false"
  7277. \end_inset
  7278. ,
  7279. \begin_inset CommandInset ref
  7280. LatexCommand ref
  7281. reference "fig:H3K4me3-neighborhood"
  7282. plural "false"
  7283. caps "false"
  7284. noprefix "false"
  7285. \end_inset
  7286. , and
  7287. \begin_inset CommandInset ref
  7288. LatexCommand ref
  7289. reference "fig:H3K27me3-neighborhood"
  7290. plural "false"
  7291. caps "false"
  7292. noprefix "false"
  7293. \end_inset
  7294. , this definition should determine a different radius for the upstream and
  7295. downstream directions.
  7296. At this point, it may be better to rename this concept
  7297. \begin_inset Quotes eld
  7298. \end_inset
  7299. effective promoter extent
  7300. \begin_inset Quotes erd
  7301. \end_inset
  7302. and avoid the word
  7303. \begin_inset Quotes eld
  7304. \end_inset
  7305. radius
  7306. \begin_inset Quotes erd
  7307. \end_inset
  7308. , since a radius implies a symmetry about the
  7309. \begin_inset Flex Glossary Term
  7310. status open
  7311. \begin_layout Plain Layout
  7312. TSS
  7313. \end_layout
  7314. \end_inset
  7315. that is not supported by the data.
  7316. \end_layout
  7317. \begin_layout Standard
  7318. Beyond improving the definition of effective promoter extent, functional
  7319. validation is necessary to show that this measure of near-TSS enrichment
  7320. has biological meaning.
  7321. Figures
  7322. \begin_inset CommandInset ref
  7323. LatexCommand ref
  7324. reference "fig:H3K4me2-neighborhood"
  7325. plural "false"
  7326. caps "false"
  7327. noprefix "false"
  7328. \end_inset
  7329. and
  7330. \begin_inset CommandInset ref
  7331. LatexCommand ref
  7332. reference "fig:H3K4me3-neighborhood"
  7333. plural "false"
  7334. caps "false"
  7335. noprefix "false"
  7336. \end_inset
  7337. already provide a very limited functional validation of the chosen promoter
  7338. extents for H3K4me2 and H3K4me3 by showing that spikes in coverage within
  7339. this region are most strongly correlated with elevated gene expression.
  7340. However, there are other ways to show functional relevance of the promoter
  7341. extent.
  7342. For example, correlations could be computed between read counts in peaks
  7343. nearby gene promoters and the expression level of those genes, and these
  7344. correlations could be plotted against the distance of the peak upstream
  7345. or downstream of the gene's
  7346. \begin_inset Flex Glossary Term
  7347. status open
  7348. \begin_layout Plain Layout
  7349. TSS
  7350. \end_layout
  7351. \end_inset
  7352. .
  7353. If the promoter extent truly defines a
  7354. \begin_inset Quotes eld
  7355. \end_inset
  7356. sphere of influence
  7357. \begin_inset Quotes erd
  7358. \end_inset
  7359. within which a histone mark is involved with the regulation of a gene,
  7360. then the correlations for peaks within this extent should be significantly
  7361. higher than those further upstream or downstream.
  7362. Peaks within these extents may also be more likely to show differential
  7363. modification than those outside genic regions of the genome.
  7364. \end_layout
  7365. \begin_layout Subsection
  7366. Design experiments to focus on post-activation convergence of naïve & memory
  7367. cells
  7368. \end_layout
  7369. \begin_layout Standard
  7370. In this study, a convergence between naïve and memory cells was observed
  7371. in both the pattern of gene expression and in epigenetic state of the 3
  7372. histone marks studied, consistent with the hypothesis that any naïve cells
  7373. remaining 14 days after activation have differentiated into memory cells,
  7374. and that both gene expression and these histone marks are involved in this
  7375. differentiation.
  7376. However, the current study was not designed with this specific hypothesis
  7377. in mind, and it therefore has some deficiencies with regard to testing
  7378. it.
  7379. The memory CD4 samples at day 14 do not resemble the memory samples at
  7380. day 0, indicating that in the specific model of activation used for this
  7381. experiment, the cells are not guaranteed to return to their original pre-activa
  7382. tion state, or perhaps this process takes substantially longer than 14 days.
  7383. This is a challenge for the convergence hypothesis because the ideal comparison
  7384. to prove that naïve cells are converging to a resting memory state would
  7385. be to compare the final naïve time point to the Day 0 memory samples, but
  7386. this comparison is only meaningful if memory cells generally return to
  7387. the same
  7388. \begin_inset Quotes eld
  7389. \end_inset
  7390. resting
  7391. \begin_inset Quotes erd
  7392. \end_inset
  7393. state that they started at.
  7394. \end_layout
  7395. \begin_layout Standard
  7396. To better study the convergence hypothesis, a new experiment should be designed
  7397. using a model system for T-cell activation that is known to allow cells
  7398. to return as closely as possible to their pre-activation state.
  7399. Alternatively, if it is not possible to find or design such a model system,
  7400. the same cell cultures could be activated serially multiple times, and
  7401. sequenced after each activation cycle right before the next activation.
  7402. It is likely that several activations in the same model system will settle
  7403. into a cyclical pattern, converging to a consistent
  7404. \begin_inset Quotes eld
  7405. \end_inset
  7406. resting
  7407. \begin_inset Quotes erd
  7408. \end_inset
  7409. state after each activation, even if this state is different from the initial
  7410. resting state at Day 0.
  7411. If so, it will be possible to compare the final states of both naïve and
  7412. memory cells to show that they converge despite different initial conditions.
  7413. \end_layout
  7414. \begin_layout Standard
  7415. In addition, if naïve-to-memory convergence is a general pattern, it should
  7416. also be detectable in other epigenetic marks, including other histone marks
  7417. and DNA methylation.
  7418. An experiment should be designed studying a large number of epigenetic
  7419. marks known or suspected to be involved in regulation of gene expression,
  7420. assaying all of these at the same pre- and post-activation time points.
  7421. Multi-dataset factor analysis methods like
  7422. \begin_inset Flex Glossary Term
  7423. status open
  7424. \begin_layout Plain Layout
  7425. MOFA
  7426. \end_layout
  7427. \end_inset
  7428. can then be used to identify coordinated patterns of regulation shared
  7429. across many epigenetic marks.
  7430. If possible, some
  7431. \begin_inset Quotes eld
  7432. \end_inset
  7433. negative control
  7434. \begin_inset Quotes erd
  7435. \end_inset
  7436. marks should be included that are known
  7437. \emph on
  7438. not
  7439. \emph default
  7440. to be involved in T-cell activation or memory formation.
  7441. Of course, CD4 T-cells are not the only adaptive immune cells with memory.
  7442. A similar study could be designed for CD8 T-cells, B-cells, and even specific
  7443. subsets of CD4 T-cells.
  7444. \end_layout
  7445. \begin_layout Subsection
  7446. Follow up on hints of interesting patterns in promoter relative coverage
  7447. profiles
  7448. \end_layout
  7449. \begin_layout Standard
  7450. \begin_inset Flex TODO Note (inline)
  7451. status open
  7452. \begin_layout Plain Layout
  7453. I think I might need to write up the negative results for the Promoter CpG
  7454. and defined pattern analysis before writing this section.
  7455. \end_layout
  7456. \end_inset
  7457. \end_layout
  7458. \begin_layout Itemize
  7459. Also find better normalizations: maybe borrow from MACS/SICER background
  7460. correction methods?
  7461. \end_layout
  7462. \begin_layout Itemize
  7463. For H3K4, define polar coordinates based on PC1 & 2: R = peak size, Theta
  7464. = peak position.
  7465. Then correlate with expression.
  7466. \end_layout
  7467. \begin_layout Itemize
  7468. Current analysis only at Day 0.
  7469. Need to study across time points.
  7470. \end_layout
  7471. \begin_layout Itemize
  7472. Integrating data across so many dimensions is a significant analysis challenge
  7473. \end_layout
  7474. \begin_layout Subsection
  7475. Investigate causes of high correlation between mutually exclusive histone
  7476. marks
  7477. \end_layout
  7478. \begin_layout Standard
  7479. The high correlation between coverage depth observed between H3K4me2 and
  7480. H3K4me3 is both expected and unexpected.
  7481. Since both marks are associated with elevated gene transcription, a positive
  7482. correlation between them is not surprising.
  7483. However, these two marks represent different post-translational modifications
  7484. of the
  7485. \emph on
  7486. same
  7487. \emph default
  7488. lysine residue on the histone H3 polypeptide, which means that they cannot
  7489. both be present on the same H3 subunit.
  7490. Thus, the high correlation between them has several potential explanations.
  7491. One possible reason is cell population heterogeneity: perhaps some genomic
  7492. loci are frequently marked with H3K4me2 in some cells, while in other cells
  7493. the same loci are marked with H3K4me3.
  7494. Another possibility is allele-specific modifications: the loci are marked
  7495. in each diploid cell with H3K4me2 on one allele and H3K4me3 on the other
  7496. allele.
  7497. Lastly, since each histone octamer contains 2 H3 subunits, it is possible
  7498. that having one H3K4me2 mark and one H3K4me3 mark on a given histone octamer
  7499. represents a distinct epigenetic state with a different function than either
  7500. double H3K4me2 or double H3K4me3.
  7501. \end_layout
  7502. \begin_layout Standard
  7503. These three hypotheses could be disentangled by single-cell
  7504. \begin_inset Flex Glossary Term
  7505. status open
  7506. \begin_layout Plain Layout
  7507. ChIP-seq
  7508. \end_layout
  7509. \end_inset
  7510. .
  7511. If the correlation between these two histone marks persists even within
  7512. the reads for each individual cell, then cell population heterogeneity
  7513. cannot explain the correlation.
  7514. Allele-specific modification can be tested for by looking at the correlation
  7515. between read coverage of the two histone marks at heterozygous loci.
  7516. If the correlation between read counts for opposite loci is low, then this
  7517. is consistent with allele-specific modification.
  7518. Finally if the modifications do not separate by either cell or allele,
  7519. the colocation of these two marks is most likely occurring at the level
  7520. of individual histones, with the heterogeneously modified histone representing
  7521. a distinct state.
  7522. \end_layout
  7523. \begin_layout Standard
  7524. However, another experiment would be required to show direct evidence of
  7525. such a heterogeneously modified state.
  7526. Specifically a
  7527. \begin_inset Quotes eld
  7528. \end_inset
  7529. double ChIP
  7530. \begin_inset Quotes erd
  7531. \end_inset
  7532. experiment would need to be performed, where the input DNA is first subjected
  7533. to an immunoprecipitation pulldown from the anti-H3K4me2 antibody, and
  7534. then the enriched material is collected, with proteins still bound, and
  7535. immunoprecipitated
  7536. \emph on
  7537. again
  7538. \emph default
  7539. using the anti-H3K4me3 antibody.
  7540. If this yields significant numbers of non-artifactual reads in the same
  7541. regions as the individual pulldowns of the two marks, this is strong evidence
  7542. that the two marks are occurring on opposite H3 subunits of the same histones.
  7543. \end_layout
  7544. \begin_layout Standard
  7545. \begin_inset Flex TODO Note (inline)
  7546. status open
  7547. \begin_layout Plain Layout
  7548. Try to see if double ChIP-seq is actually feasible, and if not, come up
  7549. with some other idea for directly detecting the mixed mod state.
  7550. Oh! Actually ChIP-seq isn't required, only double ChIP followed by quantificati
  7551. on.
  7552. That's one possible angle.
  7553. \end_layout
  7554. \end_inset
  7555. \end_layout
  7556. \begin_layout Chapter
  7557. Improving array-based diagnostics for transplant rejection by optimizing
  7558. data preprocessing
  7559. \end_layout
  7560. \begin_layout Standard
  7561. \begin_inset Note Note
  7562. status open
  7563. \begin_layout Plain Layout
  7564. Chapter author list: Me, Sunil, Tom, Padma, Dan
  7565. \end_layout
  7566. \end_inset
  7567. \end_layout
  7568. \begin_layout Standard
  7569. \begin_inset ERT
  7570. status collapsed
  7571. \begin_layout Plain Layout
  7572. \backslash
  7573. glsresetall
  7574. \end_layout
  7575. \end_inset
  7576. \end_layout
  7577. \begin_layout Section
  7578. Approach
  7579. \end_layout
  7580. \begin_layout Subsection
  7581. Proper pre-processing is essential for array data
  7582. \end_layout
  7583. \begin_layout Standard
  7584. \begin_inset Flex TODO Note (inline)
  7585. status open
  7586. \begin_layout Plain Layout
  7587. This section could probably use some citations
  7588. \end_layout
  7589. \end_inset
  7590. \end_layout
  7591. \begin_layout Standard
  7592. Microarrays, bead arrays, and similar assays produce raw data in the form
  7593. of fluorescence intensity measurements, with the each intensity measurement
  7594. proportional to the abundance of some fluorescently labelled target DNA
  7595. or RNA sequence that base pairs to a specific probe sequence.
  7596. However, these measurements for each probe are also affected my many technical
  7597. confounding factors, such as the concentration of target material, strength
  7598. of off-target binding, and the sensitivity of the imaging sensor.
  7599. Some array designs also use multiple probe sequences for each target.
  7600. Hence, extensive pre-processing of array data is necessary to normalize
  7601. out the effects of these technical factors and summarize the information
  7602. from multiple probes to arrive at a single usable estimate of abundance
  7603. or other relevant quantity, such as a ratio of two abundances, for each
  7604. target.
  7605. \end_layout
  7606. \begin_layout Standard
  7607. The choice of pre-processing algorithms used in the analysis of an array
  7608. data set can have a large effect on the results of that analysis.
  7609. However, despite their importance, these steps are often neglected or rushed
  7610. in order to get to the more scientifically interesting analysis steps involving
  7611. the actual biology of the system under study.
  7612. Hence, it is often possible to achieve substantial gains in statistical
  7613. power, model goodness-of-fit, or other relevant performance measures, by
  7614. checking the assumptions made by each preprocessing step and choosing specific
  7615. normalization methods tailored to the specific goals of the current analysis.
  7616. \end_layout
  7617. \begin_layout Subsection
  7618. Clinical diagnostic applications for microarrays require single-channel
  7619. normalization
  7620. \end_layout
  7621. \begin_layout Standard
  7622. As the cost of performing microarray assays falls, there is increasing interest
  7623. in using genomic assays for diagnostic purposes, such as distinguishing
  7624. \begin_inset ERT
  7625. status open
  7626. \begin_layout Plain Layout
  7627. \backslash
  7628. glsdisp*{TX}{healthy transplants (TX)}
  7629. \end_layout
  7630. \end_inset
  7631. \begin_inset CommandInset nomenclature
  7632. LatexCommand nomenclature
  7633. symbol "TX"
  7634. description "healthy transplant"
  7635. literal "false"
  7636. \end_inset
  7637. from transplants undergoing
  7638. \begin_inset Flex Glossary Term
  7639. status open
  7640. \begin_layout Plain Layout
  7641. AR
  7642. \end_layout
  7643. \end_inset
  7644. \begin_inset CommandInset nomenclature
  7645. LatexCommand nomenclature
  7646. symbol "AR"
  7647. description "acute rejection"
  7648. literal "false"
  7649. \end_inset
  7650. or
  7651. \begin_inset Flex Glossary Term
  7652. status open
  7653. \begin_layout Plain Layout
  7654. ADNR
  7655. \end_layout
  7656. \end_inset
  7657. \begin_inset CommandInset nomenclature
  7658. LatexCommand nomenclature
  7659. symbol "ADNR"
  7660. description "acute dysfunction with no rejection"
  7661. literal "false"
  7662. \end_inset
  7663. .
  7664. However, the the standard normalization algorithm used for microarray data,
  7665. \begin_inset Flex Glossary Term
  7666. status open
  7667. \begin_layout Plain Layout
  7668. RMA
  7669. \end_layout
  7670. \end_inset
  7671. \begin_inset CommandInset citation
  7672. LatexCommand cite
  7673. key "Irizarry2003a"
  7674. literal "false"
  7675. \end_inset
  7676. , is not applicable in a clinical setting.
  7677. Two of the steps in
  7678. \begin_inset Flex Glossary Term
  7679. status open
  7680. \begin_layout Plain Layout
  7681. RMA
  7682. \end_layout
  7683. \end_inset
  7684. , quantile normalization and probe summarization by median polish, depend
  7685. on every array in the data set being normalized.
  7686. This means that adding or removing any arrays from a data set changes the
  7687. normalized values for all arrays, and data sets that have been normalized
  7688. separately cannot be compared to each other.
  7689. Hence, when using
  7690. \begin_inset Flex Glossary Term
  7691. status open
  7692. \begin_layout Plain Layout
  7693. RMA
  7694. \end_layout
  7695. \end_inset
  7696. , any arrays to be analyzed together must also be normalized together, and
  7697. the set of arrays included in the data set must be held constant throughout
  7698. an analysis.
  7699. \end_layout
  7700. \begin_layout Standard
  7701. These limitations present serious impediments to the use of arrays as a
  7702. diagnostic tool.
  7703. When training a classifier, the samples to be classified must not be involved
  7704. in any step of the training process, lest their inclusion bias the training
  7705. process.
  7706. Once a classifier is deployed in a clinical setting, the samples to be
  7707. classified will not even
  7708. \emph on
  7709. exist
  7710. \emph default
  7711. at the time of training, so including them would be impossible even if
  7712. it were statistically justifiable.
  7713. Therefore, any machine learning application for microarrays demands that
  7714. the normalized expression values computed for an array must depend only
  7715. on information contained within that array.
  7716. This would ensure that each array's normalization is independent of every
  7717. other array, and that arrays normalized separately can still be compared
  7718. to each other without bias.
  7719. Such a normalization is commonly referred to as
  7720. \begin_inset Quotes eld
  7721. \end_inset
  7722. single-channel normalization
  7723. \begin_inset Quotes erd
  7724. \end_inset
  7725. .
  7726. \end_layout
  7727. \begin_layout Standard
  7728. \begin_inset Flex Glossary Term (Capital)
  7729. status open
  7730. \begin_layout Plain Layout
  7731. fRMA
  7732. \end_layout
  7733. \end_inset
  7734. addresses these concerns by replacing the quantile normalization and median
  7735. polish with alternatives that do not introduce inter-array dependence,
  7736. allowing each array to be normalized independently of all others
  7737. \begin_inset CommandInset citation
  7738. LatexCommand cite
  7739. key "McCall2010"
  7740. literal "false"
  7741. \end_inset
  7742. .
  7743. Quantile normalization is performed against a pre-generated set of quantiles
  7744. learned from a collection of 850 publicly available arrays sampled from
  7745. a wide variety of tissues in
  7746. \begin_inset ERT
  7747. status collapsed
  7748. \begin_layout Plain Layout
  7749. \backslash
  7750. glsdisp*{GEO}{the Gene Expression Omnibus (GEO)}
  7751. \end_layout
  7752. \end_inset
  7753. \begin_inset CommandInset nomenclature
  7754. LatexCommand nomenclature
  7755. symbol "GEO"
  7756. description "Gene Expression Omnibus"
  7757. literal "false"
  7758. \end_inset
  7759. .
  7760. Each array's probe intensity distribution is normalized against these pre-gener
  7761. ated quantiles.
  7762. The median polish step is replaced with a robust weighted average of probe
  7763. intensities, using inverse variance weights learned from the same public
  7764. \begin_inset Flex Glossary Term
  7765. status open
  7766. \begin_layout Plain Layout
  7767. GEO
  7768. \end_layout
  7769. \end_inset
  7770. data.
  7771. The result is a normalization that satisfies the requirements mentioned
  7772. above: each array is normalized independently of all others, and any two
  7773. normalized arrays can be compared directly to each other.
  7774. \end_layout
  7775. \begin_layout Standard
  7776. One important limitation of
  7777. \begin_inset Flex Glossary Term
  7778. status open
  7779. \begin_layout Plain Layout
  7780. fRMA
  7781. \end_layout
  7782. \end_inset
  7783. is that it requires a separate reference data set from which to learn the
  7784. parameters (reference quantiles and probe weights) that will be used to
  7785. normalize each array.
  7786. These parameters are specific to a given array platform, and pre-generated
  7787. parameters are only provided for the most common platforms, such as Affymetrix
  7788. hgu133plus2.
  7789. For a less common platform, such as hthgu133pluspm, is is necessary to
  7790. learn custom parameters from in-house data before
  7791. \begin_inset Flex Glossary Term
  7792. status open
  7793. \begin_layout Plain Layout
  7794. fRMA
  7795. \end_layout
  7796. \end_inset
  7797. can be used to normalize samples on that platform
  7798. \begin_inset CommandInset citation
  7799. LatexCommand cite
  7800. key "McCall2011"
  7801. literal "false"
  7802. \end_inset
  7803. .
  7804. \end_layout
  7805. \begin_layout Standard
  7806. One other option is the aptly-named
  7807. \begin_inset ERT
  7808. status open
  7809. \begin_layout Plain Layout
  7810. \backslash
  7811. glsdisp*{SCAN}{Single Channel Array Normalization (SCAN)}
  7812. \end_layout
  7813. \end_inset
  7814. , which adapts a normalization method originally designed for tiling arrays
  7815. \begin_inset CommandInset citation
  7816. LatexCommand cite
  7817. key "Piccolo2012"
  7818. literal "false"
  7819. \end_inset
  7820. .
  7821. \begin_inset Flex Glossary Term
  7822. status open
  7823. \begin_layout Plain Layout
  7824. SCAN
  7825. \end_layout
  7826. \end_inset
  7827. is truly single-channel in that it does not require a set of normalization
  7828. parameters estimated from an external set of reference samples like
  7829. \begin_inset Flex Glossary Term
  7830. status open
  7831. \begin_layout Plain Layout
  7832. fRMA
  7833. \end_layout
  7834. \end_inset
  7835. does.
  7836. \end_layout
  7837. \begin_layout Subsection
  7838. Heteroskedasticity must be accounted for in methylation array data
  7839. \end_layout
  7840. \begin_layout Standard
  7841. DNA methylation arrays are a relatively new kind of assay that uses microarrays
  7842. to measure the degree of methylation on cytosines in specific regions arrayed
  7843. across the genome.
  7844. First, bisulfite treatment converts all unmethylated cytosines to uracil
  7845. (which are read as thymine during amplification and sequencing) while leaving
  7846. methylated cytosines unaffected.
  7847. Then, each target region is interrogated with two probes: one binds to
  7848. the original genomic sequence and interrogates the level of methylated
  7849. DNA, and the other binds to the same sequence with all cytosines replaced
  7850. by thymidines and interrogates the level of unmethylated DNA.
  7851. \end_layout
  7852. \begin_layout Standard
  7853. \begin_inset Float figure
  7854. wide false
  7855. sideways false
  7856. status collapsed
  7857. \begin_layout Plain Layout
  7858. \align center
  7859. \begin_inset Graphics
  7860. filename graphics/methylvoom/sigmoid.pdf
  7861. lyxscale 50
  7862. width 60col%
  7863. groupId colwidth
  7864. \end_inset
  7865. \end_layout
  7866. \begin_layout Plain Layout
  7867. \begin_inset Caption Standard
  7868. \begin_layout Plain Layout
  7869. \begin_inset CommandInset label
  7870. LatexCommand label
  7871. name "fig:Sigmoid-beta-m-mapping"
  7872. \end_inset
  7873. \series bold
  7874. Sigmoid shape of the mapping between β and M values
  7875. \end_layout
  7876. \end_inset
  7877. \end_layout
  7878. \end_inset
  7879. \end_layout
  7880. \begin_layout Standard
  7881. After normalization, these two probe intensities are summarized in one of
  7882. two ways, each with advantages and disadvantages.
  7883. β
  7884. \series bold
  7885. \series default
  7886. values, interpreted as fraction of DNA copies methylated, range from 0 to
  7887. 1.
  7888. β
  7889. \series bold
  7890. \series default
  7891. values are conceptually easy to interpret, but the constrained range makes
  7892. them unsuitable for linear modeling, and their error distributions are
  7893. highly non-normal, which also frustrates linear modeling.
  7894. M-values, interpreted as the log ratio of methylated to unmethylated copies,
  7895. are computed by mapping the beta values from
  7896. \begin_inset Formula $[0,1]$
  7897. \end_inset
  7898. onto
  7899. \begin_inset Formula $(-\infty,+\infty)$
  7900. \end_inset
  7901. using a sigmoid curve (Figure
  7902. \begin_inset CommandInset ref
  7903. LatexCommand ref
  7904. reference "fig:Sigmoid-beta-m-mapping"
  7905. plural "false"
  7906. caps "false"
  7907. noprefix "false"
  7908. \end_inset
  7909. ).
  7910. This transformation results in values with better statistical properties:
  7911. the unconstrained range is suitable for linear modeling, and the error
  7912. distributions are more normal.
  7913. Hence, most linear modeling and other statistical testing on methylation
  7914. arrays is performed using M-values.
  7915. \end_layout
  7916. \begin_layout Standard
  7917. However, the steep slope of the sigmoid transformation near 0 and 1 tends
  7918. to over-exaggerate small differences in β values near those extremes, which
  7919. in turn amplifies the error in those values, leading to a U-shaped trend
  7920. in the mean-variance curve: extreme values have higher variances than values
  7921. near the middle.
  7922. This mean-variance dependency must be accounted for when fitting the linear
  7923. model for differential methylation, or else the variance will be systematically
  7924. overestimated for probes with moderate M-values and underestimated for
  7925. probes with extreme M-values.
  7926. This is particularly undesirable for methylation data because the intermediate
  7927. M-values are the ones of most interest, since they are more likely to represent
  7928. areas of varying methylation, whereas extreme M-values typically represent
  7929. complete methylation or complete lack of methylation.
  7930. \end_layout
  7931. \begin_layout Standard
  7932. \begin_inset Flex Glossary Term (Capital)
  7933. status open
  7934. \begin_layout Plain Layout
  7935. RNA-seq
  7936. \end_layout
  7937. \end_inset
  7938. read count data are also known to show heteroskedasticity, and the voom
  7939. method was introduced for modeling this heteroskedasticity by estimating
  7940. the mean-variance trend in the data and using this trend to assign precision
  7941. weights to each observation
  7942. \begin_inset CommandInset citation
  7943. LatexCommand cite
  7944. key "Law2013"
  7945. literal "false"
  7946. \end_inset
  7947. .
  7948. While methylation array data are not derived from counts and have a very
  7949. different mean-variance relationship from that of typical
  7950. \begin_inset Flex Glossary Term
  7951. status open
  7952. \begin_layout Plain Layout
  7953. RNA-seq
  7954. \end_layout
  7955. \end_inset
  7956. data, the voom method makes no specific assumptions on the shape of the
  7957. mean-variance relationship – it only assumes that the relationship can
  7958. be modeled as a smooth curve.
  7959. Hence, the method is sufficiently general to model the mean-variance relationsh
  7960. ip in methylation array data.
  7961. However, the standard implementation of voom assumes that the input is
  7962. given in raw read counts, and it must be adapted to run on methylation
  7963. M-values.
  7964. \end_layout
  7965. \begin_layout Section
  7966. Methods
  7967. \end_layout
  7968. \begin_layout Subsection
  7969. Evaluation of classifier performance with different normalization methods
  7970. \end_layout
  7971. \begin_layout Standard
  7972. For testing different expression microarray normalizations, a data set of
  7973. 157 hgu133plus2 arrays was used, consisting of blood samples from kidney
  7974. transplant patients whose grafts had been graded as
  7975. \begin_inset Flex Glossary Term
  7976. status open
  7977. \begin_layout Plain Layout
  7978. TX
  7979. \end_layout
  7980. \end_inset
  7981. ,
  7982. \begin_inset Flex Glossary Term
  7983. status open
  7984. \begin_layout Plain Layout
  7985. AR
  7986. \end_layout
  7987. \end_inset
  7988. , or
  7989. \begin_inset Flex Glossary Term
  7990. status open
  7991. \begin_layout Plain Layout
  7992. ADNR
  7993. \end_layout
  7994. \end_inset
  7995. via biopsy and histology (46 TX, 69 AR, 42 ADNR)
  7996. \begin_inset CommandInset citation
  7997. LatexCommand cite
  7998. key "Kurian2014"
  7999. literal "true"
  8000. \end_inset
  8001. .
  8002. Additionally, an external validation set of 75 samples was gathered from
  8003. public
  8004. \begin_inset Flex Glossary Term
  8005. status open
  8006. \begin_layout Plain Layout
  8007. GEO
  8008. \end_layout
  8009. \end_inset
  8010. data (37 TX, 38 AR, no ADNR).
  8011. \end_layout
  8012. \begin_layout Standard
  8013. \begin_inset Flex TODO Note (inline)
  8014. status open
  8015. \begin_layout Plain Layout
  8016. Find appropriate GEO identifiers if possible.
  8017. Kurian 2014 says GSE15296, but this seems to be different data.
  8018. I also need to look up the GEO accession for the external validation set.
  8019. \end_layout
  8020. \end_inset
  8021. \end_layout
  8022. \begin_layout Standard
  8023. To evaluate the effect of each normalization on classifier performance,
  8024. the same classifier training and validation procedure was used after each
  8025. normalization method.
  8026. The PAM package was used to train a nearest shrunken centroid classifier
  8027. on the training set and select the appropriate threshold for centroid shrinking.
  8028. Then the trained classifier was used to predict the class probabilities
  8029. of each validation sample.
  8030. From these class probabilities,
  8031. \begin_inset Flex Glossary Term
  8032. status open
  8033. \begin_layout Plain Layout
  8034. ROC
  8035. \end_layout
  8036. \end_inset
  8037. \begin_inset CommandInset nomenclature
  8038. LatexCommand nomenclature
  8039. symbol "ROC"
  8040. description "receiver operating characteristic"
  8041. literal "false"
  8042. \end_inset
  8043. curves and
  8044. \begin_inset Flex Glossary Term
  8045. status open
  8046. \begin_layout Plain Layout
  8047. AUC
  8048. \end_layout
  8049. \end_inset
  8050. \begin_inset CommandInset nomenclature
  8051. LatexCommand nomenclature
  8052. symbol "AUC"
  8053. description "area under ROC curve"
  8054. literal "false"
  8055. \end_inset
  8056. values were generated
  8057. \begin_inset CommandInset citation
  8058. LatexCommand cite
  8059. key "Turck2011"
  8060. literal "false"
  8061. \end_inset
  8062. .
  8063. Each normalization was tested on two different sets of training and validation
  8064. samples.
  8065. For internal validation, the 115
  8066. \begin_inset Flex Glossary Term
  8067. status open
  8068. \begin_layout Plain Layout
  8069. TX
  8070. \end_layout
  8071. \end_inset
  8072. and
  8073. \begin_inset Flex Glossary Term
  8074. status open
  8075. \begin_layout Plain Layout
  8076. AR
  8077. \end_layout
  8078. \end_inset
  8079. arrays in the internal set were split at random into two equal sized sets,
  8080. one for training and one for validation, each containing the same numbers
  8081. of
  8082. \begin_inset Flex Glossary Term
  8083. status open
  8084. \begin_layout Plain Layout
  8085. TX
  8086. \end_layout
  8087. \end_inset
  8088. and
  8089. \begin_inset Flex Glossary Term
  8090. status open
  8091. \begin_layout Plain Layout
  8092. AR
  8093. \end_layout
  8094. \end_inset
  8095. samples as the other set.
  8096. For external validation, the full set of 115
  8097. \begin_inset Flex Glossary Term
  8098. status open
  8099. \begin_layout Plain Layout
  8100. TX
  8101. \end_layout
  8102. \end_inset
  8103. and
  8104. \begin_inset Flex Glossary Term
  8105. status open
  8106. \begin_layout Plain Layout
  8107. AR
  8108. \end_layout
  8109. \end_inset
  8110. samples were used as a training set, and the 75 external
  8111. \begin_inset Flex Glossary Term
  8112. status open
  8113. \begin_layout Plain Layout
  8114. TX
  8115. \end_layout
  8116. \end_inset
  8117. and
  8118. \begin_inset Flex Glossary Term
  8119. status open
  8120. \begin_layout Plain Layout
  8121. AR
  8122. \end_layout
  8123. \end_inset
  8124. samples were used as the validation set.
  8125. Thus, 2
  8126. \begin_inset Flex Glossary Term
  8127. status open
  8128. \begin_layout Plain Layout
  8129. ROC
  8130. \end_layout
  8131. \end_inset
  8132. curves and
  8133. \begin_inset Flex Glossary Term
  8134. status open
  8135. \begin_layout Plain Layout
  8136. AUC
  8137. \end_layout
  8138. \end_inset
  8139. values were generated for each normalization method: one internal and one
  8140. external.
  8141. Because the external validation set contains no
  8142. \begin_inset Flex Glossary Term
  8143. status open
  8144. \begin_layout Plain Layout
  8145. ADNR
  8146. \end_layout
  8147. \end_inset
  8148. samples, only classification of
  8149. \begin_inset Flex Glossary Term
  8150. status open
  8151. \begin_layout Plain Layout
  8152. TX
  8153. \end_layout
  8154. \end_inset
  8155. and
  8156. \begin_inset Flex Glossary Term
  8157. status open
  8158. \begin_layout Plain Layout
  8159. AR
  8160. \end_layout
  8161. \end_inset
  8162. samples was considered.
  8163. The
  8164. \begin_inset Flex Glossary Term
  8165. status open
  8166. \begin_layout Plain Layout
  8167. ADNR
  8168. \end_layout
  8169. \end_inset
  8170. samples were included during normalization but excluded from all classifier
  8171. training and validation.
  8172. This ensures that the performance on internal and external validation sets
  8173. is directly comparable, since both are performing the same task: distinguishing
  8174. \begin_inset Flex Glossary Term
  8175. status open
  8176. \begin_layout Plain Layout
  8177. TX
  8178. \end_layout
  8179. \end_inset
  8180. from
  8181. \begin_inset Flex Glossary Term
  8182. status open
  8183. \begin_layout Plain Layout
  8184. AR
  8185. \end_layout
  8186. \end_inset
  8187. .
  8188. \end_layout
  8189. \begin_layout Standard
  8190. \begin_inset Flex TODO Note (inline)
  8191. status open
  8192. \begin_layout Plain Layout
  8193. Summarize the get.best.threshold algorithm for PAM threshold selection, or
  8194. just put the code online?
  8195. \end_layout
  8196. \end_inset
  8197. \end_layout
  8198. \begin_layout Standard
  8199. Six different normalization strategies were evaluated.
  8200. First, 2 well-known non-single-channel normalization methods were considered:
  8201. \begin_inset Flex Glossary Term
  8202. status open
  8203. \begin_layout Plain Layout
  8204. RMA
  8205. \end_layout
  8206. \end_inset
  8207. and dChip
  8208. \begin_inset CommandInset citation
  8209. LatexCommand cite
  8210. key "Li2001,Irizarry2003a"
  8211. literal "false"
  8212. \end_inset
  8213. .
  8214. Since
  8215. \begin_inset Flex Glossary Term
  8216. status open
  8217. \begin_layout Plain Layout
  8218. RMA
  8219. \end_layout
  8220. \end_inset
  8221. produces expression values on a
  8222. \begin_inset Formula $\log_{2}$
  8223. \end_inset
  8224. scale and dChip does not, the values from dChip were
  8225. \begin_inset Formula $\log_{2}$
  8226. \end_inset
  8227. transformed after normalization.
  8228. Next,
  8229. \begin_inset Flex Glossary Term
  8230. status open
  8231. \begin_layout Plain Layout
  8232. RMA
  8233. \end_layout
  8234. \end_inset
  8235. and dChip followed by
  8236. \begin_inset Flex Glossary Term
  8237. status open
  8238. \begin_layout Plain Layout
  8239. GRSN
  8240. \end_layout
  8241. \end_inset
  8242. were tested
  8243. \begin_inset CommandInset citation
  8244. LatexCommand cite
  8245. key "Pelz2008"
  8246. literal "false"
  8247. \end_inset
  8248. .
  8249. Post-processing with
  8250. \begin_inset Flex Glossary Term
  8251. status open
  8252. \begin_layout Plain Layout
  8253. GRSN
  8254. \end_layout
  8255. \end_inset
  8256. does not turn
  8257. \begin_inset Flex Glossary Term
  8258. status open
  8259. \begin_layout Plain Layout
  8260. RMA
  8261. \end_layout
  8262. \end_inset
  8263. or dChip into single-channel methods, but it may help mitigate batch effects
  8264. and is therefore useful as a benchmark.
  8265. Lastly, the two single-channel normalization methods,
  8266. \begin_inset Flex Glossary Term
  8267. status open
  8268. \begin_layout Plain Layout
  8269. fRMA
  8270. \end_layout
  8271. \end_inset
  8272. and
  8273. \begin_inset Flex Glossary Term
  8274. status open
  8275. \begin_layout Plain Layout
  8276. SCAN
  8277. \end_layout
  8278. \end_inset
  8279. , were tested
  8280. \begin_inset CommandInset citation
  8281. LatexCommand cite
  8282. key "McCall2010,Piccolo2012"
  8283. literal "false"
  8284. \end_inset
  8285. .
  8286. When evaluating internal validation performance, only the 157 internal
  8287. samples were normalized; when evaluating external validation performance,
  8288. all 157 internal samples and 75 external samples were normalized together.
  8289. \end_layout
  8290. \begin_layout Standard
  8291. For demonstrating the problem with separate normalization of training and
  8292. validation data, one additional normalization was performed: the internal
  8293. and external sets were each normalized separately using
  8294. \begin_inset Flex Glossary Term
  8295. status open
  8296. \begin_layout Plain Layout
  8297. RMA
  8298. \end_layout
  8299. \end_inset
  8300. , and the normalized data for each set were combined into a single set with
  8301. no further attempts at normalizing between the two sets.
  8302. The represents approximately how
  8303. \begin_inset Flex Glossary Term
  8304. status open
  8305. \begin_layout Plain Layout
  8306. RMA
  8307. \end_layout
  8308. \end_inset
  8309. would have to be used in a clinical setting, where the samples to be classified
  8310. are not available at the time the classifier is trained.
  8311. \end_layout
  8312. \begin_layout Subsection
  8313. Generating custom fRMA vectors for hthgu133pluspm array platform
  8314. \end_layout
  8315. \begin_layout Standard
  8316. In order to enable
  8317. \begin_inset Flex Glossary Term
  8318. status open
  8319. \begin_layout Plain Layout
  8320. fRMA
  8321. \end_layout
  8322. \end_inset
  8323. normalization for the hthgu133pluspm array platform, custom
  8324. \begin_inset Flex Glossary Term
  8325. status open
  8326. \begin_layout Plain Layout
  8327. fRMA
  8328. \end_layout
  8329. \end_inset
  8330. normalization vectors were trained using the
  8331. \begin_inset Flex Code
  8332. status open
  8333. \begin_layout Plain Layout
  8334. frmaTools
  8335. \end_layout
  8336. \end_inset
  8337. package
  8338. \begin_inset CommandInset citation
  8339. LatexCommand cite
  8340. key "McCall2011"
  8341. literal "false"
  8342. \end_inset
  8343. .
  8344. Separate vectors were created for two types of samples: kidney graft biopsy
  8345. samples and blood samples from graft recipients.
  8346. For training, a 341 kidney biopsy samples from 2 data sets and 965 blood
  8347. samples from 5 data sets were used as the reference set.
  8348. Arrays were groups into batches based on unique combinations of sample
  8349. type (blood or biopsy), diagnosis (TX, AR, etc.), data set, and scan date.
  8350. Thus, each batch represents arrays of the same kind that were run together
  8351. on the same day.
  8352. For estimating the probe inverse variance weights, frmaTools requires equal-siz
  8353. ed batches, which means a batch size must be chosen, and then batches smaller
  8354. than that size must be ignored, while batches larger than the chosen size
  8355. must be downsampled.
  8356. This downsampling is performed randomly, so the sampling process is repeated
  8357. 5 times and the resulting normalizations are compared to each other.
  8358. \end_layout
  8359. \begin_layout Standard
  8360. To evaluate the consistency of the generated normalization vectors, the
  8361. 5
  8362. \begin_inset Flex Glossary Term
  8363. status open
  8364. \begin_layout Plain Layout
  8365. fRMA
  8366. \end_layout
  8367. \end_inset
  8368. vector sets generated from 5 random batch samplings were each used to normalize
  8369. the same 20 randomly selected samples from each tissue.
  8370. Then the normalized expression values for each probe on each array were
  8371. compared across all normalizations.
  8372. Each
  8373. \begin_inset Flex Glossary Term
  8374. status open
  8375. \begin_layout Plain Layout
  8376. fRMA
  8377. \end_layout
  8378. \end_inset
  8379. normalization was also compared against the normalized expression values
  8380. obtained by normalizing the same 20 samples with ordinary
  8381. \begin_inset Flex Glossary Term
  8382. status open
  8383. \begin_layout Plain Layout
  8384. RMA
  8385. \end_layout
  8386. \end_inset
  8387. .
  8388. \end_layout
  8389. \begin_layout Subsection
  8390. Modeling methylation array M-value heteroskedasticy in linear models with
  8391. modified voom implementation
  8392. \end_layout
  8393. \begin_layout Standard
  8394. \begin_inset Flex TODO Note (inline)
  8395. status open
  8396. \begin_layout Plain Layout
  8397. Put code on Github and reference it.
  8398. \end_layout
  8399. \end_inset
  8400. \end_layout
  8401. \begin_layout Standard
  8402. To investigate the whether DNA methylation could be used to distinguish
  8403. between healthy and dysfunctional transplants, a data set of 78 Illumina
  8404. 450k methylation arrays from human kidney graft biopsies was analyzed for
  8405. differential methylation between 4 transplant statuses:
  8406. \begin_inset Flex Glossary Term
  8407. status open
  8408. \begin_layout Plain Layout
  8409. TX
  8410. \end_layout
  8411. \end_inset
  8412. , transplants undergoing
  8413. \begin_inset Flex Glossary Term
  8414. status open
  8415. \begin_layout Plain Layout
  8416. AR
  8417. \end_layout
  8418. \end_inset
  8419. ,
  8420. \begin_inset Flex Glossary Term
  8421. status open
  8422. \begin_layout Plain Layout
  8423. ADNR
  8424. \end_layout
  8425. \end_inset
  8426. , and
  8427. \begin_inset Flex Glossary Term
  8428. status open
  8429. \begin_layout Plain Layout
  8430. CAN
  8431. \end_layout
  8432. \end_inset
  8433. \begin_inset CommandInset nomenclature
  8434. LatexCommand nomenclature
  8435. symbol "CAN"
  8436. description "chronic allograft nephropathy"
  8437. literal "false"
  8438. \end_inset
  8439. .
  8440. The data consisted of 33 TX, 9 AR, 8 ADNR, and 28 CAN samples.
  8441. The uneven group sizes are a result of taking the biopsy samples before
  8442. the eventual fate of the transplant was known.
  8443. Each sample was additionally annotated with a donor ID (anonymized), sex,
  8444. age, ethnicity, creatinine level, and diabetes diagnosis (all samples in
  8445. this data set came from patients with either
  8446. \begin_inset Flex Glossary Term
  8447. status open
  8448. \begin_layout Plain Layout
  8449. T1D
  8450. \end_layout
  8451. \end_inset
  8452. \begin_inset CommandInset nomenclature
  8453. LatexCommand nomenclature
  8454. symbol "T1D"
  8455. description "Type 1 diabetes"
  8456. literal "false"
  8457. \end_inset
  8458. or
  8459. \begin_inset Flex Glossary Term
  8460. status open
  8461. \begin_layout Plain Layout
  8462. T2D
  8463. \end_layout
  8464. \end_inset
  8465. \begin_inset CommandInset nomenclature
  8466. LatexCommand nomenclature
  8467. symbol "T2D"
  8468. description "Type 2 diabetes"
  8469. literal "false"
  8470. \end_inset
  8471. ).
  8472. \end_layout
  8473. \begin_layout Standard
  8474. The intensity data were first normalized using
  8475. \begin_inset Flex Glossary Term
  8476. status open
  8477. \begin_layout Plain Layout
  8478. SWAN
  8479. \end_layout
  8480. \end_inset
  8481. \begin_inset CommandInset nomenclature
  8482. LatexCommand nomenclature
  8483. symbol "SWAN"
  8484. description "subset-quantile within array normalization"
  8485. literal "false"
  8486. \end_inset
  8487. \begin_inset CommandInset citation
  8488. LatexCommand cite
  8489. key "Maksimovic2012"
  8490. literal "false"
  8491. \end_inset
  8492. , then converted to intensity ratios (beta values)
  8493. \begin_inset CommandInset citation
  8494. LatexCommand cite
  8495. key "Aryee2014"
  8496. literal "false"
  8497. \end_inset
  8498. .
  8499. Any probes binding to loci that overlapped annotated SNPs were dropped,
  8500. and the annotated sex of each sample was verified against the sex inferred
  8501. from the ratio of median probe intensities for the X and Y chromosomes.
  8502. Then, the ratios were transformed to M-values.
  8503. \end_layout
  8504. \begin_layout Standard
  8505. \begin_inset Float table
  8506. wide false
  8507. sideways false
  8508. status open
  8509. \begin_layout Plain Layout
  8510. \align center
  8511. \begin_inset Tabular
  8512. <lyxtabular version="3" rows="4" columns="6">
  8513. <features tabularvalignment="middle">
  8514. <column alignment="center" valignment="top">
  8515. <column alignment="center" valignment="top">
  8516. <column alignment="center" valignment="top">
  8517. <column alignment="center" valignment="top">
  8518. <column alignment="center" valignment="top">
  8519. <column alignment="center" valignment="top">
  8520. <row>
  8521. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  8522. \begin_inset Text
  8523. \begin_layout Plain Layout
  8524. Analysis
  8525. \end_layout
  8526. \end_inset
  8527. </cell>
  8528. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  8529. \begin_inset Text
  8530. \begin_layout Plain Layout
  8531. random effect
  8532. \end_layout
  8533. \end_inset
  8534. </cell>
  8535. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  8536. \begin_inset Text
  8537. \begin_layout Plain Layout
  8538. eBayes
  8539. \end_layout
  8540. \end_inset
  8541. </cell>
  8542. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  8543. \begin_inset Text
  8544. \begin_layout Plain Layout
  8545. SVA
  8546. \end_layout
  8547. \end_inset
  8548. </cell>
  8549. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  8550. \begin_inset Text
  8551. \begin_layout Plain Layout
  8552. weights
  8553. \end_layout
  8554. \end_inset
  8555. </cell>
  8556. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  8557. \begin_inset Text
  8558. \begin_layout Plain Layout
  8559. voom
  8560. \end_layout
  8561. \end_inset
  8562. </cell>
  8563. </row>
  8564. <row>
  8565. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  8566. \begin_inset Text
  8567. \begin_layout Plain Layout
  8568. A
  8569. \end_layout
  8570. \end_inset
  8571. </cell>
  8572. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  8573. \begin_inset Text
  8574. \begin_layout Plain Layout
  8575. Yes
  8576. \end_layout
  8577. \end_inset
  8578. </cell>
  8579. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  8580. \begin_inset Text
  8581. \begin_layout Plain Layout
  8582. Yes
  8583. \end_layout
  8584. \end_inset
  8585. </cell>
  8586. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  8587. \begin_inset Text
  8588. \begin_layout Plain Layout
  8589. No
  8590. \end_layout
  8591. \end_inset
  8592. </cell>
  8593. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  8594. \begin_inset Text
  8595. \begin_layout Plain Layout
  8596. No
  8597. \end_layout
  8598. \end_inset
  8599. </cell>
  8600. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  8601. \begin_inset Text
  8602. \begin_layout Plain Layout
  8603. No
  8604. \end_layout
  8605. \end_inset
  8606. </cell>
  8607. </row>
  8608. <row>
  8609. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  8610. \begin_inset Text
  8611. \begin_layout Plain Layout
  8612. B
  8613. \end_layout
  8614. \end_inset
  8615. </cell>
  8616. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  8617. \begin_inset Text
  8618. \begin_layout Plain Layout
  8619. Yes
  8620. \end_layout
  8621. \end_inset
  8622. </cell>
  8623. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  8624. \begin_inset Text
  8625. \begin_layout Plain Layout
  8626. Yes
  8627. \end_layout
  8628. \end_inset
  8629. </cell>
  8630. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  8631. \begin_inset Text
  8632. \begin_layout Plain Layout
  8633. Yes
  8634. \end_layout
  8635. \end_inset
  8636. </cell>
  8637. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  8638. \begin_inset Text
  8639. \begin_layout Plain Layout
  8640. Yes
  8641. \end_layout
  8642. \end_inset
  8643. </cell>
  8644. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  8645. \begin_inset Text
  8646. \begin_layout Plain Layout
  8647. No
  8648. \end_layout
  8649. \end_inset
  8650. </cell>
  8651. </row>
  8652. <row>
  8653. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  8654. \begin_inset Text
  8655. \begin_layout Plain Layout
  8656. C
  8657. \end_layout
  8658. \end_inset
  8659. </cell>
  8660. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  8661. \begin_inset Text
  8662. \begin_layout Plain Layout
  8663. Yes
  8664. \end_layout
  8665. \end_inset
  8666. </cell>
  8667. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  8668. \begin_inset Text
  8669. \begin_layout Plain Layout
  8670. Yes
  8671. \end_layout
  8672. \end_inset
  8673. </cell>
  8674. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  8675. \begin_inset Text
  8676. \begin_layout Plain Layout
  8677. Yes
  8678. \end_layout
  8679. \end_inset
  8680. </cell>
  8681. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  8682. \begin_inset Text
  8683. \begin_layout Plain Layout
  8684. Yes
  8685. \end_layout
  8686. \end_inset
  8687. </cell>
  8688. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  8689. \begin_inset Text
  8690. \begin_layout Plain Layout
  8691. Yes
  8692. \end_layout
  8693. \end_inset
  8694. </cell>
  8695. </row>
  8696. </lyxtabular>
  8697. \end_inset
  8698. \end_layout
  8699. \begin_layout Plain Layout
  8700. \begin_inset Caption Standard
  8701. \begin_layout Plain Layout
  8702. \series bold
  8703. \begin_inset CommandInset label
  8704. LatexCommand label
  8705. name "tab:Summary-of-meth-analysis"
  8706. \end_inset
  8707. Summary of analysis variants for methylation array data.
  8708. \series default
  8709. Each analysis included a different set of steps to adjust or account for
  8710. various systematic features of the data.
  8711. Random effect: The model included a random effect accounting for correlation
  8712. between samples from the same patient
  8713. \begin_inset CommandInset citation
  8714. LatexCommand cite
  8715. key "Smyth2005a"
  8716. literal "false"
  8717. \end_inset
  8718. ; eBayes: Empirical bayes squeezing of per-probe variances toward the mean-varia
  8719. nce trend
  8720. \begin_inset CommandInset citation
  8721. LatexCommand cite
  8722. key "Ritchie2015"
  8723. literal "false"
  8724. \end_inset
  8725. ; SVA: Surrogate variable analysis to account for unobserved confounders
  8726. \begin_inset CommandInset citation
  8727. LatexCommand cite
  8728. key "Leek2007"
  8729. literal "false"
  8730. \end_inset
  8731. ; Weights: Estimate sample weights to account for differences in sample
  8732. quality
  8733. \begin_inset CommandInset citation
  8734. LatexCommand cite
  8735. key "Liu2015,Ritchie2006"
  8736. literal "false"
  8737. \end_inset
  8738. ; voom: Use mean-variance trend to assign individual sample weights
  8739. \begin_inset CommandInset citation
  8740. LatexCommand cite
  8741. key "Law2013"
  8742. literal "false"
  8743. \end_inset
  8744. .
  8745. See the text for a more detailed explanation of each step.
  8746. \end_layout
  8747. \end_inset
  8748. \end_layout
  8749. \end_inset
  8750. \end_layout
  8751. \begin_layout Standard
  8752. From the M-values, a series of parallel analyses was performed, each adding
  8753. additional steps into the model fit to accommodate a feature of the data
  8754. (see Table
  8755. \begin_inset CommandInset ref
  8756. LatexCommand ref
  8757. reference "tab:Summary-of-meth-analysis"
  8758. plural "false"
  8759. caps "false"
  8760. noprefix "false"
  8761. \end_inset
  8762. ).
  8763. For analysis A, a
  8764. \begin_inset Quotes eld
  8765. \end_inset
  8766. basic
  8767. \begin_inset Quotes erd
  8768. \end_inset
  8769. linear modeling analysis was performed, compensating for known confounders
  8770. by including terms for the factor of interest (transplant status) as well
  8771. as the known biological confounders: sex, age, ethnicity, and diabetes.
  8772. Since some samples came from the same patients at different times, the
  8773. intra-patient correlation was modeled as a random effect, estimating a
  8774. shared correlation value across all probes
  8775. \begin_inset CommandInset citation
  8776. LatexCommand cite
  8777. key "Smyth2005a"
  8778. literal "false"
  8779. \end_inset
  8780. .
  8781. Then the linear model was fit, and the variance was modeled using empirical
  8782. Bayes squeezing toward the mean-variance trend
  8783. \begin_inset CommandInset citation
  8784. LatexCommand cite
  8785. key "Ritchie2015"
  8786. literal "false"
  8787. \end_inset
  8788. .
  8789. Finally, t-tests or F-tests were performed as appropriate for each test:
  8790. t-tests for single contrasts, and F-tests for multiple contrasts.
  8791. P-values were corrected for multiple testing using the
  8792. \begin_inset Flex Glossary Term
  8793. status open
  8794. \begin_layout Plain Layout
  8795. BH
  8796. \end_layout
  8797. \end_inset
  8798. procedure for
  8799. \begin_inset Flex Glossary Term
  8800. status open
  8801. \begin_layout Plain Layout
  8802. FDR
  8803. \end_layout
  8804. \end_inset
  8805. control
  8806. \begin_inset CommandInset citation
  8807. LatexCommand cite
  8808. key "Benjamini1995"
  8809. literal "false"
  8810. \end_inset
  8811. .
  8812. \end_layout
  8813. \begin_layout Standard
  8814. For the analysis B,
  8815. \begin_inset Flex Glossary Term
  8816. status open
  8817. \begin_layout Plain Layout
  8818. SVA
  8819. \end_layout
  8820. \end_inset
  8821. was used to infer additional unobserved sources of heterogeneity in the
  8822. data
  8823. \begin_inset CommandInset citation
  8824. LatexCommand cite
  8825. key "Leek2007"
  8826. literal "false"
  8827. \end_inset
  8828. .
  8829. These surrogate variables were added to the design matrix before fitting
  8830. the linear model.
  8831. In addition, sample quality weights were estimated from the data and used
  8832. during linear modeling to down-weight the contribution of highly variable
  8833. arrays while increasing the weight to arrays with lower variability
  8834. \begin_inset CommandInset citation
  8835. LatexCommand cite
  8836. key "Ritchie2006"
  8837. literal "false"
  8838. \end_inset
  8839. .
  8840. The remainder of the analysis proceeded as in analysis A.
  8841. For analysis C, the voom method was adapted to run on methylation array
  8842. data and used to model and correct for the mean-variance trend using individual
  8843. observation weights
  8844. \begin_inset CommandInset citation
  8845. LatexCommand cite
  8846. key "Law2013"
  8847. literal "false"
  8848. \end_inset
  8849. , which were combined with the sample weights
  8850. \begin_inset CommandInset citation
  8851. LatexCommand cite
  8852. key "Liu2015,Ritchie2006"
  8853. literal "false"
  8854. \end_inset
  8855. .
  8856. Each time weights were used, they were estimated once before estimating
  8857. the random effect correlation value, and then the weights were re-estimated
  8858. taking the random effect into account.
  8859. The remainder of the analysis proceeded as in analysis B.
  8860. \end_layout
  8861. \begin_layout Section
  8862. Results
  8863. \end_layout
  8864. \begin_layout Standard
  8865. \begin_inset Flex TODO Note (inline)
  8866. status open
  8867. \begin_layout Plain Layout
  8868. Improve subsection titles in this section.
  8869. \end_layout
  8870. \end_inset
  8871. \end_layout
  8872. \begin_layout Standard
  8873. \begin_inset Flex TODO Note (inline)
  8874. status open
  8875. \begin_layout Plain Layout
  8876. Reconsider subsection organization?
  8877. \end_layout
  8878. \end_inset
  8879. \end_layout
  8880. \begin_layout Subsection
  8881. Separate normalization with RMA introduces unwanted biases in classification
  8882. \end_layout
  8883. \begin_layout Standard
  8884. \begin_inset Float figure
  8885. wide false
  8886. sideways false
  8887. status open
  8888. \begin_layout Plain Layout
  8889. \align center
  8890. \begin_inset Graphics
  8891. filename graphics/PAM/predplot.pdf
  8892. lyxscale 50
  8893. width 60col%
  8894. groupId colwidth
  8895. \end_inset
  8896. \end_layout
  8897. \begin_layout Plain Layout
  8898. \begin_inset Caption Standard
  8899. \begin_layout Plain Layout
  8900. \begin_inset CommandInset label
  8901. LatexCommand label
  8902. name "fig:Classifier-probabilities-RMA"
  8903. \end_inset
  8904. \series bold
  8905. Classifier probabilities on validation samples when normalized with RMA
  8906. together vs.
  8907. separately.
  8908. \series default
  8909. The PAM classifier algorithm was trained on the training set of arrays to
  8910. distinguish AR from TX and then used to assign class probabilities to the
  8911. validation set.
  8912. The process was performed after normalizing all samples together and after
  8913. normalizing the training and test sets separately, and the class probabilities
  8914. assigned to each sample in the validation set were plotted against each
  8915. other (PP(AR), posterior probability of being AR).
  8916. The color of each point indicates the true classification of that sample.
  8917. \end_layout
  8918. \end_inset
  8919. \end_layout
  8920. \end_inset
  8921. \end_layout
  8922. \begin_layout Standard
  8923. To demonstrate the problem with non-single-channel normalization methods,
  8924. we considered the problem of training a classifier to distinguish
  8925. \begin_inset Flex Glossary Term
  8926. status open
  8927. \begin_layout Plain Layout
  8928. TX
  8929. \end_layout
  8930. \end_inset
  8931. from
  8932. \begin_inset Flex Glossary Term
  8933. status open
  8934. \begin_layout Plain Layout
  8935. AR
  8936. \end_layout
  8937. \end_inset
  8938. using the samples from the internal set as training data, evaluating performanc
  8939. e on the external set.
  8940. First, training and evaluation were performed after normalizing all array
  8941. samples together as a single set using
  8942. \begin_inset Flex Glossary Term
  8943. status open
  8944. \begin_layout Plain Layout
  8945. RMA
  8946. \end_layout
  8947. \end_inset
  8948. , and second, the internal samples were normalized separately from the external
  8949. samples and the training and evaluation were repeated.
  8950. For each sample in the validation set, the classifier probabilities from
  8951. both classifiers were plotted against each other (Fig.
  8952. \begin_inset CommandInset ref
  8953. LatexCommand ref
  8954. reference "fig:Classifier-probabilities-RMA"
  8955. plural "false"
  8956. caps "false"
  8957. noprefix "false"
  8958. \end_inset
  8959. ).
  8960. As expected, separate normalization biases the classifier probabilities,
  8961. resulting in several misclassifications.
  8962. In this case, the bias from separate normalization causes the classifier
  8963. to assign a lower probability of
  8964. \begin_inset Flex Glossary Term
  8965. status open
  8966. \begin_layout Plain Layout
  8967. AR
  8968. \end_layout
  8969. \end_inset
  8970. to every sample.
  8971. \end_layout
  8972. \begin_layout Subsection
  8973. fRMA and SCAN maintain classification performance while eliminating dependence
  8974. on normalization strategy
  8975. \end_layout
  8976. \begin_layout Standard
  8977. \begin_inset Float figure
  8978. wide false
  8979. sideways false
  8980. status open
  8981. \begin_layout Plain Layout
  8982. \align center
  8983. \begin_inset Float figure
  8984. placement tb
  8985. wide false
  8986. sideways false
  8987. status open
  8988. \begin_layout Plain Layout
  8989. \align center
  8990. \begin_inset Graphics
  8991. filename graphics/PAM/ROC-TXvsAR-internal.pdf
  8992. lyxscale 50
  8993. height 40theight%
  8994. groupId roc-pam
  8995. \end_inset
  8996. \end_layout
  8997. \begin_layout Plain Layout
  8998. \begin_inset Caption Standard
  8999. \begin_layout Plain Layout
  9000. \begin_inset CommandInset label
  9001. LatexCommand label
  9002. name "fig:ROC-PAM-int"
  9003. \end_inset
  9004. ROC curves for PAM on internal validation data
  9005. \end_layout
  9006. \end_inset
  9007. \end_layout
  9008. \end_inset
  9009. \end_layout
  9010. \begin_layout Plain Layout
  9011. \align center
  9012. \begin_inset Float figure
  9013. placement tb
  9014. wide false
  9015. sideways false
  9016. status open
  9017. \begin_layout Plain Layout
  9018. \align center
  9019. \begin_inset Graphics
  9020. filename graphics/PAM/ROC-TXvsAR-external.pdf
  9021. lyxscale 50
  9022. height 40theight%
  9023. groupId roc-pam
  9024. \end_inset
  9025. \end_layout
  9026. \begin_layout Plain Layout
  9027. \begin_inset Caption Standard
  9028. \begin_layout Plain Layout
  9029. \begin_inset CommandInset label
  9030. LatexCommand label
  9031. name "fig:ROC-PAM-ext"
  9032. \end_inset
  9033. ROC curves for PAM on external validation data
  9034. \end_layout
  9035. \end_inset
  9036. \end_layout
  9037. \end_inset
  9038. \end_layout
  9039. \begin_layout Plain Layout
  9040. \begin_inset Caption Standard
  9041. \begin_layout Plain Layout
  9042. \series bold
  9043. \begin_inset CommandInset label
  9044. LatexCommand label
  9045. name "fig:ROC-PAM-main"
  9046. \end_inset
  9047. ROC curves for PAM using different normalization strategies.
  9048. \series default
  9049. ROC curves were generated for PAM classification of AR vs TX after 6 different
  9050. normalization strategies applied to the same data sets.
  9051. Only fRMA and SCAN are single-channel normalizations.
  9052. The other normalizations are for comparison.
  9053. \end_layout
  9054. \end_inset
  9055. \end_layout
  9056. \end_inset
  9057. \end_layout
  9058. \begin_layout Standard
  9059. \begin_inset Float table
  9060. wide false
  9061. sideways false
  9062. status open
  9063. \begin_layout Plain Layout
  9064. \align center
  9065. \begin_inset Tabular
  9066. <lyxtabular version="3" rows="7" columns="4">
  9067. <features tabularvalignment="middle">
  9068. <column alignment="center" valignment="top">
  9069. <column alignment="center" valignment="top">
  9070. <column alignment="center" valignment="top">
  9071. <column alignment="center" valignment="top">
  9072. <row>
  9073. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  9074. \begin_inset Text
  9075. \begin_layout Plain Layout
  9076. \family roman
  9077. \series medium
  9078. \shape up
  9079. \size normal
  9080. \emph off
  9081. \bar no
  9082. \strikeout off
  9083. \xout off
  9084. \uuline off
  9085. \uwave off
  9086. \noun off
  9087. \color none
  9088. Normalization
  9089. \end_layout
  9090. \end_inset
  9091. </cell>
  9092. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  9093. \begin_inset Text
  9094. \begin_layout Plain Layout
  9095. Single-channel?
  9096. \end_layout
  9097. \end_inset
  9098. </cell>
  9099. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  9100. \begin_inset Text
  9101. \begin_layout Plain Layout
  9102. \family roman
  9103. \series medium
  9104. \shape up
  9105. \size normal
  9106. \emph off
  9107. \bar no
  9108. \strikeout off
  9109. \xout off
  9110. \uuline off
  9111. \uwave off
  9112. \noun off
  9113. \color none
  9114. Internal Val.
  9115. AUC
  9116. \end_layout
  9117. \end_inset
  9118. </cell>
  9119. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  9120. \begin_inset Text
  9121. \begin_layout Plain Layout
  9122. External Val.
  9123. AUC
  9124. \end_layout
  9125. \end_inset
  9126. </cell>
  9127. </row>
  9128. <row>
  9129. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  9130. \begin_inset Text
  9131. \begin_layout Plain Layout
  9132. \family roman
  9133. \series medium
  9134. \shape up
  9135. \size normal
  9136. \emph off
  9137. \bar no
  9138. \strikeout off
  9139. \xout off
  9140. \uuline off
  9141. \uwave off
  9142. \noun off
  9143. \color none
  9144. RMA
  9145. \end_layout
  9146. \end_inset
  9147. </cell>
  9148. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  9149. \begin_inset Text
  9150. \begin_layout Plain Layout
  9151. No
  9152. \end_layout
  9153. \end_inset
  9154. </cell>
  9155. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  9156. \begin_inset Text
  9157. \begin_layout Plain Layout
  9158. \family roman
  9159. \series medium
  9160. \shape up
  9161. \size normal
  9162. \emph off
  9163. \bar no
  9164. \strikeout off
  9165. \xout off
  9166. \uuline off
  9167. \uwave off
  9168. \noun off
  9169. \color none
  9170. 0.852
  9171. \end_layout
  9172. \end_inset
  9173. </cell>
  9174. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  9175. \begin_inset Text
  9176. \begin_layout Plain Layout
  9177. \family roman
  9178. \series medium
  9179. \shape up
  9180. \size normal
  9181. \emph off
  9182. \bar no
  9183. \strikeout off
  9184. \xout off
  9185. \uuline off
  9186. \uwave off
  9187. \noun off
  9188. \color none
  9189. 0.713
  9190. \end_layout
  9191. \end_inset
  9192. </cell>
  9193. </row>
  9194. <row>
  9195. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  9196. \begin_inset Text
  9197. \begin_layout Plain Layout
  9198. \family roman
  9199. \series medium
  9200. \shape up
  9201. \size normal
  9202. \emph off
  9203. \bar no
  9204. \strikeout off
  9205. \xout off
  9206. \uuline off
  9207. \uwave off
  9208. \noun off
  9209. \color none
  9210. dChip
  9211. \end_layout
  9212. \end_inset
  9213. </cell>
  9214. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  9215. \begin_inset Text
  9216. \begin_layout Plain Layout
  9217. No
  9218. \end_layout
  9219. \end_inset
  9220. </cell>
  9221. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  9222. \begin_inset Text
  9223. \begin_layout Plain Layout
  9224. \family roman
  9225. \series medium
  9226. \shape up
  9227. \size normal
  9228. \emph off
  9229. \bar no
  9230. \strikeout off
  9231. \xout off
  9232. \uuline off
  9233. \uwave off
  9234. \noun off
  9235. \color none
  9236. 0.891
  9237. \end_layout
  9238. \end_inset
  9239. </cell>
  9240. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  9241. \begin_inset Text
  9242. \begin_layout Plain Layout
  9243. \family roman
  9244. \series medium
  9245. \shape up
  9246. \size normal
  9247. \emph off
  9248. \bar no
  9249. \strikeout off
  9250. \xout off
  9251. \uuline off
  9252. \uwave off
  9253. \noun off
  9254. \color none
  9255. 0.657
  9256. \end_layout
  9257. \end_inset
  9258. </cell>
  9259. </row>
  9260. <row>
  9261. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  9262. \begin_inset Text
  9263. \begin_layout Plain Layout
  9264. \family roman
  9265. \series medium
  9266. \shape up
  9267. \size normal
  9268. \emph off
  9269. \bar no
  9270. \strikeout off
  9271. \xout off
  9272. \uuline off
  9273. \uwave off
  9274. \noun off
  9275. \color none
  9276. RMA + GRSN
  9277. \end_layout
  9278. \end_inset
  9279. </cell>
  9280. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  9281. \begin_inset Text
  9282. \begin_layout Plain Layout
  9283. No
  9284. \end_layout
  9285. \end_inset
  9286. </cell>
  9287. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  9288. \begin_inset Text
  9289. \begin_layout Plain Layout
  9290. \family roman
  9291. \series medium
  9292. \shape up
  9293. \size normal
  9294. \emph off
  9295. \bar no
  9296. \strikeout off
  9297. \xout off
  9298. \uuline off
  9299. \uwave off
  9300. \noun off
  9301. \color none
  9302. 0.816
  9303. \end_layout
  9304. \end_inset
  9305. </cell>
  9306. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  9307. \begin_inset Text
  9308. \begin_layout Plain Layout
  9309. \family roman
  9310. \series medium
  9311. \shape up
  9312. \size normal
  9313. \emph off
  9314. \bar no
  9315. \strikeout off
  9316. \xout off
  9317. \uuline off
  9318. \uwave off
  9319. \noun off
  9320. \color none
  9321. 0.750
  9322. \end_layout
  9323. \end_inset
  9324. </cell>
  9325. </row>
  9326. <row>
  9327. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  9328. \begin_inset Text
  9329. \begin_layout Plain Layout
  9330. \family roman
  9331. \series medium
  9332. \shape up
  9333. \size normal
  9334. \emph off
  9335. \bar no
  9336. \strikeout off
  9337. \xout off
  9338. \uuline off
  9339. \uwave off
  9340. \noun off
  9341. \color none
  9342. dChip + GRSN
  9343. \end_layout
  9344. \end_inset
  9345. </cell>
  9346. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  9347. \begin_inset Text
  9348. \begin_layout Plain Layout
  9349. No
  9350. \end_layout
  9351. \end_inset
  9352. </cell>
  9353. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  9354. \begin_inset Text
  9355. \begin_layout Plain Layout
  9356. \family roman
  9357. \series medium
  9358. \shape up
  9359. \size normal
  9360. \emph off
  9361. \bar no
  9362. \strikeout off
  9363. \xout off
  9364. \uuline off
  9365. \uwave off
  9366. \noun off
  9367. \color none
  9368. 0.875
  9369. \end_layout
  9370. \end_inset
  9371. </cell>
  9372. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  9373. \begin_inset Text
  9374. \begin_layout Plain Layout
  9375. \family roman
  9376. \series medium
  9377. \shape up
  9378. \size normal
  9379. \emph off
  9380. \bar no
  9381. \strikeout off
  9382. \xout off
  9383. \uuline off
  9384. \uwave off
  9385. \noun off
  9386. \color none
  9387. 0.642
  9388. \end_layout
  9389. \end_inset
  9390. </cell>
  9391. </row>
  9392. <row>
  9393. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  9394. \begin_inset Text
  9395. \begin_layout Plain Layout
  9396. \family roman
  9397. \series medium
  9398. \shape up
  9399. \size normal
  9400. \emph off
  9401. \bar no
  9402. \strikeout off
  9403. \xout off
  9404. \uuline off
  9405. \uwave off
  9406. \noun off
  9407. \color none
  9408. fRMA
  9409. \end_layout
  9410. \end_inset
  9411. </cell>
  9412. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  9413. \begin_inset Text
  9414. \begin_layout Plain Layout
  9415. Yes
  9416. \end_layout
  9417. \end_inset
  9418. </cell>
  9419. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  9420. \begin_inset Text
  9421. \begin_layout Plain Layout
  9422. \family roman
  9423. \series medium
  9424. \shape up
  9425. \size normal
  9426. \emph off
  9427. \bar no
  9428. \strikeout off
  9429. \xout off
  9430. \uuline off
  9431. \uwave off
  9432. \noun off
  9433. \color none
  9434. 0.863
  9435. \end_layout
  9436. \end_inset
  9437. </cell>
  9438. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  9439. \begin_inset Text
  9440. \begin_layout Plain Layout
  9441. \family roman
  9442. \series medium
  9443. \shape up
  9444. \size normal
  9445. \emph off
  9446. \bar no
  9447. \strikeout off
  9448. \xout off
  9449. \uuline off
  9450. \uwave off
  9451. \noun off
  9452. \color none
  9453. 0.718
  9454. \end_layout
  9455. \end_inset
  9456. </cell>
  9457. </row>
  9458. <row>
  9459. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  9460. \begin_inset Text
  9461. \begin_layout Plain Layout
  9462. \family roman
  9463. \series medium
  9464. \shape up
  9465. \size normal
  9466. \emph off
  9467. \bar no
  9468. \strikeout off
  9469. \xout off
  9470. \uuline off
  9471. \uwave off
  9472. \noun off
  9473. \color none
  9474. SCAN
  9475. \end_layout
  9476. \end_inset
  9477. </cell>
  9478. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  9479. \begin_inset Text
  9480. \begin_layout Plain Layout
  9481. Yes
  9482. \end_layout
  9483. \end_inset
  9484. </cell>
  9485. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  9486. \begin_inset Text
  9487. \begin_layout Plain Layout
  9488. \family roman
  9489. \series medium
  9490. \shape up
  9491. \size normal
  9492. \emph off
  9493. \bar no
  9494. \strikeout off
  9495. \xout off
  9496. \uuline off
  9497. \uwave off
  9498. \noun off
  9499. \color none
  9500. 0.853
  9501. \end_layout
  9502. \end_inset
  9503. </cell>
  9504. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  9505. \begin_inset Text
  9506. \begin_layout Plain Layout
  9507. \family roman
  9508. \series medium
  9509. \shape up
  9510. \size normal
  9511. \emph off
  9512. \bar no
  9513. \strikeout off
  9514. \xout off
  9515. \uuline off
  9516. \uwave off
  9517. \noun off
  9518. \color none
  9519. 0.689
  9520. \end_layout
  9521. \end_inset
  9522. </cell>
  9523. </row>
  9524. </lyxtabular>
  9525. \end_inset
  9526. \end_layout
  9527. \begin_layout Plain Layout
  9528. \begin_inset Caption Standard
  9529. \begin_layout Plain Layout
  9530. \begin_inset CommandInset label
  9531. LatexCommand label
  9532. name "tab:AUC-PAM"
  9533. \end_inset
  9534. \series bold
  9535. ROC curve AUC values for internal and external validation with 6 different
  9536. normalization strategies.
  9537. \series default
  9538. These AUC values correspond to the ROC curves in Figure
  9539. \begin_inset CommandInset ref
  9540. LatexCommand ref
  9541. reference "fig:ROC-PAM-main"
  9542. plural "false"
  9543. caps "false"
  9544. noprefix "false"
  9545. \end_inset
  9546. .
  9547. \end_layout
  9548. \end_inset
  9549. \end_layout
  9550. \end_inset
  9551. \end_layout
  9552. \begin_layout Standard
  9553. For internal validation, the 6 methods' AUC values ranged from 0.816 to 0.891,
  9554. as shown in Table
  9555. \begin_inset CommandInset ref
  9556. LatexCommand ref
  9557. reference "tab:AUC-PAM"
  9558. plural "false"
  9559. caps "false"
  9560. noprefix "false"
  9561. \end_inset
  9562. .
  9563. Among the non-single-channel normalizations, dChip outperformed
  9564. \begin_inset Flex Glossary Term
  9565. status open
  9566. \begin_layout Plain Layout
  9567. RMA
  9568. \end_layout
  9569. \end_inset
  9570. , while
  9571. \begin_inset Flex Glossary Term
  9572. status open
  9573. \begin_layout Plain Layout
  9574. GRSN
  9575. \end_layout
  9576. \end_inset
  9577. reduced the
  9578. \begin_inset Flex Glossary Term
  9579. status open
  9580. \begin_layout Plain Layout
  9581. AUC
  9582. \end_layout
  9583. \end_inset
  9584. values for both dChip and
  9585. \begin_inset Flex Glossary Term
  9586. status open
  9587. \begin_layout Plain Layout
  9588. RMA
  9589. \end_layout
  9590. \end_inset
  9591. .
  9592. Both single-channel methods,
  9593. \begin_inset Flex Glossary Term
  9594. status open
  9595. \begin_layout Plain Layout
  9596. fRMA
  9597. \end_layout
  9598. \end_inset
  9599. and
  9600. \begin_inset Flex Glossary Term
  9601. status open
  9602. \begin_layout Plain Layout
  9603. SCAN
  9604. \end_layout
  9605. \end_inset
  9606. , slightly outperformed
  9607. \begin_inset Flex Glossary Term
  9608. status open
  9609. \begin_layout Plain Layout
  9610. RMA
  9611. \end_layout
  9612. \end_inset
  9613. , with
  9614. \begin_inset Flex Glossary Term
  9615. status open
  9616. \begin_layout Plain Layout
  9617. fRMA
  9618. \end_layout
  9619. \end_inset
  9620. ahead of
  9621. \begin_inset Flex Glossary Term
  9622. status open
  9623. \begin_layout Plain Layout
  9624. SCAN
  9625. \end_layout
  9626. \end_inset
  9627. .
  9628. However, the difference between
  9629. \begin_inset Flex Glossary Term
  9630. status open
  9631. \begin_layout Plain Layout
  9632. RMA
  9633. \end_layout
  9634. \end_inset
  9635. and
  9636. \begin_inset Flex Glossary Term
  9637. status open
  9638. \begin_layout Plain Layout
  9639. fRMA
  9640. \end_layout
  9641. \end_inset
  9642. is still quite small.
  9643. Figure
  9644. \begin_inset CommandInset ref
  9645. LatexCommand ref
  9646. reference "fig:ROC-PAM-int"
  9647. plural "false"
  9648. caps "false"
  9649. noprefix "false"
  9650. \end_inset
  9651. shows that the
  9652. \begin_inset Flex Glossary Term
  9653. status open
  9654. \begin_layout Plain Layout
  9655. ROC
  9656. \end_layout
  9657. \end_inset
  9658. curves for
  9659. \begin_inset Flex Glossary Term
  9660. status open
  9661. \begin_layout Plain Layout
  9662. RMA
  9663. \end_layout
  9664. \end_inset
  9665. , dChip, and
  9666. \begin_inset Flex Glossary Term
  9667. status open
  9668. \begin_layout Plain Layout
  9669. fRMA
  9670. \end_layout
  9671. \end_inset
  9672. look very similar and relatively smooth, while both
  9673. \begin_inset Flex Glossary Term
  9674. status open
  9675. \begin_layout Plain Layout
  9676. GRSN
  9677. \end_layout
  9678. \end_inset
  9679. curves and the curve for
  9680. \begin_inset Flex Glossary Term
  9681. status open
  9682. \begin_layout Plain Layout
  9683. SCAN
  9684. \end_layout
  9685. \end_inset
  9686. have a more jagged appearance.
  9687. \end_layout
  9688. \begin_layout Standard
  9689. For external validation, as expected, all the
  9690. \begin_inset Flex Glossary Term
  9691. status open
  9692. \begin_layout Plain Layout
  9693. AUC
  9694. \end_layout
  9695. \end_inset
  9696. values are lower than the internal validations, ranging from 0.642 to 0.750
  9697. (Table
  9698. \begin_inset CommandInset ref
  9699. LatexCommand ref
  9700. reference "tab:AUC-PAM"
  9701. plural "false"
  9702. caps "false"
  9703. noprefix "false"
  9704. \end_inset
  9705. ).
  9706. With or without
  9707. \begin_inset Flex Glossary Term
  9708. status open
  9709. \begin_layout Plain Layout
  9710. GRSN
  9711. \end_layout
  9712. \end_inset
  9713. ,
  9714. \begin_inset Flex Glossary Term
  9715. status open
  9716. \begin_layout Plain Layout
  9717. RMA
  9718. \end_layout
  9719. \end_inset
  9720. shows its dominance over dChip in this more challenging test.
  9721. Unlike in the internal validation,
  9722. \begin_inset Flex Glossary Term
  9723. status open
  9724. \begin_layout Plain Layout
  9725. GRSN
  9726. \end_layout
  9727. \end_inset
  9728. actually improves the classifier performance for
  9729. \begin_inset Flex Glossary Term
  9730. status open
  9731. \begin_layout Plain Layout
  9732. RMA
  9733. \end_layout
  9734. \end_inset
  9735. , although it does not for dChip.
  9736. Once again, both single-channel methods perform about on par with
  9737. \begin_inset Flex Glossary Term
  9738. status open
  9739. \begin_layout Plain Layout
  9740. RMA
  9741. \end_layout
  9742. \end_inset
  9743. , with
  9744. \begin_inset Flex Glossary Term
  9745. status open
  9746. \begin_layout Plain Layout
  9747. fRMA
  9748. \end_layout
  9749. \end_inset
  9750. performing slightly better and
  9751. \begin_inset Flex Glossary Term
  9752. status open
  9753. \begin_layout Plain Layout
  9754. SCAN
  9755. \end_layout
  9756. \end_inset
  9757. performing a bit worse.
  9758. Figure
  9759. \begin_inset CommandInset ref
  9760. LatexCommand ref
  9761. reference "fig:ROC-PAM-ext"
  9762. plural "false"
  9763. caps "false"
  9764. noprefix "false"
  9765. \end_inset
  9766. shows the
  9767. \begin_inset Flex Glossary Term
  9768. status open
  9769. \begin_layout Plain Layout
  9770. ROC
  9771. \end_layout
  9772. \end_inset
  9773. curves for the external validation test.
  9774. As expected, none of them are as clean-looking as the internal validation
  9775. \begin_inset Flex Glossary Term
  9776. status open
  9777. \begin_layout Plain Layout
  9778. ROC
  9779. \end_layout
  9780. \end_inset
  9781. curves.
  9782. The curves for
  9783. \begin_inset Flex Glossary Term
  9784. status open
  9785. \begin_layout Plain Layout
  9786. RMA
  9787. \end_layout
  9788. \end_inset
  9789. , RMA+GRSN, and
  9790. \begin_inset Flex Glossary Term
  9791. status open
  9792. \begin_layout Plain Layout
  9793. fRMA
  9794. \end_layout
  9795. \end_inset
  9796. all look similar, while the other curves look more divergent.
  9797. \end_layout
  9798. \begin_layout Subsection
  9799. fRMA with custom-generated vectors enables single-channel normalization
  9800. on hthgu133pluspm platform
  9801. \end_layout
  9802. \begin_layout Standard
  9803. \begin_inset Float figure
  9804. wide false
  9805. sideways false
  9806. status open
  9807. \begin_layout Plain Layout
  9808. \align center
  9809. \begin_inset Float figure
  9810. placement tb
  9811. wide false
  9812. sideways false
  9813. status collapsed
  9814. \begin_layout Plain Layout
  9815. \align center
  9816. \begin_inset Graphics
  9817. filename graphics/frma-pax-bx/batchsize_batches.pdf
  9818. lyxscale 50
  9819. height 35theight%
  9820. groupId frmatools-subfig
  9821. \end_inset
  9822. \end_layout
  9823. \begin_layout Plain Layout
  9824. \begin_inset Caption Standard
  9825. \begin_layout Plain Layout
  9826. \begin_inset CommandInset label
  9827. LatexCommand label
  9828. name "fig:batch-size-batches"
  9829. \end_inset
  9830. \series bold
  9831. Number of batches usable in fRMA probe weight learning as a function of
  9832. batch size.
  9833. \end_layout
  9834. \end_inset
  9835. \end_layout
  9836. \end_inset
  9837. \end_layout
  9838. \begin_layout Plain Layout
  9839. \align center
  9840. \begin_inset Float figure
  9841. placement tb
  9842. wide false
  9843. sideways false
  9844. status collapsed
  9845. \begin_layout Plain Layout
  9846. \align center
  9847. \begin_inset Graphics
  9848. filename graphics/frma-pax-bx/batchsize_samples.pdf
  9849. lyxscale 50
  9850. height 35theight%
  9851. groupId frmatools-subfig
  9852. \end_inset
  9853. \end_layout
  9854. \begin_layout Plain Layout
  9855. \begin_inset Caption Standard
  9856. \begin_layout Plain Layout
  9857. \begin_inset CommandInset label
  9858. LatexCommand label
  9859. name "fig:batch-size-samples"
  9860. \end_inset
  9861. \series bold
  9862. Number of samples usable in fRMA probe weight learning as a function of
  9863. batch size.
  9864. \end_layout
  9865. \end_inset
  9866. \end_layout
  9867. \end_inset
  9868. \end_layout
  9869. \begin_layout Plain Layout
  9870. \begin_inset Caption Standard
  9871. \begin_layout Plain Layout
  9872. \series bold
  9873. \begin_inset CommandInset label
  9874. LatexCommand label
  9875. name "fig:frmatools-batch-size"
  9876. \end_inset
  9877. Effect of batch size selection on number of batches and number of samples
  9878. included in fRMA probe weight learning.
  9879. \series default
  9880. For batch sizes ranging from 3 to 15, the number of batches (a) and samples
  9881. (b) included in probe weight training were plotted for biopsy (BX) and
  9882. blood (PAX) samples.
  9883. The selected batch size, 5, is marked with a dotted vertical line.
  9884. \end_layout
  9885. \end_inset
  9886. \end_layout
  9887. \end_inset
  9888. \end_layout
  9889. \begin_layout Standard
  9890. In order to enable use of
  9891. \begin_inset Flex Glossary Term
  9892. status open
  9893. \begin_layout Plain Layout
  9894. fRMA
  9895. \end_layout
  9896. \end_inset
  9897. to normalize hthgu133pluspm, a custom set of
  9898. \begin_inset Flex Glossary Term
  9899. status open
  9900. \begin_layout Plain Layout
  9901. fRMA
  9902. \end_layout
  9903. \end_inset
  9904. vectors was created.
  9905. First, an appropriate batch size was chosen by looking at the number of
  9906. batches and number of samples included as a function of batch size (Figure
  9907. \begin_inset CommandInset ref
  9908. LatexCommand ref
  9909. reference "fig:frmatools-batch-size"
  9910. plural "false"
  9911. caps "false"
  9912. noprefix "false"
  9913. \end_inset
  9914. ).
  9915. For a given batch size, all batches with fewer samples that the chosen
  9916. size must be ignored during training, while larger batches must be randomly
  9917. downsampled to the chosen size.
  9918. Hence, the number of samples included for a given batch size equals the
  9919. batch size times the number of batches with at least that many samples.
  9920. From Figure
  9921. \begin_inset CommandInset ref
  9922. LatexCommand ref
  9923. reference "fig:batch-size-samples"
  9924. plural "false"
  9925. caps "false"
  9926. noprefix "false"
  9927. \end_inset
  9928. , it is apparent that that a batch size of 8 maximizes the number of samples
  9929. included in training.
  9930. Increasing the batch size beyond this causes too many smaller batches to
  9931. be excluded, reducing the total number of samples for both tissue types.
  9932. However, a batch size of 8 is not necessarily optimal.
  9933. The article introducing frmaTools concluded that it was highly advantageous
  9934. to use a smaller batch size in order to include more batches, even at the
  9935. expense of including fewer total samples in training
  9936. \begin_inset CommandInset citation
  9937. LatexCommand cite
  9938. key "McCall2011"
  9939. literal "false"
  9940. \end_inset
  9941. .
  9942. To strike an appropriate balance between more batches and more samples,
  9943. a batch size of 5 was chosen.
  9944. For both blood and biopsy samples, this increased the number of batches
  9945. included by 10, with only a modest reduction in the number of samples compared
  9946. to a batch size of 8.
  9947. With a batch size of 5, 26 batches of biopsy samples and 46 batches of
  9948. blood samples were available.
  9949. \end_layout
  9950. \begin_layout Standard
  9951. \begin_inset Float figure
  9952. wide false
  9953. sideways false
  9954. status collapsed
  9955. \begin_layout Plain Layout
  9956. \begin_inset Float figure
  9957. wide false
  9958. sideways false
  9959. status open
  9960. \begin_layout Plain Layout
  9961. \align center
  9962. \begin_inset Graphics
  9963. filename graphics/frma-pax-bx/M-BX-violin.pdf
  9964. lyxscale 40
  9965. width 45col%
  9966. groupId m-violin
  9967. \end_inset
  9968. \end_layout
  9969. \begin_layout Plain Layout
  9970. \begin_inset Caption Standard
  9971. \begin_layout Plain Layout
  9972. \begin_inset CommandInset label
  9973. LatexCommand label
  9974. name "fig:m-bx-violin"
  9975. \end_inset
  9976. \series bold
  9977. Violin plot of inter-normalization log ratios for biopsy samples.
  9978. \end_layout
  9979. \end_inset
  9980. \end_layout
  9981. \end_inset
  9982. \begin_inset space \hfill{}
  9983. \end_inset
  9984. \begin_inset Float figure
  9985. wide false
  9986. sideways false
  9987. status collapsed
  9988. \begin_layout Plain Layout
  9989. \align center
  9990. \begin_inset Graphics
  9991. filename graphics/frma-pax-bx/M-PAX-violin.pdf
  9992. lyxscale 40
  9993. width 45col%
  9994. groupId m-violin
  9995. \end_inset
  9996. \end_layout
  9997. \begin_layout Plain Layout
  9998. \begin_inset Caption Standard
  9999. \begin_layout Plain Layout
  10000. \begin_inset CommandInset label
  10001. LatexCommand label
  10002. name "fig:m-pax-violin"
  10003. \end_inset
  10004. \series bold
  10005. Violin plot of inter-normalization log ratios for blood samples.
  10006. \end_layout
  10007. \end_inset
  10008. \end_layout
  10009. \end_inset
  10010. \end_layout
  10011. \begin_layout Plain Layout
  10012. \begin_inset Caption Standard
  10013. \begin_layout Plain Layout
  10014. \begin_inset CommandInset label
  10015. LatexCommand label
  10016. name "fig:frma-violin"
  10017. \end_inset
  10018. \series bold
  10019. Violin plot of log ratios between normalizations for 20 biopsy samples.
  10020. \series default
  10021. Each of 20 randomly selected samples was normalized with RMA and with 5
  10022. different sets of fRMA vectors.
  10023. The distribution of log ratios between normalized expression values, aggregated
  10024. across all 20 arrays, was plotted for each pair of normalizations.
  10025. \end_layout
  10026. \end_inset
  10027. \end_layout
  10028. \end_inset
  10029. \end_layout
  10030. \begin_layout Standard
  10031. Since
  10032. \begin_inset Flex Glossary Term
  10033. status open
  10034. \begin_layout Plain Layout
  10035. fRMA
  10036. \end_layout
  10037. \end_inset
  10038. training requires equal-size batches, larger batches are downsampled randomly.
  10039. This introduces a nondeterministic step in the generation of normalization
  10040. vectors.
  10041. To show that this randomness does not substantially change the outcome,
  10042. the random downsampling and subsequent vector learning was repeated 5 times,
  10043. with a different random seed each time.
  10044. 20 samples were selected at random as a test set and normalized with each
  10045. of the 5 sets of
  10046. \begin_inset Flex Glossary Term
  10047. status open
  10048. \begin_layout Plain Layout
  10049. fRMA
  10050. \end_layout
  10051. \end_inset
  10052. normalization vectors as well as ordinary RMA, and the normalized expression
  10053. values were compared across normalizations.
  10054. Figure
  10055. \begin_inset CommandInset ref
  10056. LatexCommand ref
  10057. reference "fig:m-bx-violin"
  10058. plural "false"
  10059. caps "false"
  10060. noprefix "false"
  10061. \end_inset
  10062. shows a summary of these comparisons for biopsy samples.
  10063. Comparing RMA to each of the 5
  10064. \begin_inset Flex Glossary Term
  10065. status open
  10066. \begin_layout Plain Layout
  10067. fRMA
  10068. \end_layout
  10069. \end_inset
  10070. normalizations, the distribution of log ratios is somewhat wide, indicating
  10071. that the normalizations disagree on the expression values of a fair number
  10072. of probe sets.
  10073. In contrast, comparisons of
  10074. \begin_inset Flex Glossary Term
  10075. status open
  10076. \begin_layout Plain Layout
  10077. fRMA
  10078. \end_layout
  10079. \end_inset
  10080. against
  10081. \begin_inset Flex Glossary Term
  10082. status open
  10083. \begin_layout Plain Layout
  10084. fRMA
  10085. \end_layout
  10086. \end_inset
  10087. , the vast majority of probe sets have very small log ratios, indicating
  10088. a very high agreement between the normalized values generated by the two
  10089. normalizations.
  10090. This shows that the
  10091. \begin_inset Flex Glossary Term
  10092. status open
  10093. \begin_layout Plain Layout
  10094. fRMA
  10095. \end_layout
  10096. \end_inset
  10097. normalization's behavior is not very sensitive to the random downsampling
  10098. of larger batches during training.
  10099. \end_layout
  10100. \begin_layout Standard
  10101. \begin_inset Float figure
  10102. wide false
  10103. sideways false
  10104. status open
  10105. \begin_layout Plain Layout
  10106. \align center
  10107. \begin_inset Float figure
  10108. wide false
  10109. sideways false
  10110. status collapsed
  10111. \begin_layout Plain Layout
  10112. \align center
  10113. \begin_inset Graphics
  10114. filename graphics/frma-pax-bx/MA-BX-RMA.fRMA-RASTER.png
  10115. lyxscale 10
  10116. width 45col%
  10117. groupId ma-frma
  10118. \end_inset
  10119. \end_layout
  10120. \begin_layout Plain Layout
  10121. \begin_inset Caption Standard
  10122. \begin_layout Plain Layout
  10123. \begin_inset CommandInset label
  10124. LatexCommand label
  10125. name "fig:ma-bx-rma-frma"
  10126. \end_inset
  10127. RMA vs.
  10128. fRMA for biopsy samples.
  10129. \end_layout
  10130. \end_inset
  10131. \end_layout
  10132. \end_inset
  10133. \begin_inset space \hfill{}
  10134. \end_inset
  10135. \begin_inset Float figure
  10136. wide false
  10137. sideways false
  10138. status collapsed
  10139. \begin_layout Plain Layout
  10140. \align center
  10141. \begin_inset Graphics
  10142. filename graphics/frma-pax-bx/MA-BX-fRMA.fRMA-RASTER.png
  10143. lyxscale 10
  10144. width 45col%
  10145. groupId ma-frma
  10146. \end_inset
  10147. \end_layout
  10148. \begin_layout Plain Layout
  10149. \begin_inset Caption Standard
  10150. \begin_layout Plain Layout
  10151. \begin_inset CommandInset label
  10152. LatexCommand label
  10153. name "fig:ma-bx-frma-frma"
  10154. \end_inset
  10155. fRMA vs fRMA for biopsy samples.
  10156. \end_layout
  10157. \end_inset
  10158. \end_layout
  10159. \end_inset
  10160. \end_layout
  10161. \begin_layout Plain Layout
  10162. \align center
  10163. \begin_inset Float figure
  10164. wide false
  10165. sideways false
  10166. status collapsed
  10167. \begin_layout Plain Layout
  10168. \align center
  10169. \begin_inset Graphics
  10170. filename graphics/frma-pax-bx/MA-PAX-RMA.fRMA-RASTER.png
  10171. lyxscale 10
  10172. width 45col%
  10173. groupId ma-frma
  10174. \end_inset
  10175. \end_layout
  10176. \begin_layout Plain Layout
  10177. \begin_inset Caption Standard
  10178. \begin_layout Plain Layout
  10179. \begin_inset CommandInset label
  10180. LatexCommand label
  10181. name "fig:MA-PAX-rma-frma"
  10182. \end_inset
  10183. RMA vs.
  10184. fRMA for blood samples.
  10185. \end_layout
  10186. \end_inset
  10187. \end_layout
  10188. \end_inset
  10189. \begin_inset space \hfill{}
  10190. \end_inset
  10191. \begin_inset Float figure
  10192. wide false
  10193. sideways false
  10194. status collapsed
  10195. \begin_layout Plain Layout
  10196. \align center
  10197. \begin_inset Graphics
  10198. filename graphics/frma-pax-bx/MA-PAX-fRMA.fRMA-RASTER.png
  10199. lyxscale 10
  10200. width 45col%
  10201. groupId ma-frma
  10202. \end_inset
  10203. \end_layout
  10204. \begin_layout Plain Layout
  10205. \begin_inset Caption Standard
  10206. \begin_layout Plain Layout
  10207. \begin_inset CommandInset label
  10208. LatexCommand label
  10209. name "fig:MA-PAX-frma-frma"
  10210. \end_inset
  10211. fRMA vs fRMA for blood samples.
  10212. \end_layout
  10213. \end_inset
  10214. \end_layout
  10215. \end_inset
  10216. \end_layout
  10217. \begin_layout Plain Layout
  10218. \begin_inset Caption Standard
  10219. \begin_layout Plain Layout
  10220. \series bold
  10221. \begin_inset CommandInset label
  10222. LatexCommand label
  10223. name "fig:Representative-MA-plots"
  10224. \end_inset
  10225. Representative MA plots comparing RMA and custom fRMA normalizations.
  10226. \series default
  10227. For each plot, 20 samples were normalized using 2 different normalizations,
  10228. and then averages (A) and log ratios (M) were plotted between the two different
  10229. normalizations for every probe.
  10230. For the
  10231. \begin_inset Quotes eld
  10232. \end_inset
  10233. fRMA vs fRMA
  10234. \begin_inset Quotes erd
  10235. \end_inset
  10236. plots (b & d), two different fRMA normalizations using vectors from two
  10237. independent batch samplings were compared.
  10238. Density of points is represented by blue shading, and individual outlier
  10239. points are plotted.
  10240. \end_layout
  10241. \end_inset
  10242. \end_layout
  10243. \end_inset
  10244. \end_layout
  10245. \begin_layout Standard
  10246. Figure
  10247. \begin_inset CommandInset ref
  10248. LatexCommand ref
  10249. reference "fig:ma-bx-rma-frma"
  10250. plural "false"
  10251. caps "false"
  10252. noprefix "false"
  10253. \end_inset
  10254. shows an MA plot of the RMA-normalized values against the fRMA-normalized
  10255. values for the same probe sets and arrays, corresponding to the first row
  10256. of Figure
  10257. \begin_inset CommandInset ref
  10258. LatexCommand ref
  10259. reference "fig:m-bx-violin"
  10260. plural "false"
  10261. caps "false"
  10262. noprefix "false"
  10263. \end_inset
  10264. .
  10265. This MA plot shows that not only is there a wide distribution of M-values,
  10266. but the trend of M-values is dependent on the average normalized intensity.
  10267. This is expected, since the overall trend represents the differences in
  10268. the quantile normalization step.
  10269. When running
  10270. \begin_inset Flex Glossary Term
  10271. status open
  10272. \begin_layout Plain Layout
  10273. RMA
  10274. \end_layout
  10275. \end_inset
  10276. , only the quantiles for these specific 20 arrays are used, while for
  10277. \begin_inset Flex Glossary Term
  10278. status open
  10279. \begin_layout Plain Layout
  10280. fRMA
  10281. \end_layout
  10282. \end_inset
  10283. the quantile distribution is taking from all arrays used in training.
  10284. Figure
  10285. \begin_inset CommandInset ref
  10286. LatexCommand ref
  10287. reference "fig:ma-bx-frma-frma"
  10288. plural "false"
  10289. caps "false"
  10290. noprefix "false"
  10291. \end_inset
  10292. shows a similar MA plot comparing 2 different
  10293. \begin_inset Flex Glossary Term
  10294. status open
  10295. \begin_layout Plain Layout
  10296. fRMA
  10297. \end_layout
  10298. \end_inset
  10299. normalizations, corresponding to the 6th row of Figure
  10300. \begin_inset CommandInset ref
  10301. LatexCommand ref
  10302. reference "fig:m-bx-violin"
  10303. plural "false"
  10304. caps "false"
  10305. noprefix "false"
  10306. \end_inset
  10307. .
  10308. The MA plot is very tightly centered around zero with no visible trend.
  10309. Figures
  10310. \begin_inset CommandInset ref
  10311. LatexCommand ref
  10312. reference "fig:m-pax-violin"
  10313. plural "false"
  10314. caps "false"
  10315. noprefix "false"
  10316. \end_inset
  10317. ,
  10318. \begin_inset CommandInset ref
  10319. LatexCommand ref
  10320. reference "fig:MA-PAX-rma-frma"
  10321. plural "false"
  10322. caps "false"
  10323. noprefix "false"
  10324. \end_inset
  10325. , and
  10326. \begin_inset CommandInset ref
  10327. LatexCommand ref
  10328. reference "fig:ma-bx-frma-frma"
  10329. plural "false"
  10330. caps "false"
  10331. noprefix "false"
  10332. \end_inset
  10333. show exactly the same information for the blood samples, once again comparing
  10334. the normalized expression values between normalizations for all probe sets
  10335. across 20 randomly selected test arrays.
  10336. Once again, there is a wider distribution of log ratios between RMA-normalized
  10337. values and fRMA-normalized, and a much tighter distribution when comparing
  10338. different
  10339. \begin_inset Flex Glossary Term
  10340. status open
  10341. \begin_layout Plain Layout
  10342. fRMA
  10343. \end_layout
  10344. \end_inset
  10345. normalizations to each other, indicating that the
  10346. \begin_inset Flex Glossary Term
  10347. status open
  10348. \begin_layout Plain Layout
  10349. fRMA
  10350. \end_layout
  10351. \end_inset
  10352. training process is robust to random batch downsampling for the blood samples
  10353. as well.
  10354. \end_layout
  10355. \begin_layout Subsection
  10356. SVA, voom, and array weights improve model fit for methylation array data
  10357. \end_layout
  10358. \begin_layout Standard
  10359. \begin_inset ERT
  10360. status open
  10361. \begin_layout Plain Layout
  10362. \backslash
  10363. afterpage{
  10364. \end_layout
  10365. \begin_layout Plain Layout
  10366. \backslash
  10367. begin{landscape}
  10368. \end_layout
  10369. \end_inset
  10370. \end_layout
  10371. \begin_layout Standard
  10372. \begin_inset Float figure
  10373. wide false
  10374. sideways false
  10375. status open
  10376. \begin_layout Plain Layout
  10377. \begin_inset Flex TODO Note (inline)
  10378. status open
  10379. \begin_layout Plain Layout
  10380. Fix axis labels:
  10381. \begin_inset Quotes eld
  10382. \end_inset
  10383. log2 M-value
  10384. \begin_inset Quotes erd
  10385. \end_inset
  10386. is redundant because M-values are already log scale
  10387. \end_layout
  10388. \end_inset
  10389. \end_layout
  10390. \begin_layout Plain Layout
  10391. \begin_inset Float figure
  10392. wide false
  10393. sideways false
  10394. status collapsed
  10395. \begin_layout Plain Layout
  10396. \align center
  10397. \begin_inset Graphics
  10398. filename graphics/methylvoom/unadj.dupcor/meanvar-trends-PAGE1-CROP-RASTER.png
  10399. lyxscale 15
  10400. width 30col%
  10401. groupId voomaw-subfig
  10402. \end_inset
  10403. \end_layout
  10404. \begin_layout Plain Layout
  10405. \begin_inset Caption Standard
  10406. \begin_layout Plain Layout
  10407. \begin_inset CommandInset label
  10408. LatexCommand label
  10409. name "fig:meanvar-basic"
  10410. \end_inset
  10411. Mean-variance trend for analysis A.
  10412. \end_layout
  10413. \end_inset
  10414. \end_layout
  10415. \end_inset
  10416. \begin_inset space \hfill{}
  10417. \end_inset
  10418. \begin_inset Float figure
  10419. wide false
  10420. sideways false
  10421. status collapsed
  10422. \begin_layout Plain Layout
  10423. \align center
  10424. \begin_inset Graphics
  10425. filename graphics/methylvoom/unadj.dupcor.sva.aw/meanvar-trends-PAGE1-CROP-RASTER.png
  10426. lyxscale 15
  10427. width 30col%
  10428. groupId voomaw-subfig
  10429. \end_inset
  10430. \end_layout
  10431. \begin_layout Plain Layout
  10432. \begin_inset Caption Standard
  10433. \begin_layout Plain Layout
  10434. \begin_inset CommandInset label
  10435. LatexCommand label
  10436. name "fig:meanvar-sva-aw"
  10437. \end_inset
  10438. Mean-variance trend for analysis B.
  10439. \end_layout
  10440. \end_inset
  10441. \end_layout
  10442. \end_inset
  10443. \begin_inset space \hfill{}
  10444. \end_inset
  10445. \begin_inset Float figure
  10446. wide false
  10447. sideways false
  10448. status collapsed
  10449. \begin_layout Plain Layout
  10450. \align center
  10451. \begin_inset Graphics
  10452. filename graphics/methylvoom/unadj.dupcor.sva.voomaw/meanvar-trends-PAGE2-CROP-RASTER.png
  10453. lyxscale 15
  10454. width 30col%
  10455. groupId voomaw-subfig
  10456. \end_inset
  10457. \end_layout
  10458. \begin_layout Plain Layout
  10459. \begin_inset Caption Standard
  10460. \begin_layout Plain Layout
  10461. \begin_inset CommandInset label
  10462. LatexCommand label
  10463. name "fig:meanvar-sva-voomaw"
  10464. \end_inset
  10465. Mean-variance trend after voom modeling in analysis C.
  10466. \end_layout
  10467. \end_inset
  10468. \end_layout
  10469. \end_inset
  10470. \end_layout
  10471. \begin_layout Plain Layout
  10472. \begin_inset Caption Standard
  10473. \begin_layout Plain Layout
  10474. \series bold
  10475. Mean-variance trend modeling in methylation array data.
  10476. \series default
  10477. The estimated
  10478. \begin_inset Formula $\log_{2}$
  10479. \end_inset
  10480. (standard deviation) for each probe is plotted against the probe's average
  10481. M-value across all samples as a black point, with some transparency to
  10482. make over-plotting more visible, since there are about 450,000 points.
  10483. Density of points is also indicated by the dark blue contour lines.
  10484. The prior variance trend estimated by eBayes is shown in light blue, while
  10485. the lowess trend of the points is shown in red.
  10486. \end_layout
  10487. \end_inset
  10488. \end_layout
  10489. \end_inset
  10490. \end_layout
  10491. \begin_layout Standard
  10492. \begin_inset ERT
  10493. status open
  10494. \begin_layout Plain Layout
  10495. \backslash
  10496. end{landscape}
  10497. \end_layout
  10498. \begin_layout Plain Layout
  10499. }
  10500. \end_layout
  10501. \end_inset
  10502. \end_layout
  10503. \begin_layout Standard
  10504. Figure
  10505. \begin_inset CommandInset ref
  10506. LatexCommand ref
  10507. reference "fig:meanvar-basic"
  10508. plural "false"
  10509. caps "false"
  10510. noprefix "false"
  10511. \end_inset
  10512. shows the relationship between the mean M-value and the standard deviation
  10513. calculated for each probe in the methylation array data set.
  10514. A few features of the data are apparent.
  10515. First, the data are very strongly bimodal, with peaks in the density around
  10516. M-values of +4 and -4.
  10517. These modes correspond to methylation sites that are nearly 100% methylated
  10518. and nearly 100% unmethylated, respectively.
  10519. The strong bimodality indicates that a majority of probes interrogate sites
  10520. that fall into one of these two categories.
  10521. The points in between these modes represent sites that are either partially
  10522. methylated in many samples, or are fully methylated in some samples and
  10523. fully unmethylated in other samples, or some combination.
  10524. The next visible feature of the data is the W-shaped variance trend.
  10525. The upticks in the variance trend on either side are expected, based on
  10526. the sigmoid transformation exaggerating small differences at extreme M-values
  10527. (Figure
  10528. \begin_inset CommandInset ref
  10529. LatexCommand ref
  10530. reference "fig:Sigmoid-beta-m-mapping"
  10531. plural "false"
  10532. caps "false"
  10533. noprefix "false"
  10534. \end_inset
  10535. ).
  10536. However, the uptick in the center is interesting: it indicates that sites
  10537. that are not constitutively methylated or unmethylated have a higher variance.
  10538. This could be a genuine biological effect, or it could be spurious noise
  10539. that is only observable at sites with varying methylation.
  10540. \end_layout
  10541. \begin_layout Standard
  10542. In Figure
  10543. \begin_inset CommandInset ref
  10544. LatexCommand ref
  10545. reference "fig:meanvar-sva-aw"
  10546. plural "false"
  10547. caps "false"
  10548. noprefix "false"
  10549. \end_inset
  10550. , we see the mean-variance trend for the same methylation array data, this
  10551. time with surrogate variables and sample quality weights estimated from
  10552. the data and included in the model.
  10553. As expected, the overall average variance is smaller, since the surrogate
  10554. variables account for some of the variance.
  10555. In addition, the uptick in variance in the middle of the M-value range
  10556. has disappeared, turning the W shape into a wide U shape.
  10557. This indicates that the excess variance in the probes with intermediate
  10558. M-values was explained by systematic variations not correlated with known
  10559. covariates, and these variations were modeled by the surrogate variables.
  10560. The result is a nearly flat variance trend for the entire intermediate
  10561. M-value range from about -3 to +3.
  10562. Note that this corresponds closely to the range within which the M-value
  10563. transformation shown in Figure
  10564. \begin_inset CommandInset ref
  10565. LatexCommand ref
  10566. reference "fig:Sigmoid-beta-m-mapping"
  10567. plural "false"
  10568. caps "false"
  10569. noprefix "false"
  10570. \end_inset
  10571. is nearly linear.
  10572. In contrast, the excess variance at the extremes (greater than +3 and less
  10573. than -3) was not
  10574. \begin_inset Quotes eld
  10575. \end_inset
  10576. absorbed
  10577. \begin_inset Quotes erd
  10578. \end_inset
  10579. by the surrogate variables and remains in the plot, indicating that this
  10580. variation has no systematic component: probes with extreme M-values are
  10581. uniformly more variable across all samples, as expected.
  10582. \end_layout
  10583. \begin_layout Standard
  10584. Figure
  10585. \begin_inset CommandInset ref
  10586. LatexCommand ref
  10587. reference "fig:meanvar-sva-voomaw"
  10588. plural "false"
  10589. caps "false"
  10590. noprefix "false"
  10591. \end_inset
  10592. shows the mean-variance trend after fitting the model with the observation
  10593. weights assigned by voom based on the mean-variance trend shown in Figure
  10594. \begin_inset CommandInset ref
  10595. LatexCommand ref
  10596. reference "fig:meanvar-sva-aw"
  10597. plural "false"
  10598. caps "false"
  10599. noprefix "false"
  10600. \end_inset
  10601. .
  10602. As expected, the weights exactly counteract the trend in the data, resulting
  10603. in a nearly flat trend centered vertically at 1 (i.e.
  10604. 0 on the log scale).
  10605. This shows that the observations with extreme M-values have been appropriately
  10606. down-weighted to account for the fact that the noise in those observations
  10607. has been amplified by the non-linear M-value transformation.
  10608. In turn, this gives relatively more weight to observations in the middle
  10609. region, which are more likely to correspond to probes measuring interesting
  10610. biology (not constitutively methylated or unmethylated).
  10611. \end_layout
  10612. \begin_layout Standard
  10613. \begin_inset Float table
  10614. wide false
  10615. sideways false
  10616. status open
  10617. \begin_layout Plain Layout
  10618. \align center
  10619. \begin_inset Tabular
  10620. <lyxtabular version="3" rows="5" columns="3">
  10621. <features tabularvalignment="middle">
  10622. <column alignment="center" valignment="top">
  10623. <column alignment="center" valignment="top">
  10624. <column alignment="center" valignment="top">
  10625. <row>
  10626. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  10627. \begin_inset Text
  10628. \begin_layout Plain Layout
  10629. Covariate
  10630. \end_layout
  10631. \end_inset
  10632. </cell>
  10633. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  10634. \begin_inset Text
  10635. \begin_layout Plain Layout
  10636. Test used
  10637. \end_layout
  10638. \end_inset
  10639. </cell>
  10640. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  10641. \begin_inset Text
  10642. \begin_layout Plain Layout
  10643. p-value
  10644. \end_layout
  10645. \end_inset
  10646. </cell>
  10647. </row>
  10648. <row>
  10649. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  10650. \begin_inset Text
  10651. \begin_layout Plain Layout
  10652. Transplant Status
  10653. \end_layout
  10654. \end_inset
  10655. </cell>
  10656. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  10657. \begin_inset Text
  10658. \begin_layout Plain Layout
  10659. F-test
  10660. \end_layout
  10661. \end_inset
  10662. </cell>
  10663. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  10664. \begin_inset Text
  10665. \begin_layout Plain Layout
  10666. 0.404
  10667. \end_layout
  10668. \end_inset
  10669. </cell>
  10670. </row>
  10671. <row>
  10672. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  10673. \begin_inset Text
  10674. \begin_layout Plain Layout
  10675. Diabetes Diagnosis
  10676. \end_layout
  10677. \end_inset
  10678. </cell>
  10679. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  10680. \begin_inset Text
  10681. \begin_layout Plain Layout
  10682. \emph on
  10683. t
  10684. \emph default
  10685. -test
  10686. \end_layout
  10687. \end_inset
  10688. </cell>
  10689. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  10690. \begin_inset Text
  10691. \begin_layout Plain Layout
  10692. 0.00106
  10693. \end_layout
  10694. \end_inset
  10695. </cell>
  10696. </row>
  10697. <row>
  10698. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  10699. \begin_inset Text
  10700. \begin_layout Plain Layout
  10701. Sex
  10702. \end_layout
  10703. \end_inset
  10704. </cell>
  10705. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  10706. \begin_inset Text
  10707. \begin_layout Plain Layout
  10708. \emph on
  10709. t
  10710. \emph default
  10711. -test
  10712. \end_layout
  10713. \end_inset
  10714. </cell>
  10715. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  10716. \begin_inset Text
  10717. \begin_layout Plain Layout
  10718. 0.148
  10719. \end_layout
  10720. \end_inset
  10721. </cell>
  10722. </row>
  10723. <row>
  10724. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  10725. \begin_inset Text
  10726. \begin_layout Plain Layout
  10727. Age
  10728. \end_layout
  10729. \end_inset
  10730. </cell>
  10731. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  10732. \begin_inset Text
  10733. \begin_layout Plain Layout
  10734. linear regression
  10735. \end_layout
  10736. \end_inset
  10737. </cell>
  10738. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  10739. \begin_inset Text
  10740. \begin_layout Plain Layout
  10741. 0.212
  10742. \end_layout
  10743. \end_inset
  10744. </cell>
  10745. </row>
  10746. </lyxtabular>
  10747. \end_inset
  10748. \end_layout
  10749. \begin_layout Plain Layout
  10750. \begin_inset Caption Standard
  10751. \begin_layout Plain Layout
  10752. \series bold
  10753. \begin_inset CommandInset label
  10754. LatexCommand label
  10755. name "tab:weight-covariate-tests"
  10756. \end_inset
  10757. Association of sample weights with clinical covariates in methylation array
  10758. data.
  10759. \series default
  10760. Computed sample quality log weights were tested for significant association
  10761. with each of the variables in the model (1st column).
  10762. An appropriate test was selected for each variable based on whether the
  10763. variable had 2 categories (
  10764. \emph on
  10765. t
  10766. \emph default
  10767. -test), had more than 2 categories (F-test), or was numeric (linear regression).
  10768. The test selected is shown in the 2nd column.
  10769. P-values for association with the log weights are shown in the 3rd column.
  10770. No multiple testing adjustment was performed for these p-values.
  10771. \end_layout
  10772. \end_inset
  10773. \end_layout
  10774. \end_inset
  10775. \end_layout
  10776. \begin_layout Standard
  10777. \begin_inset Float figure
  10778. wide false
  10779. sideways false
  10780. status open
  10781. \begin_layout Plain Layout
  10782. \begin_inset Flex TODO Note (inline)
  10783. status open
  10784. \begin_layout Plain Layout
  10785. Redo the sample weight boxplot with notches, and remove fill colors
  10786. \end_layout
  10787. \end_inset
  10788. \end_layout
  10789. \begin_layout Plain Layout
  10790. \align center
  10791. \begin_inset Graphics
  10792. filename graphics/methylvoom/unadj.dupcor.sva.voomaw/sample-weights-PAGE3-CROP.pdf
  10793. lyxscale 50
  10794. width 60col%
  10795. groupId colwidth
  10796. \end_inset
  10797. \end_layout
  10798. \begin_layout Plain Layout
  10799. \begin_inset Caption Standard
  10800. \begin_layout Plain Layout
  10801. \begin_inset CommandInset label
  10802. LatexCommand label
  10803. name "fig:diabetes-sample-weights"
  10804. \end_inset
  10805. \series bold
  10806. Box-and-whiskers plot of sample quality weights grouped by diabetes diagnosis.
  10807. \series default
  10808. Samples were grouped based on diabetes diagnosis, and the distribution of
  10809. sample quality weights for each diagnosis was plotted as a box-and-whiskers
  10810. plot
  10811. \begin_inset CommandInset citation
  10812. LatexCommand cite
  10813. key "McGill1978"
  10814. literal "false"
  10815. \end_inset
  10816. .
  10817. \end_layout
  10818. \end_inset
  10819. \end_layout
  10820. \begin_layout Plain Layout
  10821. \end_layout
  10822. \end_inset
  10823. \end_layout
  10824. \begin_layout Standard
  10825. To determine whether any of the known experimental factors had an impact
  10826. on data quality, the sample quality weights estimated from the data were
  10827. tested for association with each of the experimental factors (Table
  10828. \begin_inset CommandInset ref
  10829. LatexCommand ref
  10830. reference "tab:weight-covariate-tests"
  10831. plural "false"
  10832. caps "false"
  10833. noprefix "false"
  10834. \end_inset
  10835. ).
  10836. Diabetes diagnosis was found to have a potentially significant association
  10837. with the sample weights, with a t-test p-value of
  10838. \begin_inset Formula $1.06\times10^{-3}$
  10839. \end_inset
  10840. .
  10841. Figure
  10842. \begin_inset CommandInset ref
  10843. LatexCommand ref
  10844. reference "fig:diabetes-sample-weights"
  10845. plural "false"
  10846. caps "false"
  10847. noprefix "false"
  10848. \end_inset
  10849. shows the distribution of sample weights grouped by diabetes diagnosis.
  10850. The samples from patients with
  10851. \begin_inset Flex Glossary Term
  10852. status open
  10853. \begin_layout Plain Layout
  10854. T2D
  10855. \end_layout
  10856. \end_inset
  10857. were assigned significantly lower weights than those from patients with
  10858. \begin_inset Flex Glossary Term
  10859. status open
  10860. \begin_layout Plain Layout
  10861. T1D
  10862. \end_layout
  10863. \end_inset
  10864. .
  10865. This indicates that the
  10866. \begin_inset Flex Glossary Term
  10867. status open
  10868. \begin_layout Plain Layout
  10869. T2D
  10870. \end_layout
  10871. \end_inset
  10872. samples had an overall higher variance on average across all probes.
  10873. \end_layout
  10874. \begin_layout Standard
  10875. \begin_inset Float table
  10876. wide false
  10877. sideways false
  10878. status open
  10879. \begin_layout Plain Layout
  10880. \align center
  10881. \begin_inset Flex TODO Note (inline)
  10882. status open
  10883. \begin_layout Plain Layout
  10884. Consider transposing these tables
  10885. \end_layout
  10886. \end_inset
  10887. \end_layout
  10888. \begin_layout Plain Layout
  10889. \begin_inset Float table
  10890. wide false
  10891. sideways false
  10892. status open
  10893. \begin_layout Plain Layout
  10894. \align center
  10895. \begin_inset Tabular
  10896. <lyxtabular version="3" rows="5" columns="4">
  10897. <features tabularvalignment="middle">
  10898. <column alignment="center" valignment="top">
  10899. <column alignment="center" valignment="top">
  10900. <column alignment="center" valignment="top">
  10901. <column alignment="center" valignment="top">
  10902. <row>
  10903. <cell alignment="center" valignment="top" usebox="none">
  10904. \begin_inset Text
  10905. \begin_layout Plain Layout
  10906. \end_layout
  10907. \end_inset
  10908. </cell>
  10909. <cell multicolumn="1" alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  10910. \begin_inset Text
  10911. \begin_layout Plain Layout
  10912. Analysis
  10913. \end_layout
  10914. \end_inset
  10915. </cell>
  10916. <cell multicolumn="2" alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  10917. \begin_inset Text
  10918. \begin_layout Plain Layout
  10919. \end_layout
  10920. \end_inset
  10921. </cell>
  10922. <cell multicolumn="2" alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  10923. \begin_inset Text
  10924. \begin_layout Plain Layout
  10925. \end_layout
  10926. \end_inset
  10927. </cell>
  10928. </row>
  10929. <row>
  10930. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  10931. \begin_inset Text
  10932. \begin_layout Plain Layout
  10933. Contrast
  10934. \end_layout
  10935. \end_inset
  10936. </cell>
  10937. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  10938. \begin_inset Text
  10939. \begin_layout Plain Layout
  10940. A
  10941. \end_layout
  10942. \end_inset
  10943. </cell>
  10944. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  10945. \begin_inset Text
  10946. \begin_layout Plain Layout
  10947. B
  10948. \end_layout
  10949. \end_inset
  10950. </cell>
  10951. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  10952. \begin_inset Text
  10953. \begin_layout Plain Layout
  10954. C
  10955. \end_layout
  10956. \end_inset
  10957. </cell>
  10958. </row>
  10959. <row>
  10960. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  10961. \begin_inset Text
  10962. \begin_layout Plain Layout
  10963. TX vs AR
  10964. \end_layout
  10965. \end_inset
  10966. </cell>
  10967. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  10968. \begin_inset Text
  10969. \begin_layout Plain Layout
  10970. 0
  10971. \end_layout
  10972. \end_inset
  10973. </cell>
  10974. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  10975. \begin_inset Text
  10976. \begin_layout Plain Layout
  10977. 25
  10978. \end_layout
  10979. \end_inset
  10980. </cell>
  10981. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  10982. \begin_inset Text
  10983. \begin_layout Plain Layout
  10984. 22
  10985. \end_layout
  10986. \end_inset
  10987. </cell>
  10988. </row>
  10989. <row>
  10990. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  10991. \begin_inset Text
  10992. \begin_layout Plain Layout
  10993. TX vs ADNR
  10994. \end_layout
  10995. \end_inset
  10996. </cell>
  10997. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  10998. \begin_inset Text
  10999. \begin_layout Plain Layout
  11000. 7
  11001. \end_layout
  11002. \end_inset
  11003. </cell>
  11004. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  11005. \begin_inset Text
  11006. \begin_layout Plain Layout
  11007. 338
  11008. \end_layout
  11009. \end_inset
  11010. </cell>
  11011. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  11012. \begin_inset Text
  11013. \begin_layout Plain Layout
  11014. 369
  11015. \end_layout
  11016. \end_inset
  11017. </cell>
  11018. </row>
  11019. <row>
  11020. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  11021. \begin_inset Text
  11022. \begin_layout Plain Layout
  11023. TX vs CAN
  11024. \end_layout
  11025. \end_inset
  11026. </cell>
  11027. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  11028. \begin_inset Text
  11029. \begin_layout Plain Layout
  11030. 0
  11031. \end_layout
  11032. \end_inset
  11033. </cell>
  11034. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  11035. \begin_inset Text
  11036. \begin_layout Plain Layout
  11037. 231
  11038. \end_layout
  11039. \end_inset
  11040. </cell>
  11041. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  11042. \begin_inset Text
  11043. \begin_layout Plain Layout
  11044. 278
  11045. \end_layout
  11046. \end_inset
  11047. </cell>
  11048. </row>
  11049. </lyxtabular>
  11050. \end_inset
  11051. \end_layout
  11052. \begin_layout Plain Layout
  11053. \begin_inset Caption Standard
  11054. \begin_layout Plain Layout
  11055. \begin_inset CommandInset label
  11056. LatexCommand label
  11057. name "tab:methyl-num-signif"
  11058. \end_inset
  11059. Number of probes significant at 10% FDR.
  11060. \end_layout
  11061. \end_inset
  11062. \end_layout
  11063. \end_inset
  11064. \begin_inset space \hfill{}
  11065. \end_inset
  11066. \begin_inset Float table
  11067. wide false
  11068. sideways false
  11069. status open
  11070. \begin_layout Plain Layout
  11071. \align center
  11072. \begin_inset Tabular
  11073. <lyxtabular version="3" rows="5" columns="4">
  11074. <features tabularvalignment="middle">
  11075. <column alignment="center" valignment="top">
  11076. <column alignment="center" valignment="top">
  11077. <column alignment="center" valignment="top">
  11078. <column alignment="center" valignment="top">
  11079. <row>
  11080. <cell alignment="center" valignment="top" usebox="none">
  11081. \begin_inset Text
  11082. \begin_layout Plain Layout
  11083. \end_layout
  11084. \end_inset
  11085. </cell>
  11086. <cell multicolumn="1" alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  11087. \begin_inset Text
  11088. \begin_layout Plain Layout
  11089. Analysis
  11090. \end_layout
  11091. \end_inset
  11092. </cell>
  11093. <cell multicolumn="2" alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  11094. \begin_inset Text
  11095. \begin_layout Plain Layout
  11096. \end_layout
  11097. \end_inset
  11098. </cell>
  11099. <cell multicolumn="2" alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  11100. \begin_inset Text
  11101. \begin_layout Plain Layout
  11102. \end_layout
  11103. \end_inset
  11104. </cell>
  11105. </row>
  11106. <row>
  11107. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  11108. \begin_inset Text
  11109. \begin_layout Plain Layout
  11110. Contrast
  11111. \end_layout
  11112. \end_inset
  11113. </cell>
  11114. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  11115. \begin_inset Text
  11116. \begin_layout Plain Layout
  11117. A
  11118. \end_layout
  11119. \end_inset
  11120. </cell>
  11121. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  11122. \begin_inset Text
  11123. \begin_layout Plain Layout
  11124. B
  11125. \end_layout
  11126. \end_inset
  11127. </cell>
  11128. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  11129. \begin_inset Text
  11130. \begin_layout Plain Layout
  11131. C
  11132. \end_layout
  11133. \end_inset
  11134. </cell>
  11135. </row>
  11136. <row>
  11137. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  11138. \begin_inset Text
  11139. \begin_layout Plain Layout
  11140. TX vs AR
  11141. \end_layout
  11142. \end_inset
  11143. </cell>
  11144. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  11145. \begin_inset Text
  11146. \begin_layout Plain Layout
  11147. 0
  11148. \end_layout
  11149. \end_inset
  11150. </cell>
  11151. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  11152. \begin_inset Text
  11153. \begin_layout Plain Layout
  11154. 10,063
  11155. \end_layout
  11156. \end_inset
  11157. </cell>
  11158. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  11159. \begin_inset Text
  11160. \begin_layout Plain Layout
  11161. 11,225
  11162. \end_layout
  11163. \end_inset
  11164. </cell>
  11165. </row>
  11166. <row>
  11167. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  11168. \begin_inset Text
  11169. \begin_layout Plain Layout
  11170. TX vs ADNR
  11171. \end_layout
  11172. \end_inset
  11173. </cell>
  11174. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  11175. \begin_inset Text
  11176. \begin_layout Plain Layout
  11177. 27
  11178. \end_layout
  11179. \end_inset
  11180. </cell>
  11181. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  11182. \begin_inset Text
  11183. \begin_layout Plain Layout
  11184. 12,674
  11185. \end_layout
  11186. \end_inset
  11187. </cell>
  11188. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  11189. \begin_inset Text
  11190. \begin_layout Plain Layout
  11191. 13,086
  11192. \end_layout
  11193. \end_inset
  11194. </cell>
  11195. </row>
  11196. <row>
  11197. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  11198. \begin_inset Text
  11199. \begin_layout Plain Layout
  11200. TX vs CAN
  11201. \end_layout
  11202. \end_inset
  11203. </cell>
  11204. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  11205. \begin_inset Text
  11206. \begin_layout Plain Layout
  11207. 966
  11208. \end_layout
  11209. \end_inset
  11210. </cell>
  11211. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  11212. \begin_inset Text
  11213. \begin_layout Plain Layout
  11214. 20,039
  11215. \end_layout
  11216. \end_inset
  11217. </cell>
  11218. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  11219. \begin_inset Text
  11220. \begin_layout Plain Layout
  11221. 20,955
  11222. \end_layout
  11223. \end_inset
  11224. </cell>
  11225. </row>
  11226. </lyxtabular>
  11227. \end_inset
  11228. \end_layout
  11229. \begin_layout Plain Layout
  11230. \begin_inset Caption Standard
  11231. \begin_layout Plain Layout
  11232. \begin_inset CommandInset label
  11233. LatexCommand label
  11234. name "tab:methyl-est-nonnull"
  11235. \end_inset
  11236. Estimated number of non-null tests, using the method of averaging local
  11237. FDR values
  11238. \begin_inset CommandInset citation
  11239. LatexCommand cite
  11240. key "Phipson2013Thesis"
  11241. literal "false"
  11242. \end_inset
  11243. .
  11244. \end_layout
  11245. \end_inset
  11246. \end_layout
  11247. \end_inset
  11248. \end_layout
  11249. \begin_layout Plain Layout
  11250. \begin_inset Caption Standard
  11251. \begin_layout Plain Layout
  11252. \series bold
  11253. Estimates of degree of differential methylation in for each contrast in
  11254. each analysis.
  11255. \series default
  11256. For each of the analyses in Table
  11257. \begin_inset CommandInset ref
  11258. LatexCommand ref
  11259. reference "tab:Summary-of-meth-analysis"
  11260. plural "false"
  11261. caps "false"
  11262. noprefix "false"
  11263. \end_inset
  11264. , these tables show the number of probes called significantly differentially
  11265. methylated at a threshold of 10% FDR for each comparison between TX and
  11266. the other 3 transplant statuses (a) and the estimated total number of probes
  11267. that are differentially methylated (b).
  11268. \end_layout
  11269. \end_inset
  11270. \end_layout
  11271. \end_inset
  11272. \end_layout
  11273. \begin_layout Standard
  11274. \begin_inset Float figure
  11275. wide false
  11276. sideways false
  11277. status open
  11278. \begin_layout Plain Layout
  11279. \align center
  11280. \series bold
  11281. \begin_inset Float figure
  11282. wide false
  11283. sideways false
  11284. status collapsed
  11285. \begin_layout Plain Layout
  11286. \align center
  11287. \begin_inset Graphics
  11288. filename graphics/methylvoom/unadj.dupcor/pval-histograms-PAGE1.pdf
  11289. lyxscale 33
  11290. width 30col%
  11291. groupId meth-pval-hist
  11292. \end_inset
  11293. \end_layout
  11294. \begin_layout Plain Layout
  11295. \series bold
  11296. \begin_inset Caption Standard
  11297. \begin_layout Plain Layout
  11298. AR vs.
  11299. TX, Analysis A
  11300. \end_layout
  11301. \end_inset
  11302. \end_layout
  11303. \begin_layout Plain Layout
  11304. \end_layout
  11305. \end_inset
  11306. \begin_inset space \hfill{}
  11307. \end_inset
  11308. \begin_inset Float figure
  11309. wide false
  11310. sideways false
  11311. status collapsed
  11312. \begin_layout Plain Layout
  11313. \align center
  11314. \begin_inset Graphics
  11315. filename graphics/methylvoom/unadj.dupcor/pval-histograms-PAGE2.pdf
  11316. lyxscale 33
  11317. width 30col%
  11318. groupId meth-pval-hist
  11319. \end_inset
  11320. \end_layout
  11321. \begin_layout Plain Layout
  11322. \series bold
  11323. \begin_inset Caption Standard
  11324. \begin_layout Plain Layout
  11325. ADNR vs.
  11326. TX, Analysis A
  11327. \end_layout
  11328. \end_inset
  11329. \end_layout
  11330. \end_inset
  11331. \begin_inset space \hfill{}
  11332. \end_inset
  11333. \begin_inset Float figure
  11334. wide false
  11335. sideways false
  11336. status collapsed
  11337. \begin_layout Plain Layout
  11338. \align center
  11339. \begin_inset Graphics
  11340. filename graphics/methylvoom/unadj.dupcor/pval-histograms-PAGE3.pdf
  11341. lyxscale 33
  11342. width 30col%
  11343. groupId meth-pval-hist
  11344. \end_inset
  11345. \end_layout
  11346. \begin_layout Plain Layout
  11347. \series bold
  11348. \begin_inset Caption Standard
  11349. \begin_layout Plain Layout
  11350. CAN vs.
  11351. TX, Analysis A
  11352. \end_layout
  11353. \end_inset
  11354. \end_layout
  11355. \end_inset
  11356. \end_layout
  11357. \begin_layout Plain Layout
  11358. \align center
  11359. \series bold
  11360. \begin_inset Float figure
  11361. wide false
  11362. sideways false
  11363. status collapsed
  11364. \begin_layout Plain Layout
  11365. \align center
  11366. \begin_inset Graphics
  11367. filename graphics/methylvoom/unadj.dupcor.sva.aw/pval-histograms-PAGE1.pdf
  11368. lyxscale 33
  11369. width 30col%
  11370. groupId meth-pval-hist
  11371. \end_inset
  11372. \end_layout
  11373. \begin_layout Plain Layout
  11374. \series bold
  11375. \begin_inset Caption Standard
  11376. \begin_layout Plain Layout
  11377. AR vs.
  11378. TX, Analysis B
  11379. \end_layout
  11380. \end_inset
  11381. \end_layout
  11382. \end_inset
  11383. \begin_inset space \hfill{}
  11384. \end_inset
  11385. \begin_inset Float figure
  11386. wide false
  11387. sideways false
  11388. status collapsed
  11389. \begin_layout Plain Layout
  11390. \align center
  11391. \begin_inset Graphics
  11392. filename graphics/methylvoom/unadj.dupcor.sva.aw/pval-histograms-PAGE2.pdf
  11393. lyxscale 33
  11394. width 30col%
  11395. groupId meth-pval-hist
  11396. \end_inset
  11397. \end_layout
  11398. \begin_layout Plain Layout
  11399. \series bold
  11400. \begin_inset Caption Standard
  11401. \begin_layout Plain Layout
  11402. ADNR vs.
  11403. TX, Analysis B
  11404. \end_layout
  11405. \end_inset
  11406. \end_layout
  11407. \end_inset
  11408. \begin_inset space \hfill{}
  11409. \end_inset
  11410. \begin_inset Float figure
  11411. wide false
  11412. sideways false
  11413. status collapsed
  11414. \begin_layout Plain Layout
  11415. \align center
  11416. \begin_inset Graphics
  11417. filename graphics/methylvoom/unadj.dupcor.sva.aw/pval-histograms-PAGE3.pdf
  11418. lyxscale 33
  11419. width 30col%
  11420. groupId meth-pval-hist
  11421. \end_inset
  11422. \end_layout
  11423. \begin_layout Plain Layout
  11424. \series bold
  11425. \begin_inset Caption Standard
  11426. \begin_layout Plain Layout
  11427. CAN vs.
  11428. TX, Analysis B
  11429. \end_layout
  11430. \end_inset
  11431. \end_layout
  11432. \end_inset
  11433. \end_layout
  11434. \begin_layout Plain Layout
  11435. \align center
  11436. \series bold
  11437. \begin_inset Float figure
  11438. wide false
  11439. sideways false
  11440. status collapsed
  11441. \begin_layout Plain Layout
  11442. \align center
  11443. \begin_inset Graphics
  11444. filename graphics/methylvoom/unadj.dupcor.sva.voomaw/pval-histograms-PAGE1.pdf
  11445. lyxscale 33
  11446. width 30col%
  11447. groupId meth-pval-hist
  11448. \end_inset
  11449. \end_layout
  11450. \begin_layout Plain Layout
  11451. \series bold
  11452. \begin_inset Caption Standard
  11453. \begin_layout Plain Layout
  11454. AR vs.
  11455. TX, Analysis C
  11456. \end_layout
  11457. \end_inset
  11458. \end_layout
  11459. \end_inset
  11460. \begin_inset space \hfill{}
  11461. \end_inset
  11462. \begin_inset Float figure
  11463. wide false
  11464. sideways false
  11465. status collapsed
  11466. \begin_layout Plain Layout
  11467. \align center
  11468. \begin_inset Graphics
  11469. filename graphics/methylvoom/unadj.dupcor.sva.voomaw/pval-histograms-PAGE2.pdf
  11470. lyxscale 33
  11471. width 30col%
  11472. groupId meth-pval-hist
  11473. \end_inset
  11474. \end_layout
  11475. \begin_layout Plain Layout
  11476. \series bold
  11477. \begin_inset Caption Standard
  11478. \begin_layout Plain Layout
  11479. ADNR vs.
  11480. TX, Analysis C
  11481. \end_layout
  11482. \end_inset
  11483. \end_layout
  11484. \end_inset
  11485. \begin_inset space \hfill{}
  11486. \end_inset
  11487. \begin_inset Float figure
  11488. wide false
  11489. sideways false
  11490. status collapsed
  11491. \begin_layout Plain Layout
  11492. \align center
  11493. \begin_inset Graphics
  11494. filename graphics/methylvoom/unadj.dupcor.sva.voomaw/pval-histograms-PAGE3.pdf
  11495. lyxscale 33
  11496. width 30col%
  11497. groupId meth-pval-hist
  11498. \end_inset
  11499. \end_layout
  11500. \begin_layout Plain Layout
  11501. \series bold
  11502. \begin_inset Caption Standard
  11503. \begin_layout Plain Layout
  11504. CAN vs.
  11505. TX, Analysis C
  11506. \end_layout
  11507. \end_inset
  11508. \end_layout
  11509. \end_inset
  11510. \end_layout
  11511. \begin_layout Plain Layout
  11512. \begin_inset Caption Standard
  11513. \begin_layout Plain Layout
  11514. \series bold
  11515. \begin_inset CommandInset label
  11516. LatexCommand label
  11517. name "fig:meth-p-value-histograms"
  11518. \end_inset
  11519. Probe p-value histograms for each contrast in each analysis.
  11520. \series default
  11521. For each differential methylation test of interest, the distribution of
  11522. p-values across all probes is plotted as a histogram.
  11523. The red solid line indicates the density that would be expected under the
  11524. null hypothesis for all probes (a
  11525. \begin_inset Formula $\mathrm{Uniform}(0,1)$
  11526. \end_inset
  11527. distribution), while the blue dotted line indicates the fraction of p-values
  11528. that actually follow the null hypothesis (
  11529. \begin_inset Formula $\hat{\pi}_{0}$
  11530. \end_inset
  11531. ) estimated using the method of averaging local FDR values
  11532. \begin_inset CommandInset citation
  11533. LatexCommand cite
  11534. key "Phipson2013Thesis"
  11535. literal "false"
  11536. \end_inset
  11537. .
  11538. the blue line is only shown in each plot if the estimate of
  11539. \begin_inset Formula $\hat{\pi}_{0}$
  11540. \end_inset
  11541. for that p-value distribution is different from 1.
  11542. \end_layout
  11543. \end_inset
  11544. \end_layout
  11545. \end_inset
  11546. \end_layout
  11547. \begin_layout Standard
  11548. Table
  11549. \begin_inset CommandInset ref
  11550. LatexCommand ref
  11551. reference "tab:methyl-num-signif"
  11552. plural "false"
  11553. caps "false"
  11554. noprefix "false"
  11555. \end_inset
  11556. shows the number of significantly differentially methylated probes reported
  11557. by each analysis for each comparison of interest at an
  11558. \begin_inset Flex Glossary Term
  11559. status open
  11560. \begin_layout Plain Layout
  11561. FDR
  11562. \end_layout
  11563. \end_inset
  11564. of 10%.
  11565. As expected, the more elaborate analyses, B and C, report more significant
  11566. probes than the more basic analysis A, consistent with the conclusions
  11567. above that the data contain hidden systematic variations that must be modeled.
  11568. Table
  11569. \begin_inset CommandInset ref
  11570. LatexCommand ref
  11571. reference "tab:methyl-est-nonnull"
  11572. plural "false"
  11573. caps "false"
  11574. noprefix "false"
  11575. \end_inset
  11576. shows the estimated number differentially methylated probes for each test
  11577. from each analysis.
  11578. This was computed by estimating the proportion of null hypotheses that
  11579. were true using the method of
  11580. \begin_inset CommandInset citation
  11581. LatexCommand cite
  11582. key "Phipson2013Thesis"
  11583. literal "false"
  11584. \end_inset
  11585. and subtracting that fraction from the total number of probes, yielding
  11586. an estimate of the number of null hypotheses that are false based on the
  11587. distribution of p-values across the entire dataset.
  11588. Note that this does not identify which null hypotheses should be rejected
  11589. (i.e.
  11590. which probes are significant); it only estimates the true number of such
  11591. probes.
  11592. Once again, analyses B and C result it much larger estimates for the number
  11593. of differentially methylated probes.
  11594. In this case, analysis C, the only analysis that includes voom, estimates
  11595. the largest number of differentially methylated probes for all 3 contrasts.
  11596. If the assumptions of all the methods employed hold, then this represents
  11597. a gain in statistical power over the simpler analysis A.
  11598. Figure
  11599. \begin_inset CommandInset ref
  11600. LatexCommand ref
  11601. reference "fig:meth-p-value-histograms"
  11602. plural "false"
  11603. caps "false"
  11604. noprefix "false"
  11605. \end_inset
  11606. shows the p-value distributions for each test, from which the numbers in
  11607. Table
  11608. \begin_inset CommandInset ref
  11609. LatexCommand ref
  11610. reference "tab:methyl-est-nonnull"
  11611. plural "false"
  11612. caps "false"
  11613. noprefix "false"
  11614. \end_inset
  11615. were generated.
  11616. The distributions for analysis A all have a dip in density near zero, which
  11617. is a strong sign of a poor model fit.
  11618. The histograms for analyses B and C are more well-behaved, with a uniform
  11619. component stretching all the way from 0 to 1 representing the probes for
  11620. which the null hypotheses is true (no differential methylation), and a
  11621. zero-biased component representing the probes for which the null hypothesis
  11622. is false (differentially methylated).
  11623. These histograms do not indicate any major issues with the model fit.
  11624. \end_layout
  11625. \begin_layout Standard
  11626. \begin_inset Flex TODO Note (inline)
  11627. status open
  11628. \begin_layout Plain Layout
  11629. If time allows, maybe generate the PCA plots before/after SVA effect subtraction
  11630. ?
  11631. \end_layout
  11632. \end_inset
  11633. \end_layout
  11634. \begin_layout Section
  11635. Discussion
  11636. \end_layout
  11637. \begin_layout Subsection
  11638. fRMA achieves clinically applicable normalization without sacrificing classifica
  11639. tion performance
  11640. \end_layout
  11641. \begin_layout Standard
  11642. As shown in Figure
  11643. \begin_inset CommandInset ref
  11644. LatexCommand ref
  11645. reference "fig:Classifier-probabilities-RMA"
  11646. plural "false"
  11647. caps "false"
  11648. noprefix "false"
  11649. \end_inset
  11650. , improper normalization, particularly separate normalization of training
  11651. and test samples, leads to unwanted biases in classification.
  11652. In a controlled experimental context, it is always possible to correct
  11653. this issue by normalizing all experimental samples together.
  11654. However, because it is not feasible to normalize all samples together in
  11655. a clinical context, a single-channel normalization is required is required.
  11656. \end_layout
  11657. \begin_layout Standard
  11658. The major concern in using a single-channel normalization is that non-single-cha
  11659. nnel methods can share information between arrays to improve the normalization,
  11660. and single-channel methods risk sacrificing the gains in normalization
  11661. accuracy that come from this information sharing.
  11662. In the case of
  11663. \begin_inset Flex Glossary Term
  11664. status open
  11665. \begin_layout Plain Layout
  11666. RMA
  11667. \end_layout
  11668. \end_inset
  11669. , this information sharing is accomplished through quantile normalization
  11670. and median polish steps.
  11671. The need for information sharing in quantile normalization can easily be
  11672. removed by learning a fixed set of quantiles from external data and normalizing
  11673. each array to these fixed quantiles, instead of the quantiles of the data
  11674. itself.
  11675. As long as the fixed quantiles are reasonable, the result will be similar
  11676. to standard
  11677. \begin_inset Flex Glossary Term
  11678. status open
  11679. \begin_layout Plain Layout
  11680. RMA
  11681. \end_layout
  11682. \end_inset
  11683. .
  11684. However, there is no analogous way to eliminate cross-array information
  11685. sharing in the median polish step, so
  11686. \begin_inset Flex Glossary Term
  11687. status open
  11688. \begin_layout Plain Layout
  11689. fRMA
  11690. \end_layout
  11691. \end_inset
  11692. replaces this with a weighted average of probes on each array, with the
  11693. weights learned from external data.
  11694. This step of
  11695. \begin_inset Flex Glossary Term
  11696. status open
  11697. \begin_layout Plain Layout
  11698. fRMA
  11699. \end_layout
  11700. \end_inset
  11701. has the greatest potential to diverge from RMA un undesirable ways.
  11702. \end_layout
  11703. \begin_layout Standard
  11704. However, when run on real data,
  11705. \begin_inset Flex Glossary Term
  11706. status open
  11707. \begin_layout Plain Layout
  11708. fRMA
  11709. \end_layout
  11710. \end_inset
  11711. performed at least as well as
  11712. \begin_inset Flex Glossary Term
  11713. status open
  11714. \begin_layout Plain Layout
  11715. RMA
  11716. \end_layout
  11717. \end_inset
  11718. in both the internal validation and external validation tests.
  11719. This shows that
  11720. \begin_inset Flex Glossary Term
  11721. status open
  11722. \begin_layout Plain Layout
  11723. fRMA
  11724. \end_layout
  11725. \end_inset
  11726. can be used to normalize individual clinical samples in a class prediction
  11727. context without sacrificing the classifier performance that would be obtained
  11728. by using the more well-established
  11729. \begin_inset Flex Glossary Term
  11730. status open
  11731. \begin_layout Plain Layout
  11732. RMA
  11733. \end_layout
  11734. \end_inset
  11735. for normalization.
  11736. The other single-channel normalization method considered,
  11737. \begin_inset Flex Glossary Term
  11738. status open
  11739. \begin_layout Plain Layout
  11740. SCAN
  11741. \end_layout
  11742. \end_inset
  11743. , showed some loss of
  11744. \begin_inset Flex Glossary Term
  11745. status open
  11746. \begin_layout Plain Layout
  11747. AUC
  11748. \end_layout
  11749. \end_inset
  11750. in the external validation test.
  11751. Based on these results,
  11752. \begin_inset Flex Glossary Term
  11753. status open
  11754. \begin_layout Plain Layout
  11755. fRMA
  11756. \end_layout
  11757. \end_inset
  11758. is the preferred normalization for clinical samples in a class prediction
  11759. context.
  11760. \end_layout
  11761. \begin_layout Subsection
  11762. Robust fRMA vectors can be generated for new array platforms
  11763. \end_layout
  11764. \begin_layout Standard
  11765. \begin_inset Flex TODO Note (inline)
  11766. status open
  11767. \begin_layout Plain Layout
  11768. Look up the exact numbers, do a find & replace for
  11769. \begin_inset Quotes eld
  11770. \end_inset
  11771. 850
  11772. \begin_inset Quotes erd
  11773. \end_inset
  11774. \end_layout
  11775. \end_inset
  11776. \end_layout
  11777. \begin_layout Standard
  11778. The published
  11779. \begin_inset Flex Glossary Term
  11780. status open
  11781. \begin_layout Plain Layout
  11782. fRMA
  11783. \end_layout
  11784. \end_inset
  11785. normalization vectors for the hgu133plus2 platform were generated from
  11786. a set of about 850 samples chosen from a wide range of tissues, which the
  11787. authors determined was sufficient to generate a robust set of normalization
  11788. vectors that could be applied across all tissues
  11789. \begin_inset CommandInset citation
  11790. LatexCommand cite
  11791. key "McCall2010"
  11792. literal "false"
  11793. \end_inset
  11794. .
  11795. Since we only had hthgu133pluspm for 2 tissues of interest, our needs were
  11796. more modest.
  11797. Even using only 130 samples in 26 batches of 5 samples each for kidney
  11798. biopsies, we were able to train a robust set of
  11799. \begin_inset Flex Glossary Term
  11800. status open
  11801. \begin_layout Plain Layout
  11802. fRMA
  11803. \end_layout
  11804. \end_inset
  11805. normalization vectors that were not meaningfully affected by the random
  11806. selection of 5 samples from each batch.
  11807. As expected, the training process was just as robust for the blood samples
  11808. with 230 samples in 46 batches of 5 samples each.
  11809. Because these vectors were each generated using training samples from a
  11810. single tissue, they are not suitable for general use, unlike the vectors
  11811. provided with
  11812. \begin_inset Flex Glossary Term
  11813. status open
  11814. \begin_layout Plain Layout
  11815. fRMA
  11816. \end_layout
  11817. \end_inset
  11818. itself.
  11819. They are purpose-built for normalizing a specific type of sample on a specific
  11820. platform.
  11821. This is a mostly acceptable limitation in the context of developing a machine
  11822. learning classifier for diagnosing a disease based on samples of a specific
  11823. tissue.
  11824. \end_layout
  11825. \begin_layout Standard
  11826. \begin_inset Flex TODO Note (inline)
  11827. status open
  11828. \begin_layout Plain Layout
  11829. Talk about how these vectors can be used for any data from these tissues
  11830. on this platform even though they were custom made for this data set.
  11831. \end_layout
  11832. \end_inset
  11833. \end_layout
  11834. \begin_layout Standard
  11835. \begin_inset Flex TODO Note (inline)
  11836. status open
  11837. \begin_layout Plain Layout
  11838. How to bring up that these custom vectors were used in another project by
  11839. someone else that was never published?
  11840. \end_layout
  11841. \end_inset
  11842. \end_layout
  11843. \begin_layout Subsection
  11844. Methylation array data can be successfully analyzed using existing techniques,
  11845. but machine learning poses additional challenges
  11846. \end_layout
  11847. \begin_layout Standard
  11848. Both analysis strategies B and C both yield a reasonable analysis, with
  11849. a mean-variance trend that matches the expected behavior for the non-linear
  11850. M-value transformation (Figure
  11851. \begin_inset CommandInset ref
  11852. LatexCommand ref
  11853. reference "fig:meanvar-sva-aw"
  11854. plural "false"
  11855. caps "false"
  11856. noprefix "false"
  11857. \end_inset
  11858. ) and well-behaved p-value distributions (Figure
  11859. \begin_inset CommandInset ref
  11860. LatexCommand ref
  11861. reference "fig:meth-p-value-histograms"
  11862. plural "false"
  11863. caps "false"
  11864. noprefix "false"
  11865. \end_inset
  11866. ).
  11867. These two analyses also yield similar numbers of significant probes (Table
  11868. \begin_inset CommandInset ref
  11869. LatexCommand ref
  11870. reference "tab:methyl-num-signif"
  11871. plural "false"
  11872. caps "false"
  11873. noprefix "false"
  11874. \end_inset
  11875. ) and similar estimates of the number of differentially methylated probes
  11876. (Table
  11877. \begin_inset CommandInset ref
  11878. LatexCommand ref
  11879. reference "tab:methyl-est-nonnull"
  11880. plural "false"
  11881. caps "false"
  11882. noprefix "false"
  11883. \end_inset
  11884. ).
  11885. The main difference between these two analyses is the method used to account
  11886. for the mean-variance trend.
  11887. In analysis B, the trend is estimated and applied at the probe level: each
  11888. probe's estimated variance is squeezed toward the trend using an empirical
  11889. Bayes procedure (Figure
  11890. \begin_inset CommandInset ref
  11891. LatexCommand ref
  11892. reference "fig:meanvar-sva-aw"
  11893. plural "false"
  11894. caps "false"
  11895. noprefix "false"
  11896. \end_inset
  11897. ).
  11898. In analysis C, the trend is still estimated at the probe level, but instead
  11899. of estimating a single variance value shared across all observations for
  11900. a given probe, the voom method computes an initial estimate of the variance
  11901. for each observation individually based on where its model-fitted M-value
  11902. falls on the trend line and then assigns inverse-variance weights to model
  11903. the difference in variance between observations.
  11904. An overall variance is still estimated for each probe using the same empirical
  11905. Bayes method, but now the residual trend is flat (Figure
  11906. \begin_inset CommandInset ref
  11907. LatexCommand ref
  11908. reference "fig:meanvar-sva-voomaw"
  11909. plural "false"
  11910. caps "false"
  11911. noprefix "false"
  11912. \end_inset
  11913. ), indicating that the mean-variance trend is adequately modeled by scaling
  11914. the estimated variance for each observation using the weights computed
  11915. by voom.
  11916. \end_layout
  11917. \begin_layout Standard
  11918. The difference between the standard empirical Bayes trended variance modeling
  11919. (analysis B) and voom (analysis C) is analogous to the difference between
  11920. a t-test with equal variance and a t-test with unequal variance, except
  11921. that the unequal group variances used in the latter test are estimated
  11922. based on the mean-variance trend from all the probes rather than the data
  11923. for the specific probe being tested, thus stabilizing the group variance
  11924. estimates by sharing information between probes.
  11925. Allowing voom to model the variance using observation weights in this manner
  11926. allows the linear model fit to concentrate statistical power where it will
  11927. do the most good.
  11928. For example, if a particular probe's M-values are always at the extreme
  11929. of the M-value range (e.g.
  11930. less than -4) for
  11931. \begin_inset Flex Glossary Term
  11932. status open
  11933. \begin_layout Plain Layout
  11934. ADNR
  11935. \end_layout
  11936. \end_inset
  11937. samples, but the M-values for that probe in
  11938. \begin_inset Flex Glossary Term
  11939. status open
  11940. \begin_layout Plain Layout
  11941. TX
  11942. \end_layout
  11943. \end_inset
  11944. and
  11945. \begin_inset Flex Glossary Term
  11946. status open
  11947. \begin_layout Plain Layout
  11948. CAN
  11949. \end_layout
  11950. \end_inset
  11951. samples are within the flat region of the mean-variance trend (between
  11952. -3 and +3), voom is able to down-weight the contribution of the high-variance
  11953. M-values from the
  11954. \begin_inset Flex Glossary Term
  11955. status open
  11956. \begin_layout Plain Layout
  11957. ADNR
  11958. \end_layout
  11959. \end_inset
  11960. samples in order to gain more statistical power while testing for differential
  11961. methylation between
  11962. \begin_inset Flex Glossary Term
  11963. status open
  11964. \begin_layout Plain Layout
  11965. TX
  11966. \end_layout
  11967. \end_inset
  11968. and
  11969. \begin_inset Flex Glossary Term
  11970. status open
  11971. \begin_layout Plain Layout
  11972. CAN
  11973. \end_layout
  11974. \end_inset
  11975. .
  11976. In contrast, modeling the mean-variance trend only at the probe level would
  11977. combine the high-variance
  11978. \begin_inset Flex Glossary Term
  11979. status open
  11980. \begin_layout Plain Layout
  11981. ADNR
  11982. \end_layout
  11983. \end_inset
  11984. samples and lower-variance samples from other conditions and estimate an
  11985. intermediate variance for this probe.
  11986. In practice, analysis B shows that this approach is adequate, but the voom
  11987. approach in analysis C is at least as good on all model fit criteria and
  11988. yields a larger estimate for the number of differentially methylated genes,
  11989. \emph on
  11990. and
  11991. \emph default
  11992. it matches up better with the theoretical
  11993. \end_layout
  11994. \begin_layout Standard
  11995. The significant association of diabetes diagnosis with sample quality is
  11996. interesting.
  11997. The samples with
  11998. \begin_inset Flex Glossary Term
  11999. status open
  12000. \begin_layout Plain Layout
  12001. T2D
  12002. \end_layout
  12003. \end_inset
  12004. tended to have more variation, averaged across all probes, than those with
  12005. \begin_inset Flex Glossary Term
  12006. status open
  12007. \begin_layout Plain Layout
  12008. T1D
  12009. \end_layout
  12010. \end_inset
  12011. .
  12012. This is consistent with the consensus that
  12013. \begin_inset Flex Glossary Term
  12014. status open
  12015. \begin_layout Plain Layout
  12016. T2D
  12017. \end_layout
  12018. \end_inset
  12019. and the associated metabolic syndrome represent a broad dysregulation of
  12020. the body's endocrine signaling related to metabolism [citation needed].
  12021. This dysregulation could easily manifest as a greater degree of variation
  12022. in the DNA methylation patterns of affected tissues.
  12023. In contrast,
  12024. \begin_inset Flex Glossary Term
  12025. status open
  12026. \begin_layout Plain Layout
  12027. T1D
  12028. \end_layout
  12029. \end_inset
  12030. has a more specific cause and effect, so a less variable methylation signature
  12031. is expected.
  12032. \end_layout
  12033. \begin_layout Standard
  12034. This preliminary analysis suggests that some degree of differential methylation
  12035. exists between
  12036. \begin_inset Flex Glossary Term
  12037. status open
  12038. \begin_layout Plain Layout
  12039. TX
  12040. \end_layout
  12041. \end_inset
  12042. and each of the three types of transplant disfunction studied.
  12043. Hence, it may be feasible to train a classifier to diagnose transplant
  12044. disfunction from DNA methylation array data.
  12045. However, the major importance of both
  12046. \begin_inset Flex Glossary Term
  12047. status open
  12048. \begin_layout Plain Layout
  12049. SVA
  12050. \end_layout
  12051. \end_inset
  12052. and sample quality weighting for proper modeling of this data poses significant
  12053. challenges for any attempt at a machine learning on data of similar quality.
  12054. While these are easily used in a modeling context with full sample information,
  12055. neither of these methods is directly applicable in a machine learning context,
  12056. where the diagnosis is not known ahead of time.
  12057. If a machine learning approach for methylation-based diagnosis is to be
  12058. pursued, it will either require machine-learning-friendly methods to address
  12059. the same systematic trends in the data that
  12060. \begin_inset Flex Glossary Term
  12061. status open
  12062. \begin_layout Plain Layout
  12063. SVA
  12064. \end_layout
  12065. \end_inset
  12066. and sample quality weighting address, or it will require higher quality
  12067. data with substantially less systematic perturbation of the data.
  12068. \end_layout
  12069. \begin_layout Section
  12070. Future Directions
  12071. \end_layout
  12072. \begin_layout Standard
  12073. \begin_inset Flex TODO Note (inline)
  12074. status open
  12075. \begin_layout Plain Layout
  12076. Some work was already being done with the existing fRMA vectors.
  12077. Do I mention that here?
  12078. \end_layout
  12079. \end_inset
  12080. \end_layout
  12081. \begin_layout Subsection
  12082. Improving fRMA to allow training from batches of unequal size
  12083. \end_layout
  12084. \begin_layout Standard
  12085. Because the tools for building
  12086. \begin_inset Flex Glossary Term
  12087. status open
  12088. \begin_layout Plain Layout
  12089. fRMA
  12090. \end_layout
  12091. \end_inset
  12092. normalization vectors require equal-size batches, many samples must be
  12093. discarded from the training data.
  12094. This is undesirable for a few reasons.
  12095. First, more data is simply better, all other things being equal.
  12096. In this case,
  12097. \begin_inset Quotes eld
  12098. \end_inset
  12099. better
  12100. \begin_inset Quotes erd
  12101. \end_inset
  12102. means a more precise estimate of normalization parameters.
  12103. In addition, the samples to be discarded must be chosen arbitrarily, which
  12104. introduces an unnecessary element of randomness into the estimation process.
  12105. While the randomness can be made deterministic by setting a consistent
  12106. random seed, the need for equal size batches also introduces a need for
  12107. the analyst to decide on the appropriate trade-off between batch size and
  12108. the number of batches.
  12109. This introduces an unnecessary and undesirable
  12110. \begin_inset Quotes eld
  12111. \end_inset
  12112. researcher degree of freedom
  12113. \begin_inset Quotes erd
  12114. \end_inset
  12115. into the analysis, since the generated normalization vectors now depend
  12116. on the choice of batch size based on vague selection criteria and instinct,
  12117. which can unintentionally introduce bias if the researcher chooses a batch
  12118. size based on what seems to yield the most favorable downstream results
  12119. \begin_inset CommandInset citation
  12120. LatexCommand cite
  12121. key "Simmons2011"
  12122. literal "false"
  12123. \end_inset
  12124. .
  12125. \end_layout
  12126. \begin_layout Standard
  12127. Fortunately, the requirement for equal-size batches is not inherent to the
  12128. \begin_inset Flex Glossary Term
  12129. status open
  12130. \begin_layout Plain Layout
  12131. fRMA
  12132. \end_layout
  12133. \end_inset
  12134. algorithm but rather a limitation of the implementation in the
  12135. \begin_inset Flex Code
  12136. status open
  12137. \begin_layout Plain Layout
  12138. frmaTools
  12139. \end_layout
  12140. \end_inset
  12141. package.
  12142. In personal communication, the package's author, Matthew McCall, has indicated
  12143. that with some work, it should be possible to improve the implementation
  12144. to work with batches of unequal sizes.
  12145. The current implementation ignores the batch size when calculating with-batch
  12146. and between-batch residual variances, since the batch size constant cancels
  12147. out later in the calculations as long as all batches are of equal size.
  12148. Hence, the calculations of these parameters would need to be modified to
  12149. remove this optimization and properly calculate the variances using the
  12150. full formula.
  12151. Once this modification is made, a new strategy would need to be developed
  12152. for assessing the stability of parameter estimates, since the random subsamplin
  12153. g step is eliminated, meaning that different subsamplings can no longer
  12154. be compared as in Figures
  12155. \begin_inset CommandInset ref
  12156. LatexCommand ref
  12157. reference "fig:frma-violin"
  12158. plural "false"
  12159. caps "false"
  12160. noprefix "false"
  12161. \end_inset
  12162. and
  12163. \begin_inset CommandInset ref
  12164. LatexCommand ref
  12165. reference "fig:Representative-MA-plots"
  12166. plural "false"
  12167. caps "false"
  12168. noprefix "false"
  12169. \end_inset
  12170. .
  12171. Bootstrap resampling is likely a good candidate here: sample many training
  12172. sets of equal size from the existing training set with replacement, estimate
  12173. parameters from each resampled training set, and compare the estimated
  12174. parameters between bootstraps in order to quantify the variability in each
  12175. parameter's estimation.
  12176. \end_layout
  12177. \begin_layout Subsection
  12178. Developing methylation arrays as a diagnostic tool for kidney transplant
  12179. rejection
  12180. \end_layout
  12181. \begin_layout Standard
  12182. The current study has showed that DNA methylation, as assayed by Illumina
  12183. 450k methylation arrays, has some potential for diagnosing transplant dysfuncti
  12184. ons, including rejection.
  12185. However, very few probes could be confidently identified as differentially
  12186. methylated between healthy and dysfunctional transplants.
  12187. One likely explanation for this is the predominant influence of unobserved
  12188. confounding factors.
  12189. \begin_inset Flex Glossary Term
  12190. status open
  12191. \begin_layout Plain Layout
  12192. SVA
  12193. \end_layout
  12194. \end_inset
  12195. can model and correct for such factors, but the correction can never be
  12196. perfect, so some degree of unwanted systematic variation will always remain
  12197. after
  12198. \begin_inset Flex Glossary Term
  12199. status open
  12200. \begin_layout Plain Layout
  12201. SVA
  12202. \end_layout
  12203. \end_inset
  12204. correction.
  12205. If the effect size of the confounding factors was similar to that of the
  12206. factor of interest (in this case, transplant status), this would be an
  12207. acceptable limitation, since removing most of the confounding factors'
  12208. effects would allow the main effect to stand out.
  12209. However, in this data set, the confounding factors have a much larger effect
  12210. size than transplant status, which means that the small degree of remaining
  12211. variation not removed by
  12212. \begin_inset Flex Glossary Term
  12213. status open
  12214. \begin_layout Plain Layout
  12215. SVA
  12216. \end_layout
  12217. \end_inset
  12218. can still swamp the effect of interest, making it difficult to detect.
  12219. This is, of course, a major issue when the end goal is to develop a classifier
  12220. to diagnose transplant rejection from methylation data, since batch-correction
  12221. methods like
  12222. \begin_inset Flex Glossary Term
  12223. status open
  12224. \begin_layout Plain Layout
  12225. SVA
  12226. \end_layout
  12227. \end_inset
  12228. that work in a linear modeling context cannot be applied in a machine learning
  12229. context.
  12230. \end_layout
  12231. \begin_layout Standard
  12232. Currently, the source of these unwanted systematic variations in the data
  12233. is unknown.
  12234. The best solution would be to determine the cause of the variation and
  12235. eliminate it, thereby eliminating the need to model and remove that variation.
  12236. However, if this proves impractical, another option is to use
  12237. \begin_inset Flex Glossary Term
  12238. status open
  12239. \begin_layout Plain Layout
  12240. SVA
  12241. \end_layout
  12242. \end_inset
  12243. to identify probes that are highly associated with the surrogate variables
  12244. that describe the unwanted variation in the data.
  12245. These probes could be discarded prior to classifier training, in order
  12246. to maximize the chance that the training algorithm will be able to identify
  12247. highly predictive probes from those remaining.
  12248. Lastly, it is possible that some of this unwanted variation is a result
  12249. of the array-based assay being used and would be eliminated by switching
  12250. to assaying DNA methylation using bisulphite sequencing.
  12251. However, this carries the risk that the sequencing assay will have its
  12252. own set of biases that must be corrected for in a different way.
  12253. \end_layout
  12254. \begin_layout Chapter
  12255. Globin-blocking for more effective blood RNA-seq analysis in primate animal
  12256. model
  12257. \end_layout
  12258. \begin_layout Standard
  12259. \begin_inset ERT
  12260. status collapsed
  12261. \begin_layout Plain Layout
  12262. \backslash
  12263. glsresetall
  12264. \end_layout
  12265. \end_inset
  12266. \end_layout
  12267. \begin_layout Standard
  12268. \begin_inset Flex TODO Note (inline)
  12269. status open
  12270. \begin_layout Plain Layout
  12271. Choose between above and the paper title: Optimizing yield of deep RNA sequencin
  12272. g for gene expression profiling by globin reduction of peripheral blood
  12273. samples from cynomolgus monkeys (Macaca fascicularis).
  12274. \end_layout
  12275. \end_inset
  12276. \end_layout
  12277. \begin_layout Standard
  12278. \begin_inset Flex TODO Note (inline)
  12279. status open
  12280. \begin_layout Plain Layout
  12281. Chapter author list:
  12282. \begin_inset CommandInset href
  12283. LatexCommand href
  12284. target "https://tex.stackexchange.com/questions/156862/displaying-author-for-each-chapter-in-book"
  12285. \end_inset
  12286. Every chapter gets an author list, which may or may not be part of a citation
  12287. to a published/preprinted paper.
  12288. \end_layout
  12289. \end_inset
  12290. \end_layout
  12291. \begin_layout Standard
  12292. \begin_inset Flex TODO Note (inline)
  12293. status open
  12294. \begin_layout Plain Layout
  12295. Fix primes and such using math-insert
  12296. \end_layout
  12297. \end_inset
  12298. \end_layout
  12299. \begin_layout Section*
  12300. Abstract
  12301. \end_layout
  12302. \begin_layout Standard
  12303. \begin_inset Flex TODO Note (inline)
  12304. status open
  12305. \begin_layout Plain Layout
  12306. If the other chapters don't get abstracts, this one probably shouldn't either.
  12307. But parts of it can be copied into the final abstract.
  12308. \end_layout
  12309. \end_inset
  12310. \end_layout
  12311. \begin_layout Paragraph
  12312. Background
  12313. \end_layout
  12314. \begin_layout Standard
  12315. Primate blood contains high concentrations of globin
  12316. \begin_inset Flex Glossary Term
  12317. status open
  12318. \begin_layout Plain Layout
  12319. mRNA
  12320. \end_layout
  12321. \end_inset
  12322. .
  12323. Globin reduction is a standard technique used to improve the expression
  12324. results obtained by DNA microarrays on RNA from blood samples.
  12325. However, with
  12326. \begin_inset Flex Glossary Term
  12327. status open
  12328. \begin_layout Plain Layout
  12329. RNA-seq
  12330. \end_layout
  12331. \end_inset
  12332. quickly replacing microarrays for many applications, the impact of globin
  12333. reduction for
  12334. \begin_inset Flex Glossary Term
  12335. status open
  12336. \begin_layout Plain Layout
  12337. RNA-seq
  12338. \end_layout
  12339. \end_inset
  12340. has not been previously studied.
  12341. Moreover, no off-the-shelf kits are available for globin reduction in nonhuman
  12342. primates.
  12343. \end_layout
  12344. \begin_layout Paragraph
  12345. Results
  12346. \end_layout
  12347. \begin_layout Standard
  12348. Here we report a protocol for
  12349. \begin_inset Flex Glossary Term
  12350. status open
  12351. \begin_layout Plain Layout
  12352. RNA-seq
  12353. \end_layout
  12354. \end_inset
  12355. in primate blood samples that uses complimentary
  12356. \begin_inset ERT
  12357. status open
  12358. \begin_layout Plain Layout
  12359. \backslash
  12360. glspl*{oligo}
  12361. \end_layout
  12362. \end_inset
  12363. to block reverse transcription of the alpha and beta globin genes.
  12364. In test samples from cynomolgus monkeys (
  12365. \emph on
  12366. Macaca fascicularis
  12367. \emph default
  12368. ), this
  12369. \begin_inset Flex Glossary Term
  12370. status open
  12371. \begin_layout Plain Layout
  12372. GB
  12373. \end_layout
  12374. \end_inset
  12375. \begin_inset CommandInset nomenclature
  12376. LatexCommand nomenclature
  12377. symbol "GB"
  12378. description "globin blocking"
  12379. literal "false"
  12380. \end_inset
  12381. protocol approximately doubles the yield of informative (non-globin) reads
  12382. by greatly reducing the fraction of globin reads, while also improving
  12383. the consistency in sequencing depth between samples.
  12384. The increased yield enables detection of about 2000 more genes, significantly
  12385. increases the correlation in measured gene expression levels between samples,
  12386. and increases the sensitivity of differential gene expression tests.
  12387. \end_layout
  12388. \begin_layout Paragraph
  12389. Conclusions
  12390. \end_layout
  12391. \begin_layout Standard
  12392. These results show that
  12393. \begin_inset Flex Glossary Term
  12394. status open
  12395. \begin_layout Plain Layout
  12396. GB
  12397. \end_layout
  12398. \end_inset
  12399. significantly improves the cost-effectiveness of
  12400. \begin_inset Flex Glossary Term
  12401. status open
  12402. \begin_layout Plain Layout
  12403. RNA-seq
  12404. \end_layout
  12405. \end_inset
  12406. in primate blood samples by doubling the yield of useful reads, allowing
  12407. detection of more genes, and improving the precision of gene expression
  12408. measurements.
  12409. Based on these results, a globin reducing or blocking protocol is recommended
  12410. for all
  12411. \begin_inset Flex Glossary Term
  12412. status open
  12413. \begin_layout Plain Layout
  12414. RNA-seq
  12415. \end_layout
  12416. \end_inset
  12417. studies of primate blood samples.
  12418. \end_layout
  12419. \begin_layout Standard
  12420. \begin_inset ERT
  12421. status collapsed
  12422. \begin_layout Plain Layout
  12423. \backslash
  12424. glsresetall
  12425. \end_layout
  12426. \end_inset
  12427. \end_layout
  12428. \begin_layout Section
  12429. Approach
  12430. \end_layout
  12431. \begin_layout Standard
  12432. \begin_inset Note Note
  12433. status open
  12434. \begin_layout Plain Layout
  12435. Consider putting some of this in the Intro chapter
  12436. \end_layout
  12437. \begin_layout Itemize
  12438. Cynomolgus monkeys as a model organism
  12439. \end_layout
  12440. \begin_deeper
  12441. \begin_layout Itemize
  12442. Highly related to humans
  12443. \end_layout
  12444. \begin_layout Itemize
  12445. Small size and short life cycle - good research animal
  12446. \end_layout
  12447. \begin_layout Itemize
  12448. Genomics resources still in development
  12449. \end_layout
  12450. \end_deeper
  12451. \begin_layout Itemize
  12452. Inadequacy of existing blood RNA-seq protocols
  12453. \end_layout
  12454. \begin_deeper
  12455. \begin_layout Itemize
  12456. Existing protocols use a separate globin pulldown step, slowing down processing
  12457. \end_layout
  12458. \end_deeper
  12459. \end_inset
  12460. \end_layout
  12461. \begin_layout Standard
  12462. Increasingly, researchers are turning to
  12463. \begin_inset Flex Glossary Term
  12464. status open
  12465. \begin_layout Plain Layout
  12466. RNA-seq
  12467. \end_layout
  12468. \end_inset
  12469. in preference to expression microarrays for analysis of gene expression
  12470. \begin_inset CommandInset citation
  12471. LatexCommand cite
  12472. key "Mutz2012"
  12473. literal "false"
  12474. \end_inset
  12475. .
  12476. The advantages are even greater for study of model organisms with no well-estab
  12477. lished array platforms available, such as the cynomolgus monkey (Macaca
  12478. fascicularis).
  12479. High fractions of globin
  12480. \begin_inset Flex Glossary Term
  12481. status open
  12482. \begin_layout Plain Layout
  12483. mRNA
  12484. \end_layout
  12485. \end_inset
  12486. \begin_inset CommandInset nomenclature
  12487. LatexCommand nomenclature
  12488. symbol "mRNA"
  12489. description "messenger RNA"
  12490. literal "false"
  12491. \end_inset
  12492. are naturally present in mammalian peripheral blood samples (up to 70%
  12493. of total
  12494. \begin_inset Flex Glossary Term
  12495. status open
  12496. \begin_layout Plain Layout
  12497. mRNA
  12498. \end_layout
  12499. \end_inset
  12500. ) and these are known to interfere with the results of array-based expression
  12501. profiling
  12502. \begin_inset CommandInset citation
  12503. LatexCommand cite
  12504. key "Winn2010"
  12505. literal "false"
  12506. \end_inset
  12507. .
  12508. The importance of globin reduction for
  12509. \begin_inset Flex Glossary Term
  12510. status open
  12511. \begin_layout Plain Layout
  12512. RNA-seq
  12513. \end_layout
  12514. \end_inset
  12515. of blood has only been evaluated for a deepSAGE protocol on human samples
  12516. \begin_inset CommandInset citation
  12517. LatexCommand cite
  12518. key "Mastrokolias2012"
  12519. literal "false"
  12520. \end_inset
  12521. .
  12522. In the present report, we evaluated globin reduction using custom blocking
  12523. \begin_inset ERT
  12524. status open
  12525. \begin_layout Plain Layout
  12526. \backslash
  12527. glspl*{oligo}
  12528. \end_layout
  12529. \end_inset
  12530. for deep
  12531. \begin_inset Flex Glossary Term
  12532. status open
  12533. \begin_layout Plain Layout
  12534. RNA-seq
  12535. \end_layout
  12536. \end_inset
  12537. of peripheral blood samples from a nonhuman primate, cynomolgus monkey,
  12538. using the Illumina technology platform.
  12539. We demonstrate that globin reduction significantly improves the cost-effectiven
  12540. ess of
  12541. \begin_inset Flex Glossary Term
  12542. status open
  12543. \begin_layout Plain Layout
  12544. RNA-seq
  12545. \end_layout
  12546. \end_inset
  12547. in blood samples.
  12548. Thus, our protocol offers a significant advantage to any investigator planning
  12549. to use
  12550. \begin_inset Flex Glossary Term
  12551. status open
  12552. \begin_layout Plain Layout
  12553. RNA-seq
  12554. \end_layout
  12555. \end_inset
  12556. for gene expression profiling of nonhuman primate blood samples.
  12557. Our method can be generally applied to any species by designing complementary
  12558. \begin_inset Flex Glossary Term
  12559. status open
  12560. \begin_layout Plain Layout
  12561. oligo
  12562. \end_layout
  12563. \end_inset
  12564. blocking probes to the globin gene sequences of that species.
  12565. Indeed, any highly expressed but biologically uninformative transcripts
  12566. can also be blocked to further increase sequencing efficiency and value
  12567. \begin_inset CommandInset citation
  12568. LatexCommand cite
  12569. key "Arnaud2016"
  12570. literal "false"
  12571. \end_inset
  12572. .
  12573. \end_layout
  12574. \begin_layout Section
  12575. Methods
  12576. \end_layout
  12577. \begin_layout Subsection
  12578. Sample collection
  12579. \end_layout
  12580. \begin_layout Standard
  12581. All research reported here was done under IACUC-approved protocols at the
  12582. University of Miami and complied with all applicable federal and state
  12583. regulations and ethical principles for nonhuman primate research.
  12584. Blood draws occurred between 16 April 2012 and 18 June 2015.
  12585. The experimental system involved intrahepatic pancreatic islet transplantation
  12586. into Cynomolgus monkeys with induced diabetes mellitus with or without
  12587. concomitant infusion of mesenchymal stem cells.
  12588. Blood was collected at serial time points before and after transplantation
  12589. into PAXgene Blood RNA tubes (PreAnalytiX/Qiagen, Valencia, CA) at the
  12590. precise volume:volume ratio of 2.5 ml whole blood into 6.9 ml of PAX gene
  12591. additive.
  12592. \end_layout
  12593. \begin_layout Subsection
  12594. Globin Blocking
  12595. \end_layout
  12596. \begin_layout Standard
  12597. Four
  12598. \begin_inset ERT
  12599. status open
  12600. \begin_layout Plain Layout
  12601. \backslash
  12602. glspl*{oligo}
  12603. \end_layout
  12604. \end_inset
  12605. were designed to hybridize to the
  12606. \begin_inset Formula $3^{\prime}$
  12607. \end_inset
  12608. end of the transcripts for the Cynomolgus HBA1, HBA2 and HBB genes, with
  12609. two hybridization sites for HBB and 2 sites for HBA (the chosen sites were
  12610. identical in both HBA genes).
  12611. All
  12612. \begin_inset ERT
  12613. status open
  12614. \begin_layout Plain Layout
  12615. \backslash
  12616. glspl*{oligo}
  12617. \end_layout
  12618. \end_inset
  12619. were purchased from Sigma and were entirely composed of 2’O-Me bases with
  12620. a C3 spacer positioned at the
  12621. \begin_inset Formula $3^{\prime}$
  12622. \end_inset
  12623. ends to prevent any polymerase mediated primer extension.
  12624. \end_layout
  12625. \begin_layout Quote
  12626. HBA1/2 site 1: GCCCACUCAGACUUUAUUCAAAG-C3spacer
  12627. \end_layout
  12628. \begin_layout Quote
  12629. HBA1/2 site 2: GGUGCAAGGAGGGGAGGAG-C3spacer
  12630. \end_layout
  12631. \begin_layout Quote
  12632. HBB site 1: AAUGAAAAUAAAUGUUUUUUAUUAG-C3spacer
  12633. \end_layout
  12634. \begin_layout Quote
  12635. HBB site 2: CUCAAGGCCCUUCAUAAUAUCCC-C3spacer
  12636. \end_layout
  12637. \begin_layout Subsection
  12638. RNA-seq Library Preparation
  12639. \end_layout
  12640. \begin_layout Standard
  12641. \begin_inset Flex TODO Note (inline)
  12642. status open
  12643. \begin_layout Plain Layout
  12644. Add protected spaces where appropriate to prevent unwanted line breaks.
  12645. \end_layout
  12646. \end_inset
  12647. \end_layout
  12648. \begin_layout Standard
  12649. Sequencing libraries were prepared with 200
  12650. \begin_inset space ~
  12651. \end_inset
  12652. ng total RNA from each sample.
  12653. Polyadenylated
  12654. \begin_inset Flex Glossary Term
  12655. status open
  12656. \begin_layout Plain Layout
  12657. mRNA
  12658. \end_layout
  12659. \end_inset
  12660. was selected from 200 ng aliquots of cynomolgus blood-derived total RNA
  12661. using Ambion Dynabeads Oligo(dT)25 beads (Invitrogen) following manufacturer’s
  12662. recommended protocol.
  12663. PolyA selected RNA was then combined with 8 pmol of HBA1/2 (site 1), 8
  12664. pmol of HBA1/2 (site 2), 12 pmol of HBB (site 1) and 12 pmol of HBB (site
  12665. 2)
  12666. \begin_inset ERT
  12667. status open
  12668. \begin_layout Plain Layout
  12669. \backslash
  12670. glspl*{oligo}
  12671. \end_layout
  12672. \end_inset
  12673. .
  12674. In addition, 20 pmol of RT primer containing a portion of the Illumina
  12675. adapter sequence (B-oligo-dTV: GAGTTCCTTGGCACCCGAGAATTCCATTTTTTTTTTTTTTTTTTTV)
  12676. and 4 µL of 5X First Strand buffer (250 mM Tris-HCl pH 8.3, 375 mM KCl,
  12677. 15mM MgCl2) were added in a total volume of 15 µL.
  12678. The RNA was fragmented by heating this cocktail for 3 minutes at 95°C and
  12679. then placed on ice.
  12680. This was followed by the addition of 2 µL 0.1 M DTT, 1 µL RNaseOUT, 1 µL
  12681. 10mM dNTPs 10% biotin-16 aminoallyl-2’- dUTP and 10% biotin-16 aminoallyl-2’-
  12682. dCTP (TriLink Biotech, San Diego, CA), 1 µL Superscript II (200U/ µL, Thermo-Fi
  12683. sher).
  12684. A second “unblocked” library was prepared in the same way for each sample
  12685. but replacing the blocking
  12686. \begin_inset ERT
  12687. status open
  12688. \begin_layout Plain Layout
  12689. \backslash
  12690. glspl*{oligo}
  12691. \end_layout
  12692. \end_inset
  12693. with an equivalent volume of water.
  12694. The reaction was carried out at 25°C for 15 minutes and 42°C for 40 minutes,
  12695. followed by incubation at 75°C for 10 minutes to inactivate the reverse
  12696. transcriptase.
  12697. \end_layout
  12698. \begin_layout Standard
  12699. The cDNA/RNA hybrid molecules were purified using 1.8X Ampure XP beads (Agencourt
  12700. ) following supplier’s recommended protocol.
  12701. The cDNA/RNA hybrid was eluted in 25 µL of 10 mM Tris-HCl pH 8.0, and then
  12702. bound to 25 µL of M280 Magnetic Streptavidin beads washed per recommended
  12703. protocol (Thermo-Fisher).
  12704. After 30 minutes of binding, beads were washed one time in 100 µL 0.1N NaOH
  12705. to denature and remove the bound RNA, followed by two 100 µL washes with
  12706. 1X TE buffer.
  12707. \end_layout
  12708. \begin_layout Standard
  12709. Subsequent attachment of the
  12710. \begin_inset Formula $5^{\prime}$
  12711. \end_inset
  12712. Illumina A adapter was performed by on-bead random primer extension of
  12713. the following sequence (A-N8 primer: TTCAGAGTTCTACAGTCCGACGATCNNNNNNNN).
  12714. Briefly, beads were resuspended in a 20 µL reaction containing 5 µM A-N8
  12715. primer, 40mM Tris-HCl pH 7.5, 20mM MgCl2, 50mM NaCl, 0.325U/µL Sequenase
  12716. 2.0 (Affymetrix, Santa Clara, CA), 0.0025U/µL inorganic pyrophosphatase (Affymetr
  12717. ix) and 300 µM each dNTP.
  12718. Reaction was incubated at 22°C for 30 minutes, then beads were washed 2
  12719. times with 1X TE buffer (200µL).
  12720. \end_layout
  12721. \begin_layout Standard
  12722. The magnetic streptavidin beads were resuspended in 34 µL nuclease-free
  12723. water and added directly to a
  12724. \begin_inset Flex Glossary Term
  12725. status open
  12726. \begin_layout Plain Layout
  12727. PCR
  12728. \end_layout
  12729. \end_inset
  12730. \begin_inset CommandInset nomenclature
  12731. LatexCommand nomenclature
  12732. symbol "PCR"
  12733. description "polymerase chain reaction"
  12734. literal "false"
  12735. \end_inset
  12736. tube.
  12737. The two Illumina protocol-specified
  12738. \begin_inset Flex Glossary Term
  12739. status open
  12740. \begin_layout Plain Layout
  12741. PCR
  12742. \end_layout
  12743. \end_inset
  12744. primers were added at 0.53 µM (Illumina TruSeq Universal Primer 1 and Illumina
  12745. TruSeq barcoded
  12746. \begin_inset Flex Glossary Term
  12747. status open
  12748. \begin_layout Plain Layout
  12749. PCR
  12750. \end_layout
  12751. \end_inset
  12752. primer 2), along with 40 µL 2X KAPA HiFi Hotstart ReadyMix (KAPA, Willmington
  12753. MA) and thermocycled as follows: starting with 98°C (2 min-hold); 15 cycles
  12754. of 98°C, 20sec; 60°C, 30sec; 72°C, 30sec; and finished with a 72°C (2 min-hold).
  12755. \end_layout
  12756. \begin_layout Standard
  12757. \begin_inset Flex Glossary Term
  12758. status open
  12759. \begin_layout Plain Layout
  12760. PCR
  12761. \end_layout
  12762. \end_inset
  12763. products were purified with 1X Ampure Beads following manufacturer’s recommende
  12764. d protocol.
  12765. Libraries were then analyzed using the Agilent TapeStation and quantitation
  12766. of desired size range was performed by “smear analysis”.
  12767. Samples were pooled in equimolar batches of 16 samples.
  12768. Pooled libraries were size selected on 2% agarose gels (E-Gel EX Agarose
  12769. Gels; Thermo-Fisher).
  12770. Products were cut between 250 and 350 bp (corresponding to insert sizes
  12771. of 130 to 230 bps).
  12772. Finished library pools were then sequenced on the Illumina NextSeq500 instrumen
  12773. t with 75 base read lengths.
  12774. \end_layout
  12775. \begin_layout Subsection
  12776. Read alignment and counting
  12777. \end_layout
  12778. \begin_layout Standard
  12779. Reads were aligned to the cynomolgus genome using STAR
  12780. \begin_inset CommandInset citation
  12781. LatexCommand cite
  12782. key "Dobin2013,Wilson2013"
  12783. literal "false"
  12784. \end_inset
  12785. .
  12786. Counts of uniquely mapped reads were obtained for every gene in each sample
  12787. with the
  12788. \begin_inset Flex Code
  12789. status open
  12790. \begin_layout Plain Layout
  12791. featureCounts
  12792. \end_layout
  12793. \end_inset
  12794. function from the
  12795. \begin_inset Flex Code
  12796. status open
  12797. \begin_layout Plain Layout
  12798. Rsubread
  12799. \end_layout
  12800. \end_inset
  12801. package, using each of the three possibilities for the
  12802. \begin_inset Flex Code
  12803. status open
  12804. \begin_layout Plain Layout
  12805. strandSpecific
  12806. \end_layout
  12807. \end_inset
  12808. option: sense, antisense, and unstranded
  12809. \begin_inset CommandInset citation
  12810. LatexCommand cite
  12811. key "Liao2014"
  12812. literal "false"
  12813. \end_inset
  12814. .
  12815. A few artifacts in the cynomolgus genome annotation complicated read counting.
  12816. First, no ortholog is annotated for alpha globin in the cynomolgus genome,
  12817. presumably because the human genome has two alpha globin genes with nearly
  12818. identical sequences, making the orthology relationship ambiguous.
  12819. However, two loci in the cynomolgus genome are annotated as “hemoglobin
  12820. subunit alpha-like” (LOC102136192 and LOC102136846).
  12821. LOC102136192 is annotated as a pseudogene while LOC102136846 is annotated
  12822. as protein-coding.
  12823. Our globin reduction protocol was designed to include blocking of these
  12824. two genes.
  12825. Indeed, these two genes have almost the same read counts in each library
  12826. as the properly-annotated HBB gene and much larger counts than any other
  12827. gene in the unblocked libraries, giving confidence that reads derived from
  12828. the real alpha globin are mapping to both genes.
  12829. Thus, reads from both of these loci were counted as alpha globin reads
  12830. in all further analyses.
  12831. The second artifact is a small, uncharacterized non-coding RNA gene (LOC1021365
  12832. 91), which overlaps the HBA-like gene (LOC102136192) on the opposite strand.
  12833. If counting is not performed in stranded mode (or if a non-strand-specific
  12834. sequencing protocol is used), many reads mapping to the globin gene will
  12835. be discarded as ambiguous due to their overlap with this
  12836. \begin_inset Flex Glossary Term
  12837. status open
  12838. \begin_layout Plain Layout
  12839. ncRNA
  12840. \end_layout
  12841. \end_inset
  12842. \begin_inset CommandInset nomenclature
  12843. LatexCommand nomenclature
  12844. symbol "ncRNA"
  12845. description "non-coding RNA"
  12846. literal "false"
  12847. \end_inset
  12848. gene, resulting in significant undercounting of globin reads.
  12849. Therefore, stranded sense counts were used for all further analysis in
  12850. the present study to insure that we accurately accounted for globin transcript
  12851. reduction.
  12852. However, we note that stranded reads are not necessary for
  12853. \begin_inset Flex Glossary Term
  12854. status open
  12855. \begin_layout Plain Layout
  12856. RNA-seq
  12857. \end_layout
  12858. \end_inset
  12859. using our protocol in standard practice.
  12860. \end_layout
  12861. \begin_layout Subsection
  12862. Normalization and Exploratory Data Analysis
  12863. \end_layout
  12864. \begin_layout Standard
  12865. Libraries were normalized by computing scaling factors using the
  12866. \begin_inset Flex Code
  12867. status open
  12868. \begin_layout Plain Layout
  12869. edgeR
  12870. \end_layout
  12871. \end_inset
  12872. package's
  12873. \begin_inset Flex Glossary Term
  12874. status open
  12875. \begin_layout Plain Layout
  12876. TMM
  12877. \end_layout
  12878. \end_inset
  12879. method
  12880. \begin_inset CommandInset citation
  12881. LatexCommand cite
  12882. key "Robinson2010"
  12883. literal "false"
  12884. \end_inset
  12885. .
  12886. \begin_inset Flex Glossary Term (Capital)
  12887. status open
  12888. \begin_layout Plain Layout
  12889. logCPM
  12890. \end_layout
  12891. \end_inset
  12892. values were calculated using the
  12893. \begin_inset Flex Code
  12894. status open
  12895. \begin_layout Plain Layout
  12896. cpm
  12897. \end_layout
  12898. \end_inset
  12899. function in
  12900. \begin_inset Flex Code
  12901. status open
  12902. \begin_layout Plain Layout
  12903. edgeR
  12904. \end_layout
  12905. \end_inset
  12906. for individual samples and
  12907. \begin_inset Flex Code
  12908. status open
  12909. \begin_layout Plain Layout
  12910. aveLogCPM
  12911. \end_layout
  12912. \end_inset
  12913. function for averages across groups of samples, using those functions’
  12914. default prior count values to avoid taking the logarithm of 0.
  12915. Genes were considered “present” if their average normalized
  12916. \begin_inset Flex Glossary Term
  12917. status open
  12918. \begin_layout Plain Layout
  12919. logCPM
  12920. \end_layout
  12921. \end_inset
  12922. values across all libraries were at least
  12923. \begin_inset Formula $-1$
  12924. \end_inset
  12925. .
  12926. Normalizing for gene length was unnecessary because the sequencing protocol
  12927. is
  12928. \begin_inset Formula $3^{\prime}$
  12929. \end_inset
  12930. -biased and hence the expected read count for each gene is related to the
  12931. transcript’s copy number but not its length.
  12932. \end_layout
  12933. \begin_layout Standard
  12934. In order to assess the effect of blocking on reproducibility, Pearson and
  12935. Spearman correlation coefficients were computed between the
  12936. \begin_inset Flex Glossary Term
  12937. status open
  12938. \begin_layout Plain Layout
  12939. logCPM
  12940. \end_layout
  12941. \end_inset
  12942. values for every pair of libraries within the
  12943. \begin_inset Flex Glossary Term
  12944. status open
  12945. \begin_layout Plain Layout
  12946. GB
  12947. \end_layout
  12948. \end_inset
  12949. non-GB groups, and
  12950. \begin_inset Flex Code
  12951. status open
  12952. \begin_layout Plain Layout
  12953. edgeR
  12954. \end_layout
  12955. \end_inset
  12956. 's
  12957. \begin_inset Flex Code
  12958. status open
  12959. \begin_layout Plain Layout
  12960. estimateDisp
  12961. \end_layout
  12962. \end_inset
  12963. function was used to compute
  12964. \begin_inset Flex Glossary Term
  12965. status open
  12966. \begin_layout Plain Layout
  12967. NB
  12968. \end_layout
  12969. \end_inset
  12970. dispersions separately for the two groups
  12971. \begin_inset CommandInset citation
  12972. LatexCommand cite
  12973. key "Chen2014"
  12974. literal "false"
  12975. \end_inset
  12976. .
  12977. \end_layout
  12978. \begin_layout Subsection
  12979. Differential Expression Analysis
  12980. \end_layout
  12981. \begin_layout Standard
  12982. All tests for differential gene expression were performed using
  12983. \begin_inset Flex Code
  12984. status open
  12985. \begin_layout Plain Layout
  12986. edgeR
  12987. \end_layout
  12988. \end_inset
  12989. , by first fitting a
  12990. \begin_inset Flex Glossary Term
  12991. status open
  12992. \begin_layout Plain Layout
  12993. NB
  12994. \end_layout
  12995. \end_inset
  12996. \begin_inset Flex Glossary Term
  12997. status open
  12998. \begin_layout Plain Layout
  12999. GLM
  13000. \end_layout
  13001. \end_inset
  13002. to the counts and normalization factors and then performing a quasi-likelihood
  13003. F-test with robust estimation of outlier gene dispersions
  13004. \begin_inset CommandInset citation
  13005. LatexCommand cite
  13006. key "Lund2012,Phipson2016"
  13007. literal "false"
  13008. \end_inset
  13009. .
  13010. To investigate the effects of
  13011. \begin_inset Flex Glossary Term
  13012. status open
  13013. \begin_layout Plain Layout
  13014. GB
  13015. \end_layout
  13016. \end_inset
  13017. on each gene, an additive model was fit to the full data with coefficients
  13018. for
  13019. \begin_inset Flex Glossary Term
  13020. status open
  13021. \begin_layout Plain Layout
  13022. GB
  13023. \end_layout
  13024. \end_inset
  13025. and Sample ID.
  13026. To test the effect of
  13027. \begin_inset Flex Glossary Term
  13028. status open
  13029. \begin_layout Plain Layout
  13030. GB
  13031. \end_layout
  13032. \end_inset
  13033. on detection of differentially expressed genes, the
  13034. \begin_inset Flex Glossary Term
  13035. status open
  13036. \begin_layout Plain Layout
  13037. GB
  13038. \end_layout
  13039. \end_inset
  13040. samples and non-GB samples were each analyzed independently as follows:
  13041. for each animal with both a pre-transplant and a post-transplant time point
  13042. in the data set, the pre-transplant sample and the earliest post-transplant
  13043. sample were selected, and all others were excluded, yielding a pre-/post-transp
  13044. lant pair of samples for each animal (N=7 animals with paired samples).
  13045. These samples were analyzed for pre-transplant vs.
  13046. post-transplant differential gene expression while controlling for inter-animal
  13047. variation using an additive model with coefficients for transplant and
  13048. animal ID.
  13049. In all analyses, p-values were adjusted using the
  13050. \begin_inset Flex Glossary Term
  13051. status open
  13052. \begin_layout Plain Layout
  13053. BH
  13054. \end_layout
  13055. \end_inset
  13056. procedure for
  13057. \begin_inset Flex Glossary Term
  13058. status open
  13059. \begin_layout Plain Layout
  13060. FDR
  13061. \end_layout
  13062. \end_inset
  13063. control
  13064. \begin_inset CommandInset citation
  13065. LatexCommand cite
  13066. key "Benjamini1995"
  13067. literal "false"
  13068. \end_inset
  13069. .
  13070. \end_layout
  13071. \begin_layout Standard
  13072. \begin_inset Note Note
  13073. status open
  13074. \begin_layout Itemize
  13075. New blood RNA-seq protocol to block reverse transcription of globin genes
  13076. \end_layout
  13077. \begin_layout Itemize
  13078. Blood RNA-seq time course after transplants with/without MSC infusion
  13079. \end_layout
  13080. \end_inset
  13081. \end_layout
  13082. \begin_layout Section
  13083. Results
  13084. \end_layout
  13085. \begin_layout Subsection
  13086. Globin blocking yields a larger and more consistent fraction of useful reads
  13087. \end_layout
  13088. \begin_layout Standard
  13089. \begin_inset ERT
  13090. status open
  13091. \begin_layout Plain Layout
  13092. \backslash
  13093. afterpage{
  13094. \end_layout
  13095. \begin_layout Plain Layout
  13096. \backslash
  13097. begin{landscape}
  13098. \end_layout
  13099. \end_inset
  13100. \end_layout
  13101. \begin_layout Standard
  13102. \begin_inset Float table
  13103. placement p
  13104. wide false
  13105. sideways false
  13106. status open
  13107. \begin_layout Plain Layout
  13108. \align center
  13109. \begin_inset Tabular
  13110. <lyxtabular version="3" rows="4" columns="7">
  13111. <features tabularvalignment="middle">
  13112. <column alignment="center" valignment="top">
  13113. <column alignment="center" valignment="top">
  13114. <column alignment="center" valignment="top">
  13115. <column alignment="center" valignment="top">
  13116. <column alignment="center" valignment="top">
  13117. <column alignment="center" valignment="top">
  13118. <column alignment="center" valignment="top">
  13119. <row>
  13120. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  13121. \begin_inset Text
  13122. \begin_layout Plain Layout
  13123. \end_layout
  13124. \end_inset
  13125. </cell>
  13126. <cell multicolumn="1" alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  13127. \begin_inset Text
  13128. \begin_layout Plain Layout
  13129. \family roman
  13130. \series medium
  13131. \shape up
  13132. \size normal
  13133. \emph off
  13134. \bar no
  13135. \strikeout off
  13136. \xout off
  13137. \uuline off
  13138. \uwave off
  13139. \noun off
  13140. \color none
  13141. Percent of Total Reads
  13142. \end_layout
  13143. \end_inset
  13144. </cell>
  13145. <cell multicolumn="2" alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  13146. \begin_inset Text
  13147. \begin_layout Plain Layout
  13148. \end_layout
  13149. \end_inset
  13150. </cell>
  13151. <cell multicolumn="2" alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  13152. \begin_inset Text
  13153. \begin_layout Plain Layout
  13154. \end_layout
  13155. \end_inset
  13156. </cell>
  13157. <cell multicolumn="2" alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  13158. \begin_inset Text
  13159. \begin_layout Plain Layout
  13160. \end_layout
  13161. \end_inset
  13162. </cell>
  13163. <cell multicolumn="1" alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  13164. \begin_inset Text
  13165. \begin_layout Plain Layout
  13166. \family roman
  13167. \series medium
  13168. \shape up
  13169. \size normal
  13170. \emph off
  13171. \bar no
  13172. \strikeout off
  13173. \xout off
  13174. \uuline off
  13175. \uwave off
  13176. \noun off
  13177. \color none
  13178. Percent of Genic Reads
  13179. \end_layout
  13180. \end_inset
  13181. </cell>
  13182. <cell multicolumn="2" alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  13183. \begin_inset Text
  13184. \begin_layout Plain Layout
  13185. \end_layout
  13186. \end_inset
  13187. </cell>
  13188. </row>
  13189. <row>
  13190. <cell alignment="center" valignment="top" bottomline="true" leftline="true" usebox="none">
  13191. \begin_inset Text
  13192. \begin_layout Plain Layout
  13193. GB
  13194. \end_layout
  13195. \end_inset
  13196. </cell>
  13197. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  13198. \begin_inset Text
  13199. \begin_layout Plain Layout
  13200. \family roman
  13201. \series medium
  13202. \shape up
  13203. \size normal
  13204. \emph off
  13205. \bar no
  13206. \strikeout off
  13207. \xout off
  13208. \uuline off
  13209. \uwave off
  13210. \noun off
  13211. \color none
  13212. Non-globin Reads
  13213. \end_layout
  13214. \end_inset
  13215. </cell>
  13216. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  13217. \begin_inset Text
  13218. \begin_layout Plain Layout
  13219. \family roman
  13220. \series medium
  13221. \shape up
  13222. \size normal
  13223. \emph off
  13224. \bar no
  13225. \strikeout off
  13226. \xout off
  13227. \uuline off
  13228. \uwave off
  13229. \noun off
  13230. \color none
  13231. Globin Reads
  13232. \end_layout
  13233. \end_inset
  13234. </cell>
  13235. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  13236. \begin_inset Text
  13237. \begin_layout Plain Layout
  13238. \family roman
  13239. \series medium
  13240. \shape up
  13241. \size normal
  13242. \emph off
  13243. \bar no
  13244. \strikeout off
  13245. \xout off
  13246. \uuline off
  13247. \uwave off
  13248. \noun off
  13249. \color none
  13250. All Genic Reads
  13251. \end_layout
  13252. \end_inset
  13253. </cell>
  13254. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  13255. \begin_inset Text
  13256. \begin_layout Plain Layout
  13257. \family roman
  13258. \series medium
  13259. \shape up
  13260. \size normal
  13261. \emph off
  13262. \bar no
  13263. \strikeout off
  13264. \xout off
  13265. \uuline off
  13266. \uwave off
  13267. \noun off
  13268. \color none
  13269. All Aligned Reads
  13270. \end_layout
  13271. \end_inset
  13272. </cell>
  13273. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  13274. \begin_inset Text
  13275. \begin_layout Plain Layout
  13276. \family roman
  13277. \series medium
  13278. \shape up
  13279. \size normal
  13280. \emph off
  13281. \bar no
  13282. \strikeout off
  13283. \xout off
  13284. \uuline off
  13285. \uwave off
  13286. \noun off
  13287. \color none
  13288. Non-globin Reads
  13289. \end_layout
  13290. \end_inset
  13291. </cell>
  13292. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  13293. \begin_inset Text
  13294. \begin_layout Plain Layout
  13295. \family roman
  13296. \series medium
  13297. \shape up
  13298. \size normal
  13299. \emph off
  13300. \bar no
  13301. \strikeout off
  13302. \xout off
  13303. \uuline off
  13304. \uwave off
  13305. \noun off
  13306. \color none
  13307. Globin Reads
  13308. \end_layout
  13309. \end_inset
  13310. </cell>
  13311. </row>
  13312. <row>
  13313. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  13314. \begin_inset Text
  13315. \begin_layout Plain Layout
  13316. \family roman
  13317. \series medium
  13318. \shape up
  13319. \size normal
  13320. \emph off
  13321. \bar no
  13322. \strikeout off
  13323. \xout off
  13324. \uuline off
  13325. \uwave off
  13326. \noun off
  13327. \color none
  13328. Yes
  13329. \end_layout
  13330. \end_inset
  13331. </cell>
  13332. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  13333. \begin_inset Text
  13334. \begin_layout Plain Layout
  13335. \family roman
  13336. \series medium
  13337. \shape up
  13338. \size normal
  13339. \emph off
  13340. \bar no
  13341. \strikeout off
  13342. \xout off
  13343. \uuline off
  13344. \uwave off
  13345. \noun off
  13346. \color none
  13347. 50.4% ± 6.82
  13348. \end_layout
  13349. \end_inset
  13350. </cell>
  13351. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  13352. \begin_inset Text
  13353. \begin_layout Plain Layout
  13354. \family roman
  13355. \series medium
  13356. \shape up
  13357. \size normal
  13358. \emph off
  13359. \bar no
  13360. \strikeout off
  13361. \xout off
  13362. \uuline off
  13363. \uwave off
  13364. \noun off
  13365. \color none
  13366. 3.48% ± 2.94
  13367. \end_layout
  13368. \end_inset
  13369. </cell>
  13370. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  13371. \begin_inset Text
  13372. \begin_layout Plain Layout
  13373. \family roman
  13374. \series medium
  13375. \shape up
  13376. \size normal
  13377. \emph off
  13378. \bar no
  13379. \strikeout off
  13380. \xout off
  13381. \uuline off
  13382. \uwave off
  13383. \noun off
  13384. \color none
  13385. 53.9% ± 6.81
  13386. \end_layout
  13387. \end_inset
  13388. </cell>
  13389. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  13390. \begin_inset Text
  13391. \begin_layout Plain Layout
  13392. \family roman
  13393. \series medium
  13394. \shape up
  13395. \size normal
  13396. \emph off
  13397. \bar no
  13398. \strikeout off
  13399. \xout off
  13400. \uuline off
  13401. \uwave off
  13402. \noun off
  13403. \color none
  13404. 89.7% ± 2.40
  13405. \end_layout
  13406. \end_inset
  13407. </cell>
  13408. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  13409. \begin_inset Text
  13410. \begin_layout Plain Layout
  13411. \family roman
  13412. \series medium
  13413. \shape up
  13414. \size normal
  13415. \emph off
  13416. \bar no
  13417. \strikeout off
  13418. \xout off
  13419. \uuline off
  13420. \uwave off
  13421. \noun off
  13422. \color none
  13423. 93.5% ± 5.25
  13424. \end_layout
  13425. \end_inset
  13426. </cell>
  13427. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  13428. \begin_inset Text
  13429. \begin_layout Plain Layout
  13430. \family roman
  13431. \series medium
  13432. \shape up
  13433. \size normal
  13434. \emph off
  13435. \bar no
  13436. \strikeout off
  13437. \xout off
  13438. \uuline off
  13439. \uwave off
  13440. \noun off
  13441. \color none
  13442. 6.49% ± 5.25
  13443. \end_layout
  13444. \end_inset
  13445. </cell>
  13446. </row>
  13447. <row>
  13448. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  13449. \begin_inset Text
  13450. \begin_layout Plain Layout
  13451. \family roman
  13452. \series medium
  13453. \shape up
  13454. \size normal
  13455. \emph off
  13456. \bar no
  13457. \strikeout off
  13458. \xout off
  13459. \uuline off
  13460. \uwave off
  13461. \noun off
  13462. \color none
  13463. No
  13464. \end_layout
  13465. \end_inset
  13466. </cell>
  13467. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  13468. \begin_inset Text
  13469. \begin_layout Plain Layout
  13470. \family roman
  13471. \series medium
  13472. \shape up
  13473. \size normal
  13474. \emph off
  13475. \bar no
  13476. \strikeout off
  13477. \xout off
  13478. \uuline off
  13479. \uwave off
  13480. \noun off
  13481. \color none
  13482. 26.3% ± 8.95
  13483. \end_layout
  13484. \end_inset
  13485. </cell>
  13486. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  13487. \begin_inset Text
  13488. \begin_layout Plain Layout
  13489. \family roman
  13490. \series medium
  13491. \shape up
  13492. \size normal
  13493. \emph off
  13494. \bar no
  13495. \strikeout off
  13496. \xout off
  13497. \uuline off
  13498. \uwave off
  13499. \noun off
  13500. \color none
  13501. 44.6% ± 16.6
  13502. \end_layout
  13503. \end_inset
  13504. </cell>
  13505. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  13506. \begin_inset Text
  13507. \begin_layout Plain Layout
  13508. \family roman
  13509. \series medium
  13510. \shape up
  13511. \size normal
  13512. \emph off
  13513. \bar no
  13514. \strikeout off
  13515. \xout off
  13516. \uuline off
  13517. \uwave off
  13518. \noun off
  13519. \color none
  13520. 70.1% ± 9.38
  13521. \end_layout
  13522. \end_inset
  13523. </cell>
  13524. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  13525. \begin_inset Text
  13526. \begin_layout Plain Layout
  13527. \family roman
  13528. \series medium
  13529. \shape up
  13530. \size normal
  13531. \emph off
  13532. \bar no
  13533. \strikeout off
  13534. \xout off
  13535. \uuline off
  13536. \uwave off
  13537. \noun off
  13538. \color none
  13539. 90.7% ± 5.16
  13540. \end_layout
  13541. \end_inset
  13542. </cell>
  13543. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  13544. \begin_inset Text
  13545. \begin_layout Plain Layout
  13546. \family roman
  13547. \series medium
  13548. \shape up
  13549. \size normal
  13550. \emph off
  13551. \bar no
  13552. \strikeout off
  13553. \xout off
  13554. \uuline off
  13555. \uwave off
  13556. \noun off
  13557. \color none
  13558. 38.8% ± 17.1
  13559. \end_layout
  13560. \end_inset
  13561. </cell>
  13562. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  13563. \begin_inset Text
  13564. \begin_layout Plain Layout
  13565. \family roman
  13566. \series medium
  13567. \shape up
  13568. \size normal
  13569. \emph off
  13570. \bar no
  13571. \strikeout off
  13572. \xout off
  13573. \uuline off
  13574. \uwave off
  13575. \noun off
  13576. \color none
  13577. 61.2% ± 17.1
  13578. \end_layout
  13579. \end_inset
  13580. </cell>
  13581. </row>
  13582. </lyxtabular>
  13583. \end_inset
  13584. \end_layout
  13585. \begin_layout Plain Layout
  13586. \begin_inset Caption Standard
  13587. \begin_layout Plain Layout
  13588. \series bold
  13589. \begin_inset Argument 1
  13590. status collapsed
  13591. \begin_layout Plain Layout
  13592. Fractions of reads mapping to genomic features in GB and non-GB samples.
  13593. \end_layout
  13594. \end_inset
  13595. \begin_inset CommandInset label
  13596. LatexCommand label
  13597. name "tab:Fractions-of-reads"
  13598. \end_inset
  13599. Fractions of reads mapping to genomic features in GB and non-GB samples.
  13600. \series default
  13601. All values are given as mean ± standard deviation.
  13602. \end_layout
  13603. \end_inset
  13604. \end_layout
  13605. \end_inset
  13606. \end_layout
  13607. \begin_layout Standard
  13608. \begin_inset ERT
  13609. status open
  13610. \begin_layout Plain Layout
  13611. \backslash
  13612. end{landscape}
  13613. \end_layout
  13614. \begin_layout Plain Layout
  13615. }
  13616. \end_layout
  13617. \end_inset
  13618. \end_layout
  13619. \begin_layout Standard
  13620. The objective of the present study was to validate a new protocol for deep
  13621. \begin_inset Flex Glossary Term
  13622. status open
  13623. \begin_layout Plain Layout
  13624. RNA-seq
  13625. \end_layout
  13626. \end_inset
  13627. of whole blood drawn into PaxGene tubes from cynomolgus monkeys undergoing
  13628. islet transplantation, with particular focus on minimizing the loss of
  13629. useful sequencing space to uninformative globin reads.
  13630. The details of the analysis with respect to transplant outcomes and the
  13631. impact of mesenchymal stem cell treatment will be reported in a separate
  13632. manuscript (in preparation).
  13633. To focus on the efficacy of our
  13634. \begin_inset Flex Glossary Term
  13635. status open
  13636. \begin_layout Plain Layout
  13637. GB
  13638. \end_layout
  13639. \end_inset
  13640. protocol, 37 blood samples, 16 from pre-transplant and 21 from post-transplant
  13641. time points, were each prepped once with and once without
  13642. \begin_inset Flex Glossary Term
  13643. status open
  13644. \begin_layout Plain Layout
  13645. GB
  13646. \end_layout
  13647. \end_inset
  13648. \begin_inset ERT
  13649. status open
  13650. \begin_layout Plain Layout
  13651. \backslash
  13652. glspl*{oligo}
  13653. \end_layout
  13654. \end_inset
  13655. , and were then sequenced on an Illumina NextSeq500 instrument.
  13656. The number of reads aligning to each gene in the cynomolgus genome was
  13657. counted.
  13658. Table
  13659. \begin_inset CommandInset ref
  13660. LatexCommand ref
  13661. reference "tab:Fractions-of-reads"
  13662. plural "false"
  13663. caps "false"
  13664. noprefix "false"
  13665. \end_inset
  13666. summarizes the distribution of read fractions among the
  13667. \begin_inset Flex Glossary Term
  13668. status open
  13669. \begin_layout Plain Layout
  13670. GB
  13671. \end_layout
  13672. \end_inset
  13673. and non-GB libraries.
  13674. In the libraries with no
  13675. \begin_inset Flex Glossary Term
  13676. status open
  13677. \begin_layout Plain Layout
  13678. GB
  13679. \end_layout
  13680. \end_inset
  13681. , globin reads made up an average of 44.6% of total input reads, while reads
  13682. assigned to all other genes made up an average of 26.3%.
  13683. The remaining reads either aligned to intergenic regions (that include
  13684. long non-coding RNAs) or did not align with any annotated transcripts in
  13685. the current build of the cynomolgus genome.
  13686. In the
  13687. \begin_inset Flex Glossary Term
  13688. status open
  13689. \begin_layout Plain Layout
  13690. GB
  13691. \end_layout
  13692. \end_inset
  13693. libraries, globin reads made up only 3.48% and reads assigned to all other
  13694. genes increased to 50.4%.
  13695. Thus,
  13696. \begin_inset Flex Glossary Term
  13697. status open
  13698. \begin_layout Plain Layout
  13699. GB
  13700. \end_layout
  13701. \end_inset
  13702. resulted in a 92.2% reduction in globin reads and a 91.6% increase in yield
  13703. of useful non-globin reads.
  13704. \end_layout
  13705. \begin_layout Standard
  13706. This reduction is not quite as efficient as the previous analysis showed
  13707. for human samples by DeepSAGE (<0.4% globin reads after globin reduction)
  13708. \begin_inset CommandInset citation
  13709. LatexCommand cite
  13710. key "Mastrokolias2012"
  13711. literal "false"
  13712. \end_inset
  13713. .
  13714. Nonetheless, this degree of globin reduction is sufficient to nearly double
  13715. the yield of useful reads.
  13716. Thus,
  13717. \begin_inset Flex Glossary Term
  13718. status open
  13719. \begin_layout Plain Layout
  13720. GB
  13721. \end_layout
  13722. \end_inset
  13723. cuts the required sequencing effort (and costs) to achieve a target coverage
  13724. depth by almost 50%.
  13725. Consistent with this near doubling of yield, the average difference in
  13726. un-normalized
  13727. \begin_inset Flex Glossary Term
  13728. status open
  13729. \begin_layout Plain Layout
  13730. logCPM
  13731. \end_layout
  13732. \end_inset
  13733. across all genes between the
  13734. \begin_inset Flex Glossary Term
  13735. status open
  13736. \begin_layout Plain Layout
  13737. GB
  13738. \end_layout
  13739. \end_inset
  13740. libraries and non-GB libraries is approximately 1 (mean = 1.01, median =
  13741. 1.08), an overall 2-fold increase.
  13742. Un-normalized values are used here because the
  13743. \begin_inset Flex Glossary Term
  13744. status open
  13745. \begin_layout Plain Layout
  13746. TMM
  13747. \end_layout
  13748. \end_inset
  13749. normalization correctly identifies this 2-fold difference as biologically
  13750. irrelevant and removes it.
  13751. \end_layout
  13752. \begin_layout Standard
  13753. \begin_inset Float figure
  13754. wide false
  13755. sideways false
  13756. status collapsed
  13757. \begin_layout Plain Layout
  13758. \align center
  13759. \begin_inset Graphics
  13760. filename graphics/Globin Paper/figure1 - globin-fractions.pdf
  13761. lyxscale 50
  13762. width 75col%
  13763. \end_inset
  13764. \end_layout
  13765. \begin_layout Plain Layout
  13766. \begin_inset Caption Standard
  13767. \begin_layout Plain Layout
  13768. \series bold
  13769. \begin_inset Argument 1
  13770. status collapsed
  13771. \begin_layout Plain Layout
  13772. Fraction of genic reads in each sample aligned to non-globin genes, with
  13773. and without GB.
  13774. \end_layout
  13775. \end_inset
  13776. \begin_inset CommandInset label
  13777. LatexCommand label
  13778. name "fig:Fraction-of-genic-reads"
  13779. \end_inset
  13780. Fraction of genic reads in each sample aligned to non-globin genes, with
  13781. and without GB.
  13782. \series default
  13783. All reads in each sequencing library were aligned to the cyno genome, and
  13784. the number of reads uniquely aligning to each gene was counted.
  13785. For each sample, counts were summed separately for all globin genes and
  13786. for the remainder of the genes (non-globin genes), and the fraction of
  13787. genic reads aligned to non-globin genes was computed.
  13788. Each point represents an individual sample.
  13789. Gray + signs indicate the means for globin-blocked libraries and unblocked
  13790. libraries.
  13791. The overall distribution for each group is represented as a notched box
  13792. plots.
  13793. Points are randomly spread vertically to avoid excessive overlapping.
  13794. \end_layout
  13795. \end_inset
  13796. \end_layout
  13797. \end_inset
  13798. \end_layout
  13799. \begin_layout Standard
  13800. Another important aspect is that the standard deviations in Table
  13801. \begin_inset CommandInset ref
  13802. LatexCommand ref
  13803. reference "tab:Fractions-of-reads"
  13804. plural "false"
  13805. caps "false"
  13806. noprefix "false"
  13807. \end_inset
  13808. are uniformly smaller in the
  13809. \begin_inset Flex Glossary Term
  13810. status open
  13811. \begin_layout Plain Layout
  13812. GB
  13813. \end_layout
  13814. \end_inset
  13815. samples than the non-GB ones, indicating much greater consistency of yield.
  13816. This is best seen in the percentage of non-globin reads as a fraction of
  13817. total reads aligned to annotated genes (genic reads).
  13818. For the non-GB samples, this measure ranges from 10.9% to 80.9%, while for
  13819. the
  13820. \begin_inset Flex Glossary Term
  13821. status open
  13822. \begin_layout Plain Layout
  13823. GB
  13824. \end_layout
  13825. \end_inset
  13826. samples it ranges from 81.9% to 99.9% (Figure
  13827. \begin_inset CommandInset ref
  13828. LatexCommand ref
  13829. reference "fig:Fraction-of-genic-reads"
  13830. plural "false"
  13831. caps "false"
  13832. noprefix "false"
  13833. \end_inset
  13834. ).
  13835. This means that for applications where it is critical that each sample
  13836. achieve a specified minimum coverage in order to provide useful information,
  13837. it would be necessary to budget up to 10 times the sequencing depth per
  13838. sample without
  13839. \begin_inset Flex Glossary Term
  13840. status open
  13841. \begin_layout Plain Layout
  13842. GB
  13843. \end_layout
  13844. \end_inset
  13845. , even though the average yield improvement for
  13846. \begin_inset Flex Glossary Term
  13847. status open
  13848. \begin_layout Plain Layout
  13849. GB
  13850. \end_layout
  13851. \end_inset
  13852. is only 2-fold, because every sample has a chance of being 90% globin and
  13853. 10% useful reads.
  13854. Hence, the more consistent behavior of
  13855. \begin_inset Flex Glossary Term
  13856. status open
  13857. \begin_layout Plain Layout
  13858. GB
  13859. \end_layout
  13860. \end_inset
  13861. samples makes planning an experiment easier and more efficient because
  13862. it eliminates the need to over-sequence every sample in order to guard
  13863. against the worst case of a high-globin fraction.
  13864. \end_layout
  13865. \begin_layout Subsection
  13866. Globin blocking lowers the noise floor and allows detection of about 2000
  13867. more low-expression genes
  13868. \end_layout
  13869. \begin_layout Standard
  13870. \begin_inset Flex TODO Note (inline)
  13871. status open
  13872. \begin_layout Plain Layout
  13873. Remove redundant titles from figures
  13874. \end_layout
  13875. \end_inset
  13876. \end_layout
  13877. \begin_layout Standard
  13878. \begin_inset Float figure
  13879. wide false
  13880. sideways false
  13881. status collapsed
  13882. \begin_layout Plain Layout
  13883. \align center
  13884. \begin_inset Graphics
  13885. filename graphics/Globin Paper/figure2 - aveLogCPM-colored.pdf
  13886. lyxscale 50
  13887. height 60theight%
  13888. \end_inset
  13889. \end_layout
  13890. \begin_layout Plain Layout
  13891. \begin_inset Caption Standard
  13892. \begin_layout Plain Layout
  13893. \series bold
  13894. \begin_inset Argument 1
  13895. status collapsed
  13896. \begin_layout Plain Layout
  13897. Distributions of average group gene abundances when normalized separately
  13898. or together.
  13899. \end_layout
  13900. \end_inset
  13901. \begin_inset CommandInset label
  13902. LatexCommand label
  13903. name "fig:logcpm-dists"
  13904. \end_inset
  13905. Distributions of average group gene abundances when normalized separately
  13906. or together.
  13907. \series default
  13908. All reads in each sequencing library were aligned to the cyno genome, and
  13909. the number of reads uniquely aligning to each gene was counted.
  13910. Genes with zero counts in all libraries were discarded.
  13911. Libraries were normalized using the TMM method.
  13912. Libraries were split into GB and non-GB groups and the average logCPM was
  13913. computed.
  13914. The distribution of average gene logCPM values was plotted for both groups
  13915. using a kernel density plot to approximate a continuous distribution.
  13916. The GB logCPM distributions are marked in red, non-GB in blue.
  13917. The black vertical line denotes the chosen detection threshold of
  13918. \begin_inset Formula $-1$
  13919. \end_inset
  13920. .
  13921. Top panel: Libraries were split into GB and non-GB groups first and normalized
  13922. separately.
  13923. Bottom panel: Libraries were all normalized together first and then split
  13924. into groups.
  13925. \end_layout
  13926. \end_inset
  13927. \end_layout
  13928. \begin_layout Plain Layout
  13929. \end_layout
  13930. \end_inset
  13931. \end_layout
  13932. \begin_layout Standard
  13933. Since
  13934. \begin_inset Flex Glossary Term
  13935. status open
  13936. \begin_layout Plain Layout
  13937. GB
  13938. \end_layout
  13939. \end_inset
  13940. yields more usable sequencing depth, it should also allow detection of
  13941. more genes at any given threshold.
  13942. When we looked at the distribution of average normalized
  13943. \begin_inset Flex Glossary Term
  13944. status open
  13945. \begin_layout Plain Layout
  13946. logCPM
  13947. \end_layout
  13948. \end_inset
  13949. values across all libraries for genes with at least one read assigned to
  13950. them, we observed the expected bimodal distribution, with a high-abundance
  13951. "signal" peak representing detected genes and a low-abundance "noise" peak
  13952. representing genes whose read count did not rise above the noise floor
  13953. (Figure
  13954. \begin_inset CommandInset ref
  13955. LatexCommand ref
  13956. reference "fig:logcpm-dists"
  13957. plural "false"
  13958. caps "false"
  13959. noprefix "false"
  13960. \end_inset
  13961. ).
  13962. Consistent with the 2-fold increase in raw counts assigned to non-globin
  13963. genes, the signal peak for
  13964. \begin_inset Flex Glossary Term
  13965. status open
  13966. \begin_layout Plain Layout
  13967. GB
  13968. \end_layout
  13969. \end_inset
  13970. samples is shifted to the right relative to the non-GB signal peak.
  13971. When all the samples are normalized together, this difference is normalized
  13972. out, lining up the signal peaks, and this reveals that, as expected, the
  13973. noise floor for the
  13974. \begin_inset Flex Glossary Term
  13975. status open
  13976. \begin_layout Plain Layout
  13977. GB
  13978. \end_layout
  13979. \end_inset
  13980. samples is about 2-fold lower.
  13981. This greater separation between signal and noise peaks in the
  13982. \begin_inset Flex Glossary Term
  13983. status open
  13984. \begin_layout Plain Layout
  13985. GB
  13986. \end_layout
  13987. \end_inset
  13988. samples means that low-expression genes should be more easily detected
  13989. and more precisely quantified than in the non-GB samples.
  13990. \end_layout
  13991. \begin_layout Standard
  13992. \begin_inset Float figure
  13993. wide false
  13994. sideways false
  13995. status collapsed
  13996. \begin_layout Plain Layout
  13997. \align center
  13998. \begin_inset Graphics
  13999. filename graphics/Globin Paper/figure3 - detection.pdf
  14000. lyxscale 50
  14001. width 70col%
  14002. \end_inset
  14003. \end_layout
  14004. \begin_layout Plain Layout
  14005. \begin_inset Caption Standard
  14006. \begin_layout Plain Layout
  14007. \series bold
  14008. \begin_inset Argument 1
  14009. status collapsed
  14010. \begin_layout Plain Layout
  14011. Gene detections as a function of abundance thresholds in GB and non-GB samples.
  14012. \end_layout
  14013. \end_inset
  14014. \begin_inset CommandInset label
  14015. LatexCommand label
  14016. name "fig:Gene-detections"
  14017. \end_inset
  14018. Gene detections as a function of abundance thresholds in GB and non-GB samples.
  14019. \series default
  14020. Average logCPM was computed by separate group normalization as described
  14021. in Figure
  14022. \begin_inset CommandInset ref
  14023. LatexCommand ref
  14024. reference "fig:logcpm-dists"
  14025. plural "false"
  14026. caps "false"
  14027. noprefix "false"
  14028. \end_inset
  14029. for both the GB and non-GB groups, as well as for all samples considered
  14030. as one large group.
  14031. For each every integer threshold from
  14032. \begin_inset Formula $-2$
  14033. \end_inset
  14034. to 3, the number of genes detected at or above that logCPM threshold was
  14035. plotted for each group.
  14036. \end_layout
  14037. \end_inset
  14038. \end_layout
  14039. \begin_layout Plain Layout
  14040. \end_layout
  14041. \end_inset
  14042. \end_layout
  14043. \begin_layout Standard
  14044. Based on these distributions, we selected a detection threshold of
  14045. \begin_inset Formula $-1$
  14046. \end_inset
  14047. , which is approximately the leftmost edge of the trough between the signal
  14048. and noise peaks.
  14049. This represents the most liberal possible detection threshold that doesn't
  14050. call substantial numbers of noise genes as detected.
  14051. Among the full dataset, 13429 genes were detected at this threshold, and
  14052. 22276 were not.
  14053. When considering the
  14054. \begin_inset Flex Glossary Term
  14055. status open
  14056. \begin_layout Plain Layout
  14057. GB
  14058. \end_layout
  14059. \end_inset
  14060. libraries and non-GB libraries separately and re-computing normalization
  14061. factors independently within each group, 14535 genes were detected in the
  14062. \begin_inset Flex Glossary Term
  14063. status open
  14064. \begin_layout Plain Layout
  14065. GB
  14066. \end_layout
  14067. \end_inset
  14068. libraries while only 12460 were detected in the non-GB libraries.
  14069. Thus,
  14070. \begin_inset Flex Glossary Term
  14071. status open
  14072. \begin_layout Plain Layout
  14073. GB
  14074. \end_layout
  14075. \end_inset
  14076. allowed the detection of 2000 extra genes that were buried under the noise
  14077. floor without
  14078. \begin_inset Flex Glossary Term
  14079. status open
  14080. \begin_layout Plain Layout
  14081. GB
  14082. \end_layout
  14083. \end_inset
  14084. .
  14085. This pattern of at least 2000 additional genes detected with
  14086. \begin_inset Flex Glossary Term
  14087. status open
  14088. \begin_layout Plain Layout
  14089. GB
  14090. \end_layout
  14091. \end_inset
  14092. was also consistent across a wide range of possible detection thresholds,
  14093. from -2 to 3 (see Figure
  14094. \begin_inset CommandInset ref
  14095. LatexCommand ref
  14096. reference "fig:Gene-detections"
  14097. plural "false"
  14098. caps "false"
  14099. noprefix "false"
  14100. \end_inset
  14101. ).
  14102. \end_layout
  14103. \begin_layout Subsection
  14104. Globin blocking does not add significant additional noise or decrease sample
  14105. quality
  14106. \end_layout
  14107. \begin_layout Standard
  14108. One potential worry is that the
  14109. \begin_inset Flex Glossary Term
  14110. status open
  14111. \begin_layout Plain Layout
  14112. GB
  14113. \end_layout
  14114. \end_inset
  14115. protocol could perturb the levels of non-globin genes.
  14116. There are two kinds of possible perturbations: systematic and random.
  14117. The former is not a major concern for detection of differential expression,
  14118. since a 2-fold change in every sample has no effect on the relative fold
  14119. change between samples.
  14120. In contrast, random perturbations would increase the noise and obscure
  14121. the signal in the dataset, reducing the capacity to detect differential
  14122. expression.
  14123. \end_layout
  14124. \begin_layout Standard
  14125. \begin_inset Float figure
  14126. wide false
  14127. sideways false
  14128. status collapsed
  14129. \begin_layout Plain Layout
  14130. \align center
  14131. \begin_inset Graphics
  14132. filename graphics/Globin Paper/figure4 - maplot-colored.pdf
  14133. lyxscale 50
  14134. width 60col%
  14135. groupId colwidth
  14136. \end_inset
  14137. \end_layout
  14138. \begin_layout Plain Layout
  14139. \begin_inset Caption Standard
  14140. \begin_layout Plain Layout
  14141. \begin_inset Argument 1
  14142. status collapsed
  14143. \begin_layout Plain Layout
  14144. MA plot showing effects of GB on each gene's abundance.
  14145. \end_layout
  14146. \end_inset
  14147. \begin_inset CommandInset label
  14148. LatexCommand label
  14149. name "fig:MA-plot"
  14150. \end_inset
  14151. \series bold
  14152. MA plot showing effects of GB on each gene's abundance.
  14153. \series default
  14154. All libraries were normalized together as described in Figure
  14155. \begin_inset CommandInset ref
  14156. LatexCommand ref
  14157. reference "fig:logcpm-dists"
  14158. plural "false"
  14159. caps "false"
  14160. noprefix "false"
  14161. \end_inset
  14162. , and genes with an average logCPM below
  14163. \begin_inset Formula $-1$
  14164. \end_inset
  14165. were filtered out.
  14166. Each remaining gene was tested for differential abundance with respect
  14167. to
  14168. \begin_inset Flex Glossary Term (glstext)
  14169. status open
  14170. \begin_layout Plain Layout
  14171. GB
  14172. \end_layout
  14173. \end_inset
  14174. using
  14175. \begin_inset Flex Code
  14176. status open
  14177. \begin_layout Plain Layout
  14178. edgeR
  14179. \end_layout
  14180. \end_inset
  14181. ’s quasi-likelihood F-test, fitting a NB GLM to table of read counts in
  14182. each library.
  14183. For each gene,
  14184. \begin_inset Flex Code
  14185. status open
  14186. \begin_layout Plain Layout
  14187. edgeR
  14188. \end_layout
  14189. \end_inset
  14190. reported average logCPM, logFC, p-value, and BH-adjusted FDR.
  14191. Each gene's logFC was plotted against its logCPM, colored by FDR.
  14192. Red points are significant at ≤10% FDR, and blue are not significant at
  14193. that threshold.
  14194. The alpha and beta globin genes targeted for blocking are marked with large
  14195. triangles, while all other genes are represented as small points.
  14196. \end_layout
  14197. \end_inset
  14198. \end_layout
  14199. \end_inset
  14200. \end_layout
  14201. \begin_layout Standard
  14202. \begin_inset Flex TODO Note (inline)
  14203. status open
  14204. \begin_layout Plain Layout
  14205. Standardize on
  14206. \begin_inset Quotes eld
  14207. \end_inset
  14208. log2
  14209. \begin_inset Quotes erd
  14210. \end_inset
  14211. notation
  14212. \end_layout
  14213. \end_inset
  14214. \end_layout
  14215. \begin_layout Standard
  14216. The data do indeed show small systematic perturbations in gene levels (Figure
  14217. \begin_inset CommandInset ref
  14218. LatexCommand ref
  14219. reference "fig:MA-plot"
  14220. plural "false"
  14221. caps "false"
  14222. noprefix "false"
  14223. \end_inset
  14224. ).
  14225. Other than the 3 designated alpha and beta globin genes, two other genes
  14226. stand out as having especially large negative
  14227. \begin_inset ERT
  14228. status open
  14229. \begin_layout Plain Layout
  14230. \backslash
  14231. glspl*{logFC}
  14232. \end_layout
  14233. \end_inset
  14234. : HBD and LOC1021365.
  14235. HBD, delta globin, is most likely targeted by the blocking
  14236. \begin_inset ERT
  14237. status open
  14238. \begin_layout Plain Layout
  14239. \backslash
  14240. glspl*{oligo}
  14241. \end_layout
  14242. \end_inset
  14243. due to high sequence homology with the other globin genes.
  14244. LOC1021365 is the aforementioned
  14245. \begin_inset Flex Glossary Term
  14246. status open
  14247. \begin_layout Plain Layout
  14248. ncRNA
  14249. \end_layout
  14250. \end_inset
  14251. that is reverse-complementary to one of the alpha-like genes and that would
  14252. be expected to be removed during the
  14253. \begin_inset Flex Glossary Term
  14254. status open
  14255. \begin_layout Plain Layout
  14256. GB
  14257. \end_layout
  14258. \end_inset
  14259. step.
  14260. All other genes appear in a cluster centered vertically at 0, and the vast
  14261. majority of genes in this cluster show an absolute
  14262. \begin_inset Flex Glossary Term
  14263. status open
  14264. \begin_layout Plain Layout
  14265. logFC
  14266. \end_layout
  14267. \end_inset
  14268. of 0.5 or less.
  14269. Nevertheless, many of these small perturbations are still statistically
  14270. significant, indicating that the
  14271. \begin_inset Flex Glossary Term
  14272. status open
  14273. \begin_layout Plain Layout
  14274. GB
  14275. \end_layout
  14276. \end_inset
  14277. \begin_inset ERT
  14278. status open
  14279. \begin_layout Plain Layout
  14280. \backslash
  14281. glspl*{oligo}
  14282. \end_layout
  14283. \end_inset
  14284. likely cause very small but non-zero systematic perturbations in measured
  14285. gene expression levels.
  14286. \end_layout
  14287. \begin_layout Standard
  14288. \begin_inset Float figure
  14289. wide false
  14290. sideways false
  14291. status collapsed
  14292. \begin_layout Plain Layout
  14293. \align center
  14294. \begin_inset Graphics
  14295. filename graphics/Globin Paper/figure5 - corrplot.pdf
  14296. lyxscale 50
  14297. width 70col%
  14298. \end_inset
  14299. \end_layout
  14300. \begin_layout Plain Layout
  14301. \begin_inset Caption Standard
  14302. \begin_layout Plain Layout
  14303. \series bold
  14304. \begin_inset Argument 1
  14305. status collapsed
  14306. \begin_layout Plain Layout
  14307. Comparison of inter-sample gene abundance correlations with and without
  14308. GB.
  14309. \end_layout
  14310. \end_inset
  14311. \begin_inset CommandInset label
  14312. LatexCommand label
  14313. name "fig:gene-abundance-correlations"
  14314. \end_inset
  14315. Comparison of inter-sample gene abundance correlations with and without
  14316. GB.
  14317. \series default
  14318. All libraries were normalized together as described in Figure 2, and genes
  14319. with an average logCPM less than
  14320. \begin_inset Formula $-1$
  14321. \end_inset
  14322. were filtered out.
  14323. Each gene’s logCPM was computed in each library using
  14324. \begin_inset Flex Code
  14325. status open
  14326. \begin_layout Plain Layout
  14327. edgeR
  14328. \end_layout
  14329. \end_inset
  14330. 's
  14331. \begin_inset Flex Code
  14332. status open
  14333. \begin_layout Plain Layout
  14334. cpm
  14335. \end_layout
  14336. \end_inset
  14337. function.
  14338. For each pair of biological samples, the Pearson correlation between those
  14339. samples' GB libraries was plotted against the correlation between the same
  14340. samples’ non-GB libraries.
  14341. Each point represents an unique pair of samples.
  14342. The solid gray line shows a quantile-quantile plot of distribution of GB
  14343. correlations vs.
  14344. that of non-GB correlations.
  14345. The thin dashed line is the identity line, provided for reference.
  14346. \end_layout
  14347. \end_inset
  14348. \end_layout
  14349. \begin_layout Plain Layout
  14350. \end_layout
  14351. \end_inset
  14352. \end_layout
  14353. \begin_layout Standard
  14354. \begin_inset Flex TODO Note (inline)
  14355. status open
  14356. \begin_layout Plain Layout
  14357. Give these numbers the LaTeX math treatment
  14358. \end_layout
  14359. \end_inset
  14360. \end_layout
  14361. \begin_layout Standard
  14362. To evaluate the possibility of
  14363. \begin_inset Flex Glossary Term
  14364. status open
  14365. \begin_layout Plain Layout
  14366. GB
  14367. \end_layout
  14368. \end_inset
  14369. causing random perturbations and reducing sample quality, we computed the
  14370. Pearson correlation between
  14371. \begin_inset Flex Glossary Term
  14372. status open
  14373. \begin_layout Plain Layout
  14374. logCPM
  14375. \end_layout
  14376. \end_inset
  14377. values for every pair of samples with and without
  14378. \begin_inset Flex Glossary Term
  14379. status open
  14380. \begin_layout Plain Layout
  14381. GB
  14382. \end_layout
  14383. \end_inset
  14384. and plotted them against each other (Figure
  14385. \begin_inset CommandInset ref
  14386. LatexCommand ref
  14387. reference "fig:gene-abundance-correlations"
  14388. plural "false"
  14389. caps "false"
  14390. noprefix "false"
  14391. \end_inset
  14392. ).
  14393. The plot indicated that the
  14394. \begin_inset Flex Glossary Term
  14395. status open
  14396. \begin_layout Plain Layout
  14397. GB
  14398. \end_layout
  14399. \end_inset
  14400. libraries have higher sample-to-sample correlations than the non-GB libraries.
  14401. Parametric and nonparametric tests for differences between the correlations
  14402. with and without
  14403. \begin_inset Flex Glossary Term
  14404. status open
  14405. \begin_layout Plain Layout
  14406. GB
  14407. \end_layout
  14408. \end_inset
  14409. both confirmed that this difference was highly significant (2-sided paired
  14410. t-test: t = 37.2, df = 665, P ≪ 2.2e-16; 2-sided Wilcoxon sign-rank test:
  14411. V = 2195, P ≪ 2.2e-16).
  14412. Performing the same tests on the Spearman correlations gave the same conclusion
  14413. (t-test: t = 26.8, df = 665, P ≪ 2.2e-16; sign-rank test: V = 8781, P ≪ 2.2e-16).
  14414. The
  14415. \begin_inset Flex Code
  14416. status open
  14417. \begin_layout Plain Layout
  14418. edgeR
  14419. \end_layout
  14420. \end_inset
  14421. package was used to compute the overall
  14422. \begin_inset Flex Glossary Term
  14423. status open
  14424. \begin_layout Plain Layout
  14425. BCV
  14426. \end_layout
  14427. \end_inset
  14428. for
  14429. \begin_inset Flex Glossary Term
  14430. status open
  14431. \begin_layout Plain Layout
  14432. GB
  14433. \end_layout
  14434. \end_inset
  14435. and non-GB libraries, and found that
  14436. \begin_inset Flex Glossary Term
  14437. status open
  14438. \begin_layout Plain Layout
  14439. GB
  14440. \end_layout
  14441. \end_inset
  14442. resulted in a negligible increase in the
  14443. \begin_inset Flex Glossary Term
  14444. status open
  14445. \begin_layout Plain Layout
  14446. BCV
  14447. \end_layout
  14448. \end_inset
  14449. (0.417 with GB vs.
  14450. 0.400 without).
  14451. The near equality of the
  14452. \begin_inset Flex Glossary Term
  14453. status open
  14454. \begin_layout Plain Layout
  14455. BCV
  14456. \end_layout
  14457. \end_inset
  14458. for both sets indicates that the higher correlations in the GB libraries
  14459. are most likely a result of the increased yield of useful reads, which
  14460. reduces the contribution of Poisson counting uncertainty to the overall
  14461. variance of the
  14462. \begin_inset Flex Glossary Term
  14463. status open
  14464. \begin_layout Plain Layout
  14465. logCPM
  14466. \end_layout
  14467. \end_inset
  14468. values
  14469. \begin_inset CommandInset citation
  14470. LatexCommand cite
  14471. key "McCarthy2012"
  14472. literal "false"
  14473. \end_inset
  14474. .
  14475. This improves the precision of expression measurements and more than offsets
  14476. the negligible increase in
  14477. \begin_inset Flex Glossary Term
  14478. status open
  14479. \begin_layout Plain Layout
  14480. BCV
  14481. \end_layout
  14482. \end_inset
  14483. .
  14484. \end_layout
  14485. \begin_layout Subsection
  14486. More differentially expressed genes are detected with globin blocking
  14487. \end_layout
  14488. \begin_layout Standard
  14489. \begin_inset Float table
  14490. wide false
  14491. sideways false
  14492. status collapsed
  14493. \begin_layout Plain Layout
  14494. \align center
  14495. \begin_inset Tabular
  14496. <lyxtabular version="3" rows="5" columns="5">
  14497. <features tabularvalignment="middle">
  14498. <column alignment="center" valignment="top">
  14499. <column alignment="center" valignment="top">
  14500. <column alignment="center" valignment="top">
  14501. <column alignment="center" valignment="top">
  14502. <column alignment="center" valignment="top">
  14503. <row>
  14504. <cell alignment="center" valignment="top" usebox="none">
  14505. \begin_inset Text
  14506. \begin_layout Plain Layout
  14507. \end_layout
  14508. \end_inset
  14509. </cell>
  14510. <cell alignment="center" valignment="top" usebox="none">
  14511. \begin_inset Text
  14512. \begin_layout Plain Layout
  14513. \end_layout
  14514. \end_inset
  14515. </cell>
  14516. <cell multicolumn="1" alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  14517. \begin_inset Text
  14518. \begin_layout Plain Layout
  14519. \series bold
  14520. No Globin Blocking
  14521. \end_layout
  14522. \end_inset
  14523. </cell>
  14524. <cell multicolumn="2" alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  14525. \begin_inset Text
  14526. \begin_layout Plain Layout
  14527. \end_layout
  14528. \end_inset
  14529. </cell>
  14530. <cell multicolumn="2" alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  14531. \begin_inset Text
  14532. \begin_layout Plain Layout
  14533. \end_layout
  14534. \end_inset
  14535. </cell>
  14536. </row>
  14537. <row>
  14538. <cell alignment="center" valignment="top" usebox="none">
  14539. \begin_inset Text
  14540. \begin_layout Plain Layout
  14541. \end_layout
  14542. \end_inset
  14543. </cell>
  14544. <cell alignment="center" valignment="top" usebox="none">
  14545. \begin_inset Text
  14546. \begin_layout Plain Layout
  14547. \end_layout
  14548. \end_inset
  14549. </cell>
  14550. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  14551. \begin_inset Text
  14552. \begin_layout Plain Layout
  14553. \series bold
  14554. Up
  14555. \end_layout
  14556. \end_inset
  14557. </cell>
  14558. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  14559. \begin_inset Text
  14560. \begin_layout Plain Layout
  14561. \series bold
  14562. NS
  14563. \end_layout
  14564. \end_inset
  14565. </cell>
  14566. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  14567. \begin_inset Text
  14568. \begin_layout Plain Layout
  14569. \series bold
  14570. Down
  14571. \end_layout
  14572. \end_inset
  14573. </cell>
  14574. </row>
  14575. <row>
  14576. <cell multirow="3" alignment="center" valignment="middle" topline="true" bottomline="true" leftline="true" usebox="none">
  14577. \begin_inset Text
  14578. \begin_layout Plain Layout
  14579. \series bold
  14580. Globin-Blocking
  14581. \end_layout
  14582. \end_inset
  14583. </cell>
  14584. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  14585. \begin_inset Text
  14586. \begin_layout Plain Layout
  14587. \series bold
  14588. Up
  14589. \end_layout
  14590. \end_inset
  14591. </cell>
  14592. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  14593. \begin_inset Text
  14594. \begin_layout Plain Layout
  14595. \family roman
  14596. \series medium
  14597. \shape up
  14598. \size normal
  14599. \emph off
  14600. \bar no
  14601. \strikeout off
  14602. \xout off
  14603. \uuline off
  14604. \uwave off
  14605. \noun off
  14606. \color none
  14607. 231
  14608. \end_layout
  14609. \end_inset
  14610. </cell>
  14611. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  14612. \begin_inset Text
  14613. \begin_layout Plain Layout
  14614. \family roman
  14615. \series medium
  14616. \shape up
  14617. \size normal
  14618. \emph off
  14619. \bar no
  14620. \strikeout off
  14621. \xout off
  14622. \uuline off
  14623. \uwave off
  14624. \noun off
  14625. \color none
  14626. 515
  14627. \end_layout
  14628. \end_inset
  14629. </cell>
  14630. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  14631. \begin_inset Text
  14632. \begin_layout Plain Layout
  14633. \family roman
  14634. \series medium
  14635. \shape up
  14636. \size normal
  14637. \emph off
  14638. \bar no
  14639. \strikeout off
  14640. \xout off
  14641. \uuline off
  14642. \uwave off
  14643. \noun off
  14644. \color none
  14645. 2
  14646. \end_layout
  14647. \end_inset
  14648. </cell>
  14649. </row>
  14650. <row>
  14651. <cell multirow="4" alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  14652. \begin_inset Text
  14653. \begin_layout Plain Layout
  14654. \end_layout
  14655. \end_inset
  14656. </cell>
  14657. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  14658. \begin_inset Text
  14659. \begin_layout Plain Layout
  14660. \series bold
  14661. NS
  14662. \end_layout
  14663. \end_inset
  14664. </cell>
  14665. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  14666. \begin_inset Text
  14667. \begin_layout Plain Layout
  14668. \family roman
  14669. \series medium
  14670. \shape up
  14671. \size normal
  14672. \emph off
  14673. \bar no
  14674. \strikeout off
  14675. \xout off
  14676. \uuline off
  14677. \uwave off
  14678. \noun off
  14679. \color none
  14680. 160
  14681. \end_layout
  14682. \end_inset
  14683. </cell>
  14684. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  14685. \begin_inset Text
  14686. \begin_layout Plain Layout
  14687. \family roman
  14688. \series medium
  14689. \shape up
  14690. \size normal
  14691. \emph off
  14692. \bar no
  14693. \strikeout off
  14694. \xout off
  14695. \uuline off
  14696. \uwave off
  14697. \noun off
  14698. \color none
  14699. 11235
  14700. \end_layout
  14701. \end_inset
  14702. </cell>
  14703. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  14704. \begin_inset Text
  14705. \begin_layout Plain Layout
  14706. \family roman
  14707. \series medium
  14708. \shape up
  14709. \size normal
  14710. \emph off
  14711. \bar no
  14712. \strikeout off
  14713. \xout off
  14714. \uuline off
  14715. \uwave off
  14716. \noun off
  14717. \color none
  14718. 136
  14719. \end_layout
  14720. \end_inset
  14721. </cell>
  14722. </row>
  14723. <row>
  14724. <cell multirow="4" alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  14725. \begin_inset Text
  14726. \begin_layout Plain Layout
  14727. \end_layout
  14728. \end_inset
  14729. </cell>
  14730. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  14731. \begin_inset Text
  14732. \begin_layout Plain Layout
  14733. \series bold
  14734. Down
  14735. \end_layout
  14736. \end_inset
  14737. </cell>
  14738. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  14739. \begin_inset Text
  14740. \begin_layout Plain Layout
  14741. \family roman
  14742. \series medium
  14743. \shape up
  14744. \size normal
  14745. \emph off
  14746. \bar no
  14747. \strikeout off
  14748. \xout off
  14749. \uuline off
  14750. \uwave off
  14751. \noun off
  14752. \color none
  14753. 0
  14754. \end_layout
  14755. \end_inset
  14756. </cell>
  14757. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  14758. \begin_inset Text
  14759. \begin_layout Plain Layout
  14760. \family roman
  14761. \series medium
  14762. \shape up
  14763. \size normal
  14764. \emph off
  14765. \bar no
  14766. \strikeout off
  14767. \xout off
  14768. \uuline off
  14769. \uwave off
  14770. \noun off
  14771. \color none
  14772. 548
  14773. \end_layout
  14774. \end_inset
  14775. </cell>
  14776. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  14777. \begin_inset Text
  14778. \begin_layout Plain Layout
  14779. \family roman
  14780. \series medium
  14781. \shape up
  14782. \size normal
  14783. \emph off
  14784. \bar no
  14785. \strikeout off
  14786. \xout off
  14787. \uuline off
  14788. \uwave off
  14789. \noun off
  14790. \color none
  14791. 127
  14792. \end_layout
  14793. \end_inset
  14794. </cell>
  14795. </row>
  14796. </lyxtabular>
  14797. \end_inset
  14798. \end_layout
  14799. \begin_layout Plain Layout
  14800. \begin_inset Caption Standard
  14801. \begin_layout Plain Layout
  14802. \series bold
  14803. \begin_inset Argument 1
  14804. status open
  14805. \begin_layout Plain Layout
  14806. Comparison of significantly differentially expressed genes with and without
  14807. globin blocking.
  14808. \end_layout
  14809. \end_inset
  14810. \begin_inset CommandInset label
  14811. LatexCommand label
  14812. name "tab:Comparison-of-significant"
  14813. \end_inset
  14814. Comparison of significantly differentially expressed genes with and without
  14815. globin blocking.
  14816. \series default
  14817. Up, Down: Genes significantly up/down-regulated in post-transplant samples
  14818. relative to pre-transplant samples, with a false discovery rate of 10%
  14819. or less.
  14820. NS: Non-significant genes (false discovery rate greater than 10%).
  14821. \end_layout
  14822. \end_inset
  14823. \end_layout
  14824. \begin_layout Plain Layout
  14825. \end_layout
  14826. \end_inset
  14827. \end_layout
  14828. \begin_layout Standard
  14829. To compare performance on differential gene expression tests, we took subsets
  14830. of both the
  14831. \begin_inset Flex Glossary Term
  14832. status open
  14833. \begin_layout Plain Layout
  14834. GB
  14835. \end_layout
  14836. \end_inset
  14837. and non-GB libraries with exactly one pre-transplant and one post-transplant
  14838. sample for each animal that had paired samples available for analysis (N=7
  14839. animals, N=14 samples in each subset).
  14840. The same test for pre- vs.
  14841. post-transplant differential gene expression was performed on the same
  14842. 7 pairs of samples from
  14843. \begin_inset Flex Glossary Term
  14844. status open
  14845. \begin_layout Plain Layout
  14846. GB
  14847. \end_layout
  14848. \end_inset
  14849. libraries and non-GB libraries, in each case using an
  14850. \begin_inset Flex Glossary Term
  14851. status open
  14852. \begin_layout Plain Layout
  14853. FDR
  14854. \end_layout
  14855. \end_inset
  14856. of 10% as the threshold of significance.
  14857. Out of 12954 genes that passed the detection threshold in both subsets,
  14858. 358 were called significantly differentially expressed in the same direction
  14859. in both sets; 1063 were differentially expressed in the
  14860. \begin_inset Flex Glossary Term
  14861. status open
  14862. \begin_layout Plain Layout
  14863. GB
  14864. \end_layout
  14865. \end_inset
  14866. set only; 296 were differentially expressed in the non-GB set only; 2 genes
  14867. were called significantly up in the
  14868. \begin_inset Flex Glossary Term
  14869. status open
  14870. \begin_layout Plain Layout
  14871. GB
  14872. \end_layout
  14873. \end_inset
  14874. set but significantly down in the non-GB set; and the remaining 11235 were
  14875. not called differentially expressed in either set.
  14876. These data are summarized in Table
  14877. \begin_inset CommandInset ref
  14878. LatexCommand ref
  14879. reference "tab:Comparison-of-significant"
  14880. plural "false"
  14881. caps "false"
  14882. noprefix "false"
  14883. \end_inset
  14884. .
  14885. The differences in
  14886. \begin_inset Flex Glossary Term
  14887. status open
  14888. \begin_layout Plain Layout
  14889. BCV
  14890. \end_layout
  14891. \end_inset
  14892. calculated by
  14893. \begin_inset Flex Code
  14894. status open
  14895. \begin_layout Plain Layout
  14896. edgeR
  14897. \end_layout
  14898. \end_inset
  14899. for these subsets of samples were negligible (
  14900. \begin_inset Formula $\textrm{BCV}=0.302$
  14901. \end_inset
  14902. for
  14903. \begin_inset Flex Glossary Term
  14904. status open
  14905. \begin_layout Plain Layout
  14906. GB
  14907. \end_layout
  14908. \end_inset
  14909. and 0.297 for non-GB).
  14910. \end_layout
  14911. \begin_layout Standard
  14912. The key point is that the
  14913. \begin_inset Flex Glossary Term
  14914. status open
  14915. \begin_layout Plain Layout
  14916. GB
  14917. \end_layout
  14918. \end_inset
  14919. data results in substantially more differentially expressed calls than
  14920. the non-GB data.
  14921. Since there is no gold standard for this dataset, it is impossible to be
  14922. certain whether this is due to under-calling of differential expression
  14923. in the non-GB samples or over-calling in the
  14924. \begin_inset Flex Glossary Term
  14925. status open
  14926. \begin_layout Plain Layout
  14927. GB
  14928. \end_layout
  14929. \end_inset
  14930. samples.
  14931. However, given that both datasets are derived from the same biological
  14932. samples and have nearly equal
  14933. \begin_inset ERT
  14934. status collapsed
  14935. \begin_layout Plain Layout
  14936. \backslash
  14937. glspl*{BCV}
  14938. \end_layout
  14939. \end_inset
  14940. , it is more likely that the larger number of DE calls in the
  14941. \begin_inset Flex Glossary Term
  14942. status open
  14943. \begin_layout Plain Layout
  14944. GB
  14945. \end_layout
  14946. \end_inset
  14947. samples are genuine detections that were enabled by the higher sequencing
  14948. depth and measurement precision of the
  14949. \begin_inset Flex Glossary Term
  14950. status open
  14951. \begin_layout Plain Layout
  14952. GB
  14953. \end_layout
  14954. \end_inset
  14955. samples.
  14956. Note that the same set of genes was considered in both subsets, so the
  14957. larger number of differentially expressed gene calls in the
  14958. \begin_inset Flex Glossary Term
  14959. status open
  14960. \begin_layout Plain Layout
  14961. GB
  14962. \end_layout
  14963. \end_inset
  14964. data set reflects a greater sensitivity to detect significant differential
  14965. gene expression and not simply the larger total number of detected genes
  14966. in
  14967. \begin_inset Flex Glossary Term
  14968. status open
  14969. \begin_layout Plain Layout
  14970. GB
  14971. \end_layout
  14972. \end_inset
  14973. samples described earlier.
  14974. \end_layout
  14975. \begin_layout Section
  14976. Discussion
  14977. \end_layout
  14978. \begin_layout Standard
  14979. The original experience with whole blood gene expression profiling on DNA
  14980. microarrays demonstrated that the high concentration of globin transcripts
  14981. reduced the sensitivity to detect genes with relatively low expression
  14982. levels, in effect, significantly reducing the sensitivity.
  14983. To address this limitation, commercial protocols for globin reduction were
  14984. developed based on strategies to block globin transcript amplification
  14985. during labeling or physically removing globin transcripts by affinity bead
  14986. methods
  14987. \begin_inset CommandInset citation
  14988. LatexCommand cite
  14989. key "Winn2010"
  14990. literal "false"
  14991. \end_inset
  14992. .
  14993. More recently, using the latest generation of labeling protocols and arrays,
  14994. it was determined that globin reduction was no longer necessary to obtain
  14995. sufficient sensitivity to detect differential transcript expression
  14996. \begin_inset CommandInset citation
  14997. LatexCommand cite
  14998. key "NuGEN2010"
  14999. literal "false"
  15000. \end_inset
  15001. .
  15002. However, we are not aware of any publications using these currently available
  15003. protocols the with latest generation of microarrays that actually compare
  15004. the detection sensitivity with and without globin reduction.
  15005. However, in practice this has now been adopted generally primarily driven
  15006. by concerns for cost control.
  15007. The main objective of our work was to directly test the impact of globin
  15008. gene transcripts and a new
  15009. \begin_inset Flex Glossary Term
  15010. status open
  15011. \begin_layout Plain Layout
  15012. GB
  15013. \end_layout
  15014. \end_inset
  15015. protocol for application to the newest generation of differential gene
  15016. expression profiling determined using next generation sequencing.
  15017. \end_layout
  15018. \begin_layout Standard
  15019. The challenge of doing global gene expression profiling in cynomolgus monkeys
  15020. is that the current available arrays were never designed to comprehensively
  15021. cover this genome and have not been updated since the first assemblies
  15022. of the cynomolgus genome were published.
  15023. Therefore, we determined that the best strategy for peripheral blood profiling
  15024. was to do deep
  15025. \begin_inset Flex Glossary Term
  15026. status open
  15027. \begin_layout Plain Layout
  15028. RNA-seq
  15029. \end_layout
  15030. \end_inset
  15031. and inform the workflow using the latest available genome assembly and
  15032. annotation
  15033. \begin_inset CommandInset citation
  15034. LatexCommand cite
  15035. key "Wilson2013"
  15036. literal "false"
  15037. \end_inset
  15038. .
  15039. However, it was not immediately clear whether globin reduction was necessary
  15040. for
  15041. \begin_inset Flex Glossary Term
  15042. status open
  15043. \begin_layout Plain Layout
  15044. RNA-seq
  15045. \end_layout
  15046. \end_inset
  15047. or how much improvement in efficiency or sensitivity to detect differential
  15048. gene expression would be achieved for the added cost and work.
  15049. \end_layout
  15050. \begin_layout Standard
  15051. We only found one report that demonstrated that globin reduction significantly
  15052. improved the effective read yields for sequencing of human peripheral blood
  15053. cell RNA using a DeepSAGE protocol
  15054. \begin_inset CommandInset citation
  15055. LatexCommand cite
  15056. key "Mastrokolias2012"
  15057. literal "false"
  15058. \end_inset
  15059. .
  15060. The DeepSAGE method involves two different restriction enzymes that purify
  15061. and then tag small fragments of transcripts at specific locations and thus
  15062. significantly reduces the complexity of the transcriptome.
  15063. Therefore, we could not determine how DeepSAGE results would translate
  15064. to the common strategy in the field for assaying the entire transcript
  15065. population by whole-transcriptome
  15066. \begin_inset Formula $3^{\prime}$
  15067. \end_inset
  15068. -end
  15069. \begin_inset Flex Glossary Term
  15070. status open
  15071. \begin_layout Plain Layout
  15072. RNA-seq
  15073. \end_layout
  15074. \end_inset
  15075. .
  15076. Furthermore, if globin reduction is necessary, we also needed a globin
  15077. reduction method specific to cynomolgus globin sequences that would work
  15078. an organism for which no kit is available off the shelf.
  15079. \end_layout
  15080. \begin_layout Standard
  15081. As mentioned above, the addition of
  15082. \begin_inset Flex Glossary Term
  15083. status open
  15084. \begin_layout Plain Layout
  15085. GB
  15086. \end_layout
  15087. \end_inset
  15088. \begin_inset ERT
  15089. status open
  15090. \begin_layout Plain Layout
  15091. \backslash
  15092. glspl*{oligo}
  15093. \end_layout
  15094. \end_inset
  15095. has a very small impact on measured expression levels of gene expression.
  15096. However, this is a non-issue for the purposes of differential expression
  15097. testing, since a systematic change in a gene in all samples does not affect
  15098. relative expression levels between samples.
  15099. However, we must acknowledge that simple comparisons of gene expression
  15100. data obtained by
  15101. \begin_inset Flex Glossary Term
  15102. status open
  15103. \begin_layout Plain Layout
  15104. GB
  15105. \end_layout
  15106. \end_inset
  15107. and non-GB protocols are not possible without additional normalization.
  15108. \end_layout
  15109. \begin_layout Standard
  15110. More importantly,
  15111. \begin_inset Flex Glossary Term
  15112. status open
  15113. \begin_layout Plain Layout
  15114. GB
  15115. \end_layout
  15116. \end_inset
  15117. not only nearly doubles the yield of usable reads, it also increases inter-samp
  15118. le correlation and sensitivity to detect differential gene expression relative
  15119. to the same set of samples profiled without blocking.
  15120. In addition,
  15121. \begin_inset Flex Glossary Term
  15122. status open
  15123. \begin_layout Plain Layout
  15124. GB
  15125. \end_layout
  15126. \end_inset
  15127. does not add a significant amount of random noise to the data.
  15128. Globin blocking thus represents a cost-effective way to squeeze more data
  15129. and statistical power out of the same blood samples and the same amount
  15130. of sequencing.
  15131. In conclusion, globin reduction greatly increases the yield of useful
  15132. \begin_inset Flex Glossary Term
  15133. status open
  15134. \begin_layout Plain Layout
  15135. RNA-seq
  15136. \end_layout
  15137. \end_inset
  15138. reads mapping to the rest of the genome, with minimal perturbations in
  15139. the relative levels of non-globin genes.
  15140. Based on these results, globin transcript reduction using sequence-specific,
  15141. complementary blocking
  15142. \begin_inset ERT
  15143. status open
  15144. \begin_layout Plain Layout
  15145. \backslash
  15146. glspl*{oligo}
  15147. \end_layout
  15148. \end_inset
  15149. is recommended for all deep
  15150. \begin_inset Flex Glossary Term
  15151. status open
  15152. \begin_layout Plain Layout
  15153. RNA-seq
  15154. \end_layout
  15155. \end_inset
  15156. of cynomolgus and other nonhuman primate blood samples.
  15157. \end_layout
  15158. \begin_layout Section
  15159. Future Directions
  15160. \end_layout
  15161. \begin_layout Standard
  15162. One drawback of the
  15163. \begin_inset Flex Glossary Term
  15164. status open
  15165. \begin_layout Plain Layout
  15166. GB
  15167. \end_layout
  15168. \end_inset
  15169. method presented in this analysis is a poor yield of genic reads, only
  15170. around 50%.
  15171. In a separate experiment, the reagent mixture was modified so as to address
  15172. this drawback, resulting in a method that produces an even better reduction
  15173. in globin reads without reducing the overall fraction of genic reads.
  15174. However, the data showing this improvement consists of only a few test
  15175. samples, so the larger data set analyzed above was chosen in order to demonstra
  15176. te the effectiveness of the method in reducing globin reads while preserving
  15177. the biological signal.
  15178. \end_layout
  15179. \begin_layout Standard
  15180. The motivation for developing a fast practical way to enrich for non-globin
  15181. reads in cyno blood samples was to enable a large-scale
  15182. \begin_inset Flex Glossary Term
  15183. status open
  15184. \begin_layout Plain Layout
  15185. RNA-seq
  15186. \end_layout
  15187. \end_inset
  15188. experiment investigating the effects of mesenchymal stem cell infusion
  15189. on blood gene expression in cynomologus transplant recipients in a time
  15190. course after transplantation.
  15191. With the
  15192. \begin_inset Flex Glossary Term
  15193. status open
  15194. \begin_layout Plain Layout
  15195. GB
  15196. \end_layout
  15197. \end_inset
  15198. method in place, the way is now clear for this experiment to proceed.
  15199. \end_layout
  15200. \begin_layout Chapter
  15201. Future Directions
  15202. \end_layout
  15203. \begin_layout Standard
  15204. \begin_inset Flex TODO Note (inline)
  15205. status open
  15206. \begin_layout Plain Layout
  15207. If there are any chapter-independent future directions, put them here.
  15208. Otherwise, delete this section.
  15209. \end_layout
  15210. \end_inset
  15211. \end_layout
  15212. \begin_layout Chapter
  15213. Closing remarks
  15214. \end_layout
  15215. \begin_layout Standard
  15216. \begin_inset ERT
  15217. status collapsed
  15218. \begin_layout Plain Layout
  15219. % Use "References" as the title of the Bibliography
  15220. \end_layout
  15221. \begin_layout Plain Layout
  15222. \backslash
  15223. renewcommand{
  15224. \backslash
  15225. bibname}{References}
  15226. \end_layout
  15227. \end_inset
  15228. \end_layout
  15229. \begin_layout Standard
  15230. \begin_inset CommandInset bibtex
  15231. LatexCommand bibtex
  15232. btprint "btPrintCited"
  15233. bibfiles "code-refs,refs-PROCESSED"
  15234. options "bibtotoc,unsrt"
  15235. \end_inset
  15236. \end_layout
  15237. \begin_layout Standard
  15238. \begin_inset Flex TODO Note (inline)
  15239. status open
  15240. \begin_layout Plain Layout
  15241. Check bib entry formatting & sort order
  15242. \end_layout
  15243. \end_inset
  15244. \end_layout
  15245. \begin_layout Standard
  15246. \begin_inset Flex TODO Note (inline)
  15247. status open
  15248. \begin_layout Plain Layout
  15249. Check in-text citation format.
  15250. Probably don't just want [1], [2], etc.
  15251. \end_layout
  15252. \end_inset
  15253. \end_layout
  15254. \end_body
  15255. \end_document