thesis.lyx 341 KB

12345678910111213141516171819202122232425262728293031323334353637383940414243444546474849505152535455565758596061626364656667686970717273747576777879808182838485868788899091929394959697989910010110210310410510610710810911011111211311411511611711811912012112212312412512612712812913013113213313413513613713813914014114214314414514614714814915015115215315415515615715815916016116216316416516616716816917017117217317417517617717817918018118218318418518618718818919019119219319419519619719819920020120220320420520620720820921021121221321421521621721821922022122222322422522622722822923023123223323423523623723823924024124224324424524624724824925025125225325425525625725825926026126226326426526626726826927027127227327427527627727827928028128228328428528628728828929029129229329429529629729829930030130230330430530630730830931031131231331431531631731831932032132232332432532632732832933033133233333433533633733833934034134234334434534634734834935035135235335435535635735835936036136236336436536636736836937037137237337437537637737837938038138238338438538638738838939039139239339439539639739839940040140240340440540640740840941041141241341441541641741841942042142242342442542642742842943043143243343443543643743843944044144244344444544644744844945045145245345445545645745845946046146246346446546646746846947047147247347447547647747847948048148248348448548648748848949049149249349449549649749849950050150250350450550650750850951051151251351451551651751851952052152252352452552652752852953053153253353453553653753853954054154254354454554654754854955055155255355455555655755855956056156256356456556656756856957057157257357457557657757857958058158258358458558658758858959059159259359459559659759859960060160260360460560660760860961061161261361461561661761861962062162262362462562662762862963063163263363463563663763863964064164264364464564664764864965065165265365465565665765865966066166266366466566666766866967067167267367467567667767867968068168268368468568668768868969069169269369469569669769869970070170270370470570670770870971071171271371471571671771871972072172272372472572672772872973073173273373473573673773873974074174274374474574674774874975075175275375475575675775875976076176276376476576676776876977077177277377477577677777877978078178278378478578678778878979079179279379479579679779879980080180280380480580680780880981081181281381481581681781881982082182282382482582682782882983083183283383483583683783883984084184284384484584684784884985085185285385485585685785885986086186286386486586686786886987087187287387487587687787887988088188288388488588688788888989089189289389489589689789889990090190290390490590690790890991091191291391491591691791891992092192292392492592692792892993093193293393493593693793893994094194294394494594694794894995095195295395495595695795895996096196296396496596696796896997097197297397497597697797897998098198298398498598698798898999099199299399499599699799899910001001100210031004100510061007100810091010101110121013101410151016101710181019102010211022102310241025102610271028102910301031103210331034103510361037103810391040104110421043104410451046104710481049105010511052105310541055105610571058105910601061106210631064106510661067106810691070107110721073107410751076107710781079108010811082108310841085108610871088108910901091109210931094109510961097109810991100110111021103110411051106110711081109111011111112111311141115111611171118111911201121112211231124112511261127112811291130113111321133113411351136113711381139114011411142114311441145114611471148114911501151115211531154115511561157115811591160116111621163116411651166116711681169117011711172117311741175117611771178117911801181118211831184118511861187118811891190119111921193119411951196119711981199120012011202120312041205120612071208120912101211121212131214121512161217121812191220122112221223122412251226122712281229123012311232123312341235123612371238123912401241124212431244124512461247124812491250125112521253125412551256125712581259126012611262126312641265126612671268126912701271127212731274127512761277127812791280128112821283128412851286128712881289129012911292129312941295129612971298129913001301130213031304130513061307130813091310131113121313131413151316131713181319132013211322132313241325132613271328132913301331133213331334133513361337133813391340134113421343134413451346134713481349135013511352135313541355135613571358135913601361136213631364136513661367136813691370137113721373137413751376137713781379138013811382138313841385138613871388138913901391139213931394139513961397139813991400140114021403140414051406140714081409141014111412141314141415141614171418141914201421142214231424142514261427142814291430143114321433143414351436143714381439144014411442144314441445144614471448144914501451145214531454145514561457145814591460146114621463146414651466146714681469147014711472147314741475147614771478147914801481148214831484148514861487148814891490149114921493149414951496149714981499150015011502150315041505150615071508150915101511151215131514151515161517151815191520152115221523152415251526152715281529153015311532153315341535153615371538153915401541154215431544154515461547154815491550155115521553155415551556155715581559156015611562156315641565156615671568156915701571157215731574157515761577157815791580158115821583158415851586158715881589159015911592159315941595159615971598159916001601160216031604160516061607160816091610161116121613161416151616161716181619162016211622162316241625162616271628162916301631163216331634163516361637163816391640164116421643164416451646164716481649165016511652165316541655165616571658165916601661166216631664166516661667166816691670167116721673167416751676167716781679168016811682168316841685168616871688168916901691169216931694169516961697169816991700170117021703170417051706170717081709171017111712171317141715171617171718171917201721172217231724172517261727172817291730173117321733173417351736173717381739174017411742174317441745174617471748174917501751175217531754175517561757175817591760176117621763176417651766176717681769177017711772177317741775177617771778177917801781178217831784178517861787178817891790179117921793179417951796179717981799180018011802180318041805180618071808180918101811181218131814181518161817181818191820182118221823182418251826182718281829183018311832183318341835183618371838183918401841184218431844184518461847184818491850185118521853185418551856185718581859186018611862186318641865186618671868186918701871187218731874187518761877187818791880188118821883188418851886188718881889189018911892189318941895189618971898189919001901190219031904190519061907190819091910191119121913191419151916191719181919192019211922192319241925192619271928192919301931193219331934193519361937193819391940194119421943194419451946194719481949195019511952195319541955195619571958195919601961196219631964196519661967196819691970197119721973197419751976197719781979198019811982198319841985198619871988198919901991199219931994199519961997199819992000200120022003200420052006200720082009201020112012201320142015201620172018201920202021202220232024202520262027202820292030203120322033203420352036203720382039204020412042204320442045204620472048204920502051205220532054205520562057205820592060206120622063206420652066206720682069207020712072207320742075207620772078207920802081208220832084208520862087208820892090209120922093209420952096209720982099210021012102210321042105210621072108210921102111211221132114211521162117211821192120212121222123212421252126212721282129213021312132213321342135213621372138213921402141214221432144214521462147214821492150215121522153215421552156215721582159216021612162216321642165216621672168216921702171217221732174217521762177217821792180218121822183218421852186218721882189219021912192219321942195219621972198219922002201220222032204220522062207220822092210221122122213221422152216221722182219222022212222222322242225222622272228222922302231223222332234223522362237223822392240224122422243224422452246224722482249225022512252225322542255225622572258225922602261226222632264226522662267226822692270227122722273227422752276227722782279228022812282228322842285228622872288228922902291229222932294229522962297229822992300230123022303230423052306230723082309231023112312231323142315231623172318231923202321232223232324232523262327232823292330233123322333233423352336233723382339234023412342234323442345234623472348234923502351235223532354235523562357235823592360236123622363236423652366236723682369237023712372237323742375237623772378237923802381238223832384238523862387238823892390239123922393239423952396239723982399240024012402240324042405240624072408240924102411241224132414241524162417241824192420242124222423242424252426242724282429243024312432243324342435243624372438243924402441244224432444244524462447244824492450245124522453245424552456245724582459246024612462246324642465246624672468246924702471247224732474247524762477247824792480248124822483248424852486248724882489249024912492249324942495249624972498249925002501250225032504250525062507250825092510251125122513251425152516251725182519252025212522252325242525252625272528252925302531253225332534253525362537253825392540254125422543254425452546254725482549255025512552255325542555255625572558255925602561256225632564256525662567256825692570257125722573257425752576257725782579258025812582258325842585258625872588258925902591259225932594259525962597259825992600260126022603260426052606260726082609261026112612261326142615261626172618261926202621262226232624262526262627262826292630263126322633263426352636263726382639264026412642264326442645264626472648264926502651265226532654265526562657265826592660266126622663266426652666266726682669267026712672267326742675267626772678267926802681268226832684268526862687268826892690269126922693269426952696269726982699270027012702270327042705270627072708270927102711271227132714271527162717271827192720272127222723272427252726272727282729273027312732273327342735273627372738273927402741274227432744274527462747274827492750275127522753275427552756275727582759276027612762276327642765276627672768276927702771277227732774277527762777277827792780278127822783278427852786278727882789279027912792279327942795279627972798279928002801280228032804280528062807280828092810281128122813281428152816281728182819282028212822282328242825282628272828282928302831283228332834283528362837283828392840284128422843284428452846284728482849285028512852285328542855285628572858285928602861286228632864286528662867286828692870287128722873287428752876287728782879288028812882288328842885288628872888288928902891289228932894289528962897289828992900290129022903290429052906290729082909291029112912291329142915291629172918291929202921292229232924292529262927292829292930293129322933293429352936293729382939294029412942294329442945294629472948294929502951295229532954295529562957295829592960296129622963296429652966296729682969297029712972297329742975297629772978297929802981298229832984298529862987298829892990299129922993299429952996299729982999300030013002300330043005300630073008300930103011301230133014301530163017301830193020302130223023302430253026302730283029303030313032303330343035303630373038303930403041304230433044304530463047304830493050305130523053305430553056305730583059306030613062306330643065306630673068306930703071307230733074307530763077307830793080308130823083308430853086308730883089309030913092309330943095309630973098309931003101310231033104310531063107310831093110311131123113311431153116311731183119312031213122312331243125312631273128312931303131313231333134313531363137313831393140314131423143314431453146314731483149315031513152315331543155315631573158315931603161316231633164316531663167316831693170317131723173317431753176317731783179318031813182318331843185318631873188318931903191319231933194319531963197319831993200320132023203320432053206320732083209321032113212321332143215321632173218321932203221322232233224322532263227322832293230323132323233323432353236323732383239324032413242324332443245324632473248324932503251325232533254325532563257325832593260326132623263326432653266326732683269327032713272327332743275327632773278327932803281328232833284328532863287328832893290329132923293329432953296329732983299330033013302330333043305330633073308330933103311331233133314331533163317331833193320332133223323332433253326332733283329333033313332333333343335333633373338333933403341334233433344334533463347334833493350335133523353335433553356335733583359336033613362336333643365336633673368336933703371337233733374337533763377337833793380338133823383338433853386338733883389339033913392339333943395339633973398339934003401340234033404340534063407340834093410341134123413341434153416341734183419342034213422342334243425342634273428342934303431343234333434343534363437343834393440344134423443344434453446344734483449345034513452345334543455345634573458345934603461346234633464346534663467346834693470347134723473347434753476347734783479348034813482348334843485348634873488348934903491349234933494349534963497349834993500350135023503350435053506350735083509351035113512351335143515351635173518351935203521352235233524352535263527352835293530353135323533353435353536353735383539354035413542354335443545354635473548354935503551355235533554355535563557355835593560356135623563356435653566356735683569357035713572357335743575357635773578357935803581358235833584358535863587358835893590359135923593359435953596359735983599360036013602360336043605360636073608360936103611361236133614361536163617361836193620362136223623362436253626362736283629363036313632363336343635363636373638363936403641364236433644364536463647364836493650365136523653365436553656365736583659366036613662366336643665366636673668366936703671367236733674367536763677367836793680368136823683368436853686368736883689369036913692369336943695369636973698369937003701370237033704370537063707370837093710371137123713371437153716371737183719372037213722372337243725372637273728372937303731373237333734373537363737373837393740374137423743374437453746374737483749375037513752375337543755375637573758375937603761376237633764376537663767376837693770377137723773377437753776377737783779378037813782378337843785378637873788378937903791379237933794379537963797379837993800380138023803380438053806380738083809381038113812381338143815381638173818381938203821382238233824382538263827382838293830383138323833383438353836383738383839384038413842384338443845384638473848384938503851385238533854385538563857385838593860386138623863386438653866386738683869387038713872387338743875387638773878387938803881388238833884388538863887388838893890389138923893389438953896389738983899390039013902390339043905390639073908390939103911391239133914391539163917391839193920392139223923392439253926392739283929393039313932393339343935393639373938393939403941394239433944394539463947394839493950395139523953395439553956395739583959396039613962396339643965396639673968396939703971397239733974397539763977397839793980398139823983398439853986398739883989399039913992399339943995399639973998399940004001400240034004400540064007400840094010401140124013401440154016401740184019402040214022402340244025402640274028402940304031403240334034403540364037403840394040404140424043404440454046404740484049405040514052405340544055405640574058405940604061406240634064406540664067406840694070407140724073407440754076407740784079408040814082408340844085408640874088408940904091409240934094409540964097409840994100410141024103410441054106410741084109411041114112411341144115411641174118411941204121412241234124412541264127412841294130413141324133413441354136413741384139414041414142414341444145414641474148414941504151415241534154415541564157415841594160416141624163416441654166416741684169417041714172417341744175417641774178417941804181418241834184418541864187418841894190419141924193419441954196419741984199420042014202420342044205420642074208420942104211421242134214421542164217421842194220422142224223422442254226422742284229423042314232423342344235423642374238423942404241424242434244424542464247424842494250425142524253425442554256425742584259426042614262426342644265426642674268426942704271427242734274427542764277427842794280428142824283428442854286428742884289429042914292429342944295429642974298429943004301430243034304430543064307430843094310431143124313431443154316431743184319432043214322432343244325432643274328432943304331433243334334433543364337433843394340434143424343434443454346434743484349435043514352435343544355435643574358435943604361436243634364436543664367436843694370437143724373437443754376437743784379438043814382438343844385438643874388438943904391439243934394439543964397439843994400440144024403440444054406440744084409441044114412441344144415441644174418441944204421442244234424442544264427442844294430443144324433443444354436443744384439444044414442444344444445444644474448444944504451445244534454445544564457445844594460446144624463446444654466446744684469447044714472447344744475447644774478447944804481448244834484448544864487448844894490449144924493449444954496449744984499450045014502450345044505450645074508450945104511451245134514451545164517451845194520452145224523452445254526452745284529453045314532453345344535453645374538453945404541454245434544454545464547454845494550455145524553455445554556455745584559456045614562456345644565456645674568456945704571457245734574457545764577457845794580458145824583458445854586458745884589459045914592459345944595459645974598459946004601460246034604460546064607460846094610461146124613461446154616461746184619462046214622462346244625462646274628462946304631463246334634463546364637463846394640464146424643464446454646464746484649465046514652465346544655465646574658465946604661466246634664466546664667466846694670467146724673467446754676467746784679468046814682468346844685468646874688468946904691469246934694469546964697469846994700470147024703470447054706470747084709471047114712471347144715471647174718471947204721472247234724472547264727472847294730473147324733473447354736473747384739474047414742474347444745474647474748474947504751475247534754475547564757475847594760476147624763476447654766476747684769477047714772477347744775477647774778477947804781478247834784478547864787478847894790479147924793479447954796479747984799480048014802480348044805480648074808480948104811481248134814481548164817481848194820482148224823482448254826482748284829483048314832483348344835483648374838483948404841484248434844484548464847484848494850485148524853485448554856485748584859486048614862486348644865486648674868486948704871487248734874487548764877487848794880488148824883488448854886488748884889489048914892489348944895489648974898489949004901490249034904490549064907490849094910491149124913491449154916491749184919492049214922492349244925492649274928492949304931493249334934493549364937493849394940494149424943494449454946494749484949495049514952495349544955495649574958495949604961496249634964496549664967496849694970497149724973497449754976497749784979498049814982498349844985498649874988498949904991499249934994499549964997499849995000500150025003500450055006500750085009501050115012501350145015501650175018501950205021502250235024502550265027502850295030503150325033503450355036503750385039504050415042504350445045504650475048504950505051505250535054505550565057505850595060506150625063506450655066506750685069507050715072507350745075507650775078507950805081508250835084508550865087508850895090509150925093509450955096509750985099510051015102510351045105510651075108510951105111511251135114511551165117511851195120512151225123512451255126512751285129513051315132513351345135513651375138513951405141514251435144514551465147514851495150515151525153515451555156515751585159516051615162516351645165516651675168516951705171517251735174517551765177517851795180518151825183518451855186518751885189519051915192519351945195519651975198519952005201520252035204520552065207520852095210521152125213521452155216521752185219522052215222522352245225522652275228522952305231523252335234523552365237523852395240524152425243524452455246524752485249525052515252525352545255525652575258525952605261526252635264526552665267526852695270527152725273527452755276527752785279528052815282528352845285528652875288528952905291529252935294529552965297529852995300530153025303530453055306530753085309531053115312531353145315531653175318531953205321532253235324532553265327532853295330533153325333533453355336533753385339534053415342534353445345534653475348534953505351535253535354535553565357535853595360536153625363536453655366536753685369537053715372537353745375537653775378537953805381538253835384538553865387538853895390539153925393539453955396539753985399540054015402540354045405540654075408540954105411541254135414541554165417541854195420542154225423542454255426542754285429543054315432543354345435543654375438543954405441544254435444544554465447544854495450545154525453545454555456545754585459546054615462546354645465546654675468546954705471547254735474547554765477547854795480548154825483548454855486548754885489549054915492549354945495549654975498549955005501550255035504550555065507550855095510551155125513551455155516551755185519552055215522552355245525552655275528552955305531553255335534553555365537553855395540554155425543554455455546554755485549555055515552555355545555555655575558555955605561556255635564556555665567556855695570557155725573557455755576557755785579558055815582558355845585558655875588558955905591559255935594559555965597559855995600560156025603560456055606560756085609561056115612561356145615561656175618561956205621562256235624562556265627562856295630563156325633563456355636563756385639564056415642564356445645564656475648564956505651565256535654565556565657565856595660566156625663566456655666566756685669567056715672567356745675567656775678567956805681568256835684568556865687568856895690569156925693569456955696569756985699570057015702570357045705570657075708570957105711571257135714571557165717571857195720572157225723572457255726572757285729573057315732573357345735573657375738573957405741574257435744574557465747574857495750575157525753575457555756575757585759576057615762576357645765576657675768576957705771577257735774577557765777577857795780578157825783578457855786578757885789579057915792579357945795579657975798579958005801580258035804580558065807580858095810581158125813581458155816581758185819582058215822582358245825582658275828582958305831583258335834583558365837583858395840584158425843584458455846584758485849585058515852585358545855585658575858585958605861586258635864586558665867586858695870587158725873587458755876587758785879588058815882588358845885588658875888588958905891589258935894589558965897589858995900590159025903590459055906590759085909591059115912591359145915591659175918591959205921592259235924592559265927592859295930593159325933593459355936593759385939594059415942594359445945594659475948594959505951595259535954595559565957595859595960596159625963596459655966596759685969597059715972597359745975597659775978597959805981598259835984598559865987598859895990599159925993599459955996599759985999600060016002600360046005600660076008600960106011601260136014601560166017601860196020602160226023602460256026602760286029603060316032603360346035603660376038603960406041604260436044604560466047604860496050605160526053605460556056605760586059606060616062606360646065606660676068606960706071607260736074607560766077607860796080608160826083608460856086608760886089609060916092609360946095609660976098609961006101610261036104610561066107610861096110611161126113611461156116611761186119612061216122612361246125612661276128612961306131613261336134613561366137613861396140614161426143614461456146614761486149615061516152615361546155615661576158615961606161616261636164616561666167616861696170617161726173617461756176617761786179618061816182618361846185618661876188618961906191619261936194619561966197619861996200620162026203620462056206620762086209621062116212621362146215621662176218621962206221622262236224622562266227622862296230623162326233623462356236623762386239624062416242624362446245624662476248624962506251625262536254625562566257625862596260626162626263626462656266626762686269627062716272627362746275627662776278627962806281628262836284628562866287628862896290629162926293629462956296629762986299630063016302630363046305630663076308630963106311631263136314631563166317631863196320632163226323632463256326632763286329633063316332633363346335633663376338633963406341634263436344634563466347634863496350635163526353635463556356635763586359636063616362636363646365636663676368636963706371637263736374637563766377637863796380638163826383638463856386638763886389639063916392639363946395639663976398639964006401640264036404640564066407640864096410641164126413641464156416641764186419642064216422642364246425642664276428642964306431643264336434643564366437643864396440644164426443644464456446644764486449645064516452645364546455645664576458645964606461646264636464646564666467646864696470647164726473647464756476647764786479648064816482648364846485648664876488648964906491649264936494649564966497649864996500650165026503650465056506650765086509651065116512651365146515651665176518651965206521652265236524652565266527652865296530653165326533653465356536653765386539654065416542654365446545654665476548654965506551655265536554655565566557655865596560656165626563656465656566656765686569657065716572657365746575657665776578657965806581658265836584658565866587658865896590659165926593659465956596659765986599660066016602660366046605660666076608660966106611661266136614661566166617661866196620662166226623662466256626662766286629663066316632663366346635663666376638663966406641664266436644664566466647664866496650665166526653665466556656665766586659666066616662666366646665666666676668666966706671667266736674667566766677667866796680668166826683668466856686668766886689669066916692669366946695669666976698669967006701670267036704670567066707670867096710671167126713671467156716671767186719672067216722672367246725672667276728672967306731673267336734673567366737673867396740674167426743674467456746674767486749675067516752675367546755675667576758675967606761676267636764676567666767676867696770677167726773677467756776677767786779678067816782678367846785678667876788678967906791679267936794679567966797679867996800680168026803680468056806680768086809681068116812681368146815681668176818681968206821682268236824682568266827682868296830683168326833683468356836683768386839684068416842684368446845684668476848684968506851685268536854685568566857685868596860686168626863686468656866686768686869687068716872687368746875687668776878687968806881688268836884688568866887688868896890689168926893689468956896689768986899690069016902690369046905690669076908690969106911691269136914691569166917691869196920692169226923692469256926692769286929693069316932693369346935693669376938693969406941694269436944694569466947694869496950695169526953695469556956695769586959696069616962696369646965696669676968696969706971697269736974697569766977697869796980698169826983698469856986698769886989699069916992699369946995699669976998699970007001700270037004700570067007700870097010701170127013701470157016701770187019702070217022702370247025702670277028702970307031703270337034703570367037703870397040704170427043704470457046704770487049705070517052705370547055705670577058705970607061706270637064706570667067706870697070707170727073707470757076707770787079708070817082708370847085708670877088708970907091709270937094709570967097709870997100710171027103710471057106710771087109711071117112711371147115711671177118711971207121712271237124712571267127712871297130713171327133713471357136713771387139714071417142714371447145714671477148714971507151715271537154715571567157715871597160716171627163716471657166716771687169717071717172717371747175717671777178717971807181718271837184718571867187718871897190719171927193719471957196719771987199720072017202720372047205720672077208720972107211721272137214721572167217721872197220722172227223722472257226722772287229723072317232723372347235723672377238723972407241724272437244724572467247724872497250725172527253725472557256725772587259726072617262726372647265726672677268726972707271727272737274727572767277727872797280728172827283728472857286728772887289729072917292729372947295729672977298729973007301730273037304730573067307730873097310731173127313731473157316731773187319732073217322732373247325732673277328732973307331733273337334733573367337733873397340734173427343734473457346734773487349735073517352735373547355735673577358735973607361736273637364736573667367736873697370737173727373737473757376737773787379738073817382738373847385738673877388738973907391739273937394739573967397739873997400740174027403740474057406740774087409741074117412741374147415741674177418741974207421742274237424742574267427742874297430743174327433743474357436743774387439744074417442744374447445744674477448744974507451745274537454745574567457745874597460746174627463746474657466746774687469747074717472747374747475747674777478747974807481748274837484748574867487748874897490749174927493749474957496749774987499750075017502750375047505750675077508750975107511751275137514751575167517751875197520752175227523752475257526752775287529753075317532753375347535753675377538753975407541754275437544754575467547754875497550755175527553755475557556755775587559756075617562756375647565756675677568756975707571757275737574757575767577757875797580758175827583758475857586758775887589759075917592759375947595759675977598759976007601760276037604760576067607760876097610761176127613761476157616761776187619762076217622762376247625762676277628762976307631763276337634763576367637763876397640764176427643764476457646764776487649765076517652765376547655765676577658765976607661766276637664766576667667766876697670767176727673767476757676767776787679768076817682768376847685768676877688768976907691769276937694769576967697769876997700770177027703770477057706770777087709771077117712771377147715771677177718771977207721772277237724772577267727772877297730773177327733773477357736773777387739774077417742774377447745774677477748774977507751775277537754775577567757775877597760776177627763776477657766776777687769777077717772777377747775777677777778777977807781778277837784778577867787778877897790779177927793779477957796779777987799780078017802780378047805780678077808780978107811781278137814781578167817781878197820782178227823782478257826782778287829783078317832783378347835783678377838783978407841784278437844784578467847784878497850785178527853785478557856785778587859786078617862786378647865786678677868786978707871787278737874787578767877787878797880788178827883788478857886788778887889789078917892789378947895789678977898789979007901790279037904790579067907790879097910791179127913791479157916791779187919792079217922792379247925792679277928792979307931793279337934793579367937793879397940794179427943794479457946794779487949795079517952795379547955795679577958795979607961796279637964796579667967796879697970797179727973797479757976797779787979798079817982798379847985798679877988798979907991799279937994799579967997799879998000800180028003800480058006800780088009801080118012801380148015801680178018801980208021802280238024802580268027802880298030803180328033803480358036803780388039804080418042804380448045804680478048804980508051805280538054805580568057805880598060806180628063806480658066806780688069807080718072807380748075807680778078807980808081808280838084808580868087808880898090809180928093809480958096809780988099810081018102810381048105810681078108810981108111811281138114811581168117811881198120812181228123812481258126812781288129813081318132813381348135813681378138813981408141814281438144814581468147814881498150815181528153815481558156815781588159816081618162816381648165816681678168816981708171817281738174817581768177817881798180818181828183818481858186818781888189819081918192819381948195819681978198819982008201820282038204820582068207820882098210821182128213821482158216821782188219822082218222822382248225822682278228822982308231823282338234823582368237823882398240824182428243824482458246824782488249825082518252825382548255825682578258825982608261826282638264826582668267826882698270827182728273827482758276827782788279828082818282828382848285828682878288828982908291829282938294829582968297829882998300830183028303830483058306830783088309831083118312831383148315831683178318831983208321832283238324832583268327832883298330833183328333833483358336833783388339834083418342834383448345834683478348834983508351835283538354835583568357835883598360836183628363836483658366836783688369837083718372837383748375837683778378837983808381838283838384838583868387838883898390839183928393839483958396839783988399840084018402840384048405840684078408840984108411841284138414841584168417841884198420842184228423842484258426842784288429843084318432843384348435843684378438843984408441844284438444844584468447844884498450845184528453845484558456845784588459846084618462846384648465846684678468846984708471847284738474847584768477847884798480848184828483848484858486848784888489849084918492849384948495849684978498849985008501850285038504850585068507850885098510851185128513851485158516851785188519852085218522852385248525852685278528852985308531853285338534853585368537853885398540854185428543854485458546854785488549855085518552855385548555855685578558855985608561856285638564856585668567856885698570857185728573857485758576857785788579858085818582858385848585858685878588858985908591859285938594859585968597859885998600860186028603860486058606860786088609861086118612861386148615861686178618861986208621862286238624862586268627862886298630863186328633863486358636863786388639864086418642864386448645864686478648864986508651865286538654865586568657865886598660866186628663866486658666866786688669867086718672867386748675867686778678867986808681868286838684868586868687868886898690869186928693869486958696869786988699870087018702870387048705870687078708870987108711871287138714871587168717871887198720872187228723872487258726872787288729873087318732873387348735873687378738873987408741874287438744874587468747874887498750875187528753875487558756875787588759876087618762876387648765876687678768876987708771877287738774877587768777877887798780878187828783878487858786878787888789879087918792879387948795879687978798879988008801880288038804880588068807880888098810881188128813881488158816881788188819882088218822882388248825882688278828882988308831883288338834883588368837883888398840884188428843884488458846884788488849885088518852885388548855885688578858885988608861886288638864886588668867886888698870887188728873887488758876887788788879888088818882888388848885888688878888888988908891889288938894889588968897889888998900890189028903890489058906890789088909891089118912891389148915891689178918891989208921892289238924892589268927892889298930893189328933893489358936893789388939894089418942894389448945894689478948894989508951895289538954895589568957895889598960896189628963896489658966896789688969897089718972897389748975897689778978897989808981898289838984898589868987898889898990899189928993899489958996899789988999900090019002900390049005900690079008900990109011901290139014901590169017901890199020902190229023902490259026902790289029903090319032903390349035903690379038903990409041904290439044904590469047904890499050905190529053905490559056905790589059906090619062906390649065906690679068906990709071907290739074907590769077907890799080908190829083908490859086908790889089909090919092909390949095909690979098909991009101910291039104910591069107910891099110911191129113911491159116911791189119912091219122912391249125912691279128912991309131913291339134913591369137913891399140914191429143914491459146914791489149915091519152915391549155915691579158915991609161916291639164916591669167916891699170917191729173917491759176917791789179918091819182918391849185918691879188918991909191919291939194919591969197919891999200920192029203920492059206920792089209921092119212921392149215921692179218921992209221922292239224922592269227922892299230923192329233923492359236923792389239924092419242924392449245924692479248924992509251925292539254925592569257925892599260926192629263926492659266926792689269927092719272927392749275927692779278927992809281928292839284928592869287928892899290929192929293929492959296929792989299930093019302930393049305930693079308930993109311931293139314931593169317931893199320932193229323932493259326932793289329933093319332933393349335933693379338933993409341934293439344934593469347934893499350935193529353935493559356935793589359936093619362936393649365936693679368936993709371937293739374937593769377937893799380938193829383938493859386938793889389939093919392939393949395939693979398939994009401940294039404940594069407940894099410941194129413941494159416941794189419942094219422942394249425942694279428942994309431943294339434943594369437943894399440944194429443944494459446944794489449945094519452945394549455945694579458945994609461946294639464946594669467946894699470947194729473947494759476947794789479948094819482948394849485948694879488948994909491949294939494949594969497949894999500950195029503950495059506950795089509951095119512951395149515951695179518951995209521952295239524952595269527952895299530953195329533953495359536953795389539954095419542954395449545954695479548954995509551955295539554955595569557955895599560956195629563956495659566956795689569957095719572957395749575957695779578957995809581958295839584958595869587958895899590959195929593959495959596959795989599960096019602960396049605960696079608960996109611961296139614961596169617961896199620962196229623962496259626962796289629963096319632963396349635963696379638963996409641964296439644964596469647964896499650965196529653965496559656965796589659966096619662966396649665966696679668966996709671967296739674967596769677967896799680968196829683968496859686968796889689969096919692969396949695969696979698969997009701970297039704970597069707970897099710971197129713971497159716971797189719972097219722972397249725972697279728972997309731973297339734973597369737973897399740974197429743974497459746974797489749975097519752975397549755975697579758975997609761976297639764976597669767976897699770977197729773977497759776977797789779978097819782978397849785978697879788978997909791979297939794979597969797979897999800980198029803980498059806980798089809981098119812981398149815981698179818981998209821982298239824982598269827982898299830983198329833983498359836983798389839984098419842984398449845984698479848984998509851985298539854985598569857985898599860986198629863986498659866986798689869987098719872987398749875987698779878987998809881988298839884988598869887988898899890989198929893989498959896989798989899990099019902990399049905990699079908990999109911991299139914991599169917991899199920992199229923992499259926992799289929993099319932993399349935993699379938993999409941994299439944994599469947994899499950995199529953995499559956995799589959996099619962996399649965996699679968996999709971997299739974997599769977997899799980998199829983998499859986998799889989999099919992999399949995999699979998999910000100011000210003100041000510006100071000810009100101001110012100131001410015100161001710018100191002010021100221002310024100251002610027100281002910030100311003210033100341003510036100371003810039100401004110042100431004410045100461004710048100491005010051100521005310054100551005610057100581005910060100611006210063100641006510066100671006810069100701007110072100731007410075100761007710078100791008010081100821008310084100851008610087100881008910090100911009210093100941009510096100971009810099101001010110102101031010410105101061010710108101091011010111101121011310114101151011610117101181011910120101211012210123101241012510126101271012810129101301013110132101331013410135101361013710138101391014010141101421014310144101451014610147101481014910150101511015210153101541015510156101571015810159101601016110162101631016410165101661016710168101691017010171101721017310174101751017610177101781017910180101811018210183101841018510186101871018810189101901019110192101931019410195101961019710198101991020010201102021020310204102051020610207102081020910210102111021210213102141021510216102171021810219102201022110222102231022410225102261022710228102291023010231102321023310234102351023610237102381023910240102411024210243102441024510246102471024810249102501025110252102531025410255102561025710258102591026010261102621026310264102651026610267102681026910270102711027210273102741027510276102771027810279102801028110282102831028410285102861028710288102891029010291102921029310294102951029610297102981029910300103011030210303103041030510306103071030810309103101031110312103131031410315103161031710318103191032010321103221032310324103251032610327103281032910330103311033210333103341033510336103371033810339103401034110342103431034410345103461034710348103491035010351103521035310354103551035610357103581035910360103611036210363103641036510366103671036810369103701037110372103731037410375103761037710378103791038010381103821038310384103851038610387103881038910390103911039210393103941039510396103971039810399104001040110402104031040410405104061040710408104091041010411104121041310414104151041610417104181041910420104211042210423104241042510426104271042810429104301043110432104331043410435104361043710438104391044010441104421044310444104451044610447104481044910450104511045210453104541045510456104571045810459104601046110462104631046410465104661046710468104691047010471104721047310474104751047610477104781047910480104811048210483104841048510486104871048810489104901049110492104931049410495104961049710498104991050010501105021050310504105051050610507105081050910510105111051210513105141051510516105171051810519105201052110522105231052410525105261052710528105291053010531105321053310534105351053610537105381053910540105411054210543105441054510546105471054810549105501055110552105531055410555105561055710558105591056010561105621056310564105651056610567105681056910570105711057210573105741057510576105771057810579105801058110582105831058410585105861058710588105891059010591105921059310594105951059610597105981059910600106011060210603106041060510606106071060810609106101061110612106131061410615106161061710618106191062010621106221062310624106251062610627106281062910630106311063210633106341063510636106371063810639106401064110642106431064410645106461064710648106491065010651106521065310654106551065610657106581065910660106611066210663106641066510666106671066810669106701067110672106731067410675106761067710678106791068010681106821068310684106851068610687106881068910690106911069210693106941069510696106971069810699107001070110702107031070410705107061070710708107091071010711107121071310714107151071610717107181071910720107211072210723107241072510726107271072810729107301073110732107331073410735107361073710738107391074010741107421074310744107451074610747107481074910750107511075210753107541075510756107571075810759107601076110762107631076410765107661076710768107691077010771107721077310774107751077610777107781077910780107811078210783107841078510786107871078810789107901079110792107931079410795107961079710798107991080010801108021080310804108051080610807108081080910810108111081210813108141081510816108171081810819108201082110822108231082410825108261082710828108291083010831108321083310834108351083610837108381083910840108411084210843108441084510846108471084810849108501085110852108531085410855108561085710858108591086010861108621086310864108651086610867108681086910870108711087210873108741087510876108771087810879108801088110882108831088410885108861088710888108891089010891108921089310894108951089610897108981089910900109011090210903109041090510906109071090810909109101091110912109131091410915109161091710918109191092010921109221092310924109251092610927109281092910930109311093210933109341093510936109371093810939109401094110942109431094410945109461094710948109491095010951109521095310954109551095610957109581095910960109611096210963109641096510966109671096810969109701097110972109731097410975109761097710978109791098010981109821098310984109851098610987109881098910990109911099210993109941099510996109971099810999110001100111002110031100411005110061100711008110091101011011110121101311014110151101611017110181101911020110211102211023110241102511026110271102811029110301103111032110331103411035110361103711038110391104011041110421104311044110451104611047110481104911050110511105211053110541105511056110571105811059110601106111062110631106411065110661106711068110691107011071110721107311074110751107611077110781107911080110811108211083110841108511086110871108811089110901109111092110931109411095110961109711098110991110011101111021110311104111051110611107111081110911110111111111211113111141111511116111171111811119111201112111122111231112411125111261112711128111291113011131111321113311134111351113611137111381113911140111411114211143111441114511146111471114811149111501115111152111531115411155111561115711158111591116011161111621116311164111651116611167111681116911170111711117211173111741117511176111771117811179111801118111182111831118411185111861118711188111891119011191111921119311194111951119611197111981119911200112011120211203112041120511206112071120811209112101121111212112131121411215112161121711218112191122011221112221122311224112251122611227112281122911230112311123211233112341123511236112371123811239112401124111242112431124411245112461124711248112491125011251112521125311254112551125611257112581125911260112611126211263112641126511266112671126811269112701127111272112731127411275112761127711278112791128011281112821128311284112851128611287112881128911290112911129211293112941129511296112971129811299113001130111302113031130411305113061130711308113091131011311113121131311314113151131611317113181131911320113211132211323113241132511326113271132811329113301133111332113331133411335113361133711338113391134011341113421134311344113451134611347113481134911350113511135211353113541135511356113571135811359113601136111362113631136411365113661136711368113691137011371113721137311374113751137611377113781137911380113811138211383113841138511386113871138811389113901139111392113931139411395113961139711398113991140011401114021140311404114051140611407114081140911410114111141211413114141141511416114171141811419114201142111422114231142411425114261142711428114291143011431114321143311434114351143611437114381143911440114411144211443114441144511446114471144811449114501145111452114531145411455114561145711458114591146011461114621146311464114651146611467114681146911470114711147211473114741147511476114771147811479114801148111482114831148411485114861148711488114891149011491114921149311494114951149611497114981149911500115011150211503115041150511506115071150811509115101151111512115131151411515115161151711518115191152011521115221152311524115251152611527115281152911530115311153211533115341153511536115371153811539115401154111542115431154411545115461154711548115491155011551115521155311554115551155611557115581155911560115611156211563115641156511566115671156811569115701157111572115731157411575115761157711578115791158011581115821158311584115851158611587115881158911590115911159211593115941159511596115971159811599116001160111602116031160411605116061160711608116091161011611116121161311614116151161611617116181161911620116211162211623116241162511626116271162811629116301163111632116331163411635116361163711638116391164011641116421164311644116451164611647116481164911650116511165211653116541165511656116571165811659116601166111662116631166411665116661166711668116691167011671116721167311674116751167611677116781167911680116811168211683116841168511686116871168811689116901169111692116931169411695116961169711698116991170011701117021170311704117051170611707117081170911710117111171211713117141171511716117171171811719117201172111722117231172411725117261172711728117291173011731117321173311734117351173611737117381173911740117411174211743117441174511746117471174811749117501175111752117531175411755117561175711758117591176011761117621176311764117651176611767117681176911770117711177211773117741177511776117771177811779117801178111782117831178411785117861178711788117891179011791117921179311794117951179611797117981179911800118011180211803118041180511806118071180811809118101181111812118131181411815118161181711818118191182011821118221182311824118251182611827118281182911830118311183211833118341183511836118371183811839118401184111842118431184411845118461184711848118491185011851118521185311854118551185611857118581185911860118611186211863118641186511866118671186811869118701187111872118731187411875118761187711878118791188011881118821188311884118851188611887118881188911890118911189211893118941189511896118971189811899119001190111902119031190411905119061190711908119091191011911119121191311914119151191611917119181191911920119211192211923119241192511926119271192811929119301193111932119331193411935119361193711938119391194011941119421194311944119451194611947119481194911950119511195211953119541195511956119571195811959119601196111962119631196411965119661196711968119691197011971119721197311974119751197611977119781197911980119811198211983119841198511986119871198811989119901199111992119931199411995119961199711998119991200012001120021200312004120051200612007120081200912010120111201212013120141201512016120171201812019120201202112022120231202412025120261202712028120291203012031120321203312034120351203612037120381203912040120411204212043120441204512046120471204812049120501205112052120531205412055120561205712058120591206012061120621206312064120651206612067120681206912070120711207212073120741207512076120771207812079120801208112082120831208412085120861208712088120891209012091120921209312094120951209612097120981209912100121011210212103121041210512106121071210812109121101211112112121131211412115121161211712118121191212012121121221212312124121251212612127121281212912130121311213212133121341213512136121371213812139121401214112142121431214412145121461214712148121491215012151121521215312154121551215612157121581215912160121611216212163121641216512166121671216812169121701217112172121731217412175121761217712178121791218012181121821218312184121851218612187121881218912190121911219212193121941219512196121971219812199122001220112202122031220412205122061220712208122091221012211122121221312214122151221612217122181221912220122211222212223122241222512226122271222812229122301223112232122331223412235122361223712238122391224012241122421224312244122451224612247122481224912250122511225212253122541225512256122571225812259122601226112262122631226412265122661226712268122691227012271122721227312274122751227612277122781227912280122811228212283122841228512286122871228812289122901229112292122931229412295122961229712298122991230012301123021230312304123051230612307123081230912310123111231212313123141231512316123171231812319123201232112322123231232412325123261232712328123291233012331123321233312334123351233612337123381233912340123411234212343123441234512346123471234812349123501235112352123531235412355123561235712358123591236012361123621236312364123651236612367123681236912370123711237212373123741237512376123771237812379123801238112382123831238412385123861238712388123891239012391123921239312394123951239612397123981239912400124011240212403124041240512406124071240812409124101241112412124131241412415124161241712418124191242012421124221242312424124251242612427124281242912430124311243212433124341243512436124371243812439124401244112442124431244412445124461244712448124491245012451124521245312454124551245612457124581245912460124611246212463124641246512466124671246812469124701247112472124731247412475124761247712478124791248012481124821248312484124851248612487124881248912490124911249212493124941249512496124971249812499125001250112502125031250412505125061250712508125091251012511125121251312514125151251612517125181251912520125211252212523125241252512526125271252812529125301253112532125331253412535125361253712538125391254012541125421254312544125451254612547125481254912550125511255212553125541255512556125571255812559125601256112562125631256412565125661256712568125691257012571125721257312574125751257612577125781257912580125811258212583125841258512586125871258812589125901259112592125931259412595125961259712598125991260012601126021260312604126051260612607126081260912610126111261212613126141261512616126171261812619126201262112622126231262412625126261262712628126291263012631126321263312634126351263612637126381263912640126411264212643126441264512646126471264812649126501265112652126531265412655126561265712658126591266012661126621266312664126651266612667126681266912670126711267212673126741267512676126771267812679126801268112682126831268412685126861268712688126891269012691126921269312694126951269612697126981269912700127011270212703127041270512706127071270812709127101271112712127131271412715127161271712718127191272012721127221272312724127251272612727127281272912730127311273212733127341273512736127371273812739127401274112742127431274412745127461274712748127491275012751127521275312754127551275612757127581275912760127611276212763127641276512766127671276812769127701277112772127731277412775127761277712778127791278012781127821278312784127851278612787127881278912790127911279212793127941279512796127971279812799128001280112802128031280412805128061280712808128091281012811128121281312814128151281612817128181281912820128211282212823128241282512826128271282812829128301283112832128331283412835128361283712838128391284012841128421284312844128451284612847128481284912850128511285212853128541285512856128571285812859128601286112862128631286412865128661286712868128691287012871128721287312874128751287612877128781287912880128811288212883128841288512886128871288812889128901289112892128931289412895128961289712898128991290012901129021290312904129051290612907129081290912910129111291212913129141291512916129171291812919129201292112922129231292412925129261292712928129291293012931129321293312934129351293612937129381293912940129411294212943129441294512946129471294812949129501295112952129531295412955129561295712958129591296012961129621296312964129651296612967129681296912970129711297212973129741297512976129771297812979129801298112982129831298412985129861298712988129891299012991129921299312994129951299612997129981299913000130011300213003130041300513006130071300813009130101301113012130131301413015130161301713018130191302013021130221302313024130251302613027130281302913030130311303213033130341303513036130371303813039130401304113042130431304413045130461304713048130491305013051130521305313054130551305613057130581305913060130611306213063130641306513066130671306813069130701307113072130731307413075130761307713078130791308013081130821308313084130851308613087130881308913090130911309213093130941309513096130971309813099131001310113102131031310413105131061310713108131091311013111131121311313114131151311613117131181311913120131211312213123131241312513126131271312813129131301313113132131331313413135131361313713138131391314013141131421314313144131451314613147131481314913150131511315213153131541315513156131571315813159131601316113162131631316413165131661316713168131691317013171131721317313174131751317613177131781317913180131811318213183131841318513186131871318813189131901319113192131931319413195131961319713198131991320013201132021320313204132051320613207132081320913210132111321213213132141321513216132171321813219132201322113222132231322413225132261322713228132291323013231132321323313234132351323613237132381323913240132411324213243132441324513246132471324813249132501325113252132531325413255132561325713258132591326013261132621326313264132651326613267132681326913270132711327213273132741327513276132771327813279132801328113282132831328413285132861328713288132891329013291132921329313294132951329613297132981329913300133011330213303133041330513306133071330813309133101331113312133131331413315133161331713318133191332013321133221332313324133251332613327133281332913330133311333213333133341333513336133371333813339133401334113342133431334413345133461334713348133491335013351133521335313354133551335613357133581335913360133611336213363133641336513366133671336813369133701337113372133731337413375133761337713378133791338013381133821338313384133851338613387133881338913390133911339213393133941339513396133971339813399134001340113402134031340413405134061340713408134091341013411134121341313414134151341613417134181341913420134211342213423134241342513426134271342813429134301343113432134331343413435134361343713438134391344013441134421344313444134451344613447134481344913450134511345213453134541345513456134571345813459134601346113462134631346413465134661346713468134691347013471134721347313474134751347613477134781347913480134811348213483134841348513486134871348813489134901349113492134931349413495134961349713498134991350013501135021350313504135051350613507135081350913510135111351213513135141351513516135171351813519135201352113522135231352413525135261352713528135291353013531135321353313534135351353613537135381353913540135411354213543135441354513546135471354813549135501355113552135531355413555135561355713558135591356013561135621356313564135651356613567135681356913570135711357213573135741357513576135771357813579135801358113582135831358413585135861358713588135891359013591135921359313594135951359613597135981359913600136011360213603136041360513606136071360813609136101361113612136131361413615136161361713618136191362013621136221362313624136251362613627136281362913630136311363213633136341363513636136371363813639136401364113642136431364413645136461364713648136491365013651136521365313654136551365613657136581365913660136611366213663136641366513666136671366813669136701367113672136731367413675136761367713678136791368013681136821368313684136851368613687136881368913690136911369213693136941369513696136971369813699137001370113702137031370413705137061370713708137091371013711137121371313714137151371613717137181371913720137211372213723137241372513726137271372813729137301373113732137331373413735137361373713738137391374013741137421374313744137451374613747137481374913750137511375213753137541375513756137571375813759137601376113762137631376413765137661376713768137691377013771137721377313774137751377613777137781377913780137811378213783137841378513786137871378813789137901379113792137931379413795137961379713798137991380013801138021380313804138051380613807138081380913810138111381213813138141381513816138171381813819138201382113822138231382413825138261382713828138291383013831138321383313834138351383613837138381383913840138411384213843138441384513846138471384813849138501385113852138531385413855138561385713858138591386013861138621386313864138651386613867138681386913870138711387213873138741387513876138771387813879138801388113882138831388413885138861388713888138891389013891138921389313894138951389613897138981389913900139011390213903139041390513906139071390813909139101391113912139131391413915139161391713918139191392013921139221392313924139251392613927139281392913930139311393213933139341393513936139371393813939139401394113942139431394413945139461394713948139491395013951139521395313954139551395613957139581395913960139611396213963139641396513966139671396813969139701397113972139731397413975139761397713978139791398013981139821398313984139851398613987139881398913990139911399213993139941399513996139971399813999140001400114002140031400414005140061400714008140091401014011140121401314014140151401614017140181401914020140211402214023140241402514026140271402814029140301403114032140331403414035140361403714038140391404014041140421404314044140451404614047140481404914050140511405214053140541405514056140571405814059140601406114062140631406414065140661406714068140691407014071140721407314074140751407614077140781407914080140811408214083140841408514086140871408814089140901409114092140931409414095140961409714098140991410014101141021410314104141051410614107141081410914110141111411214113141141411514116141171411814119141201412114122141231412414125141261412714128141291413014131141321413314134141351413614137141381413914140141411414214143141441414514146141471414814149141501415114152141531415414155141561415714158141591416014161141621416314164141651416614167141681416914170141711417214173141741417514176141771417814179141801418114182141831418414185141861418714188141891419014191141921419314194141951419614197141981419914200142011420214203142041420514206142071420814209142101421114212142131421414215142161421714218142191422014221142221422314224142251422614227142281422914230142311423214233142341423514236142371423814239142401424114242142431424414245142461424714248142491425014251142521425314254142551425614257142581425914260142611426214263142641426514266142671426814269142701427114272142731427414275142761427714278142791428014281142821428314284142851428614287142881428914290142911429214293142941429514296142971429814299143001430114302143031430414305143061430714308143091431014311143121431314314143151431614317143181431914320143211432214323143241432514326143271432814329143301433114332143331433414335143361433714338143391434014341143421434314344143451434614347143481434914350143511435214353143541435514356143571435814359143601436114362143631436414365143661436714368143691437014371143721437314374143751437614377143781437914380143811438214383143841438514386143871438814389143901439114392143931439414395143961439714398143991440014401144021440314404144051440614407144081440914410144111441214413144141441514416144171441814419144201442114422144231442414425144261442714428144291443014431144321443314434144351443614437144381443914440144411444214443144441444514446144471444814449144501445114452144531445414455144561445714458144591446014461144621446314464144651446614467144681446914470144711447214473144741447514476144771447814479144801448114482144831448414485144861448714488144891449014491144921449314494144951449614497144981449914500145011450214503145041450514506145071450814509145101451114512145131451414515145161451714518145191452014521145221452314524145251452614527145281452914530145311453214533145341453514536145371453814539145401454114542145431454414545145461454714548145491455014551145521455314554145551455614557145581455914560145611456214563145641456514566145671456814569145701457114572145731457414575145761457714578145791458014581145821458314584145851458614587145881458914590145911459214593145941459514596145971459814599146001460114602146031460414605146061460714608146091461014611146121461314614146151461614617146181461914620146211462214623146241462514626146271462814629146301463114632146331463414635146361463714638146391464014641146421464314644146451464614647146481464914650146511465214653146541465514656146571465814659146601466114662146631466414665146661466714668146691467014671146721467314674146751467614677146781467914680146811468214683146841468514686146871468814689146901469114692146931469414695146961469714698146991470014701147021470314704147051470614707147081470914710147111471214713147141471514716147171471814719147201472114722147231472414725147261472714728147291473014731147321473314734147351473614737147381473914740147411474214743147441474514746147471474814749147501475114752147531475414755147561475714758147591476014761147621476314764147651476614767147681476914770147711477214773147741477514776147771477814779147801478114782147831478414785147861478714788147891479014791147921479314794147951479614797147981479914800148011480214803148041480514806148071480814809148101481114812148131481414815148161481714818148191482014821148221482314824148251482614827148281482914830148311483214833148341483514836148371483814839148401484114842148431484414845148461484714848148491485014851148521485314854148551485614857148581485914860148611486214863148641486514866148671486814869148701487114872148731487414875148761487714878148791488014881148821488314884148851488614887148881488914890148911489214893148941489514896148971489814899149001490114902149031490414905149061490714908149091491014911149121491314914149151491614917149181491914920149211492214923149241492514926149271492814929149301493114932149331493414935149361493714938149391494014941149421494314944149451494614947149481494914950149511495214953149541495514956149571495814959149601496114962149631496414965149661496714968149691497014971149721497314974149751497614977149781497914980149811498214983149841498514986149871498814989149901499114992149931499414995149961499714998149991500015001150021500315004150051500615007150081500915010150111501215013150141501515016150171501815019150201502115022150231502415025150261502715028150291503015031150321503315034150351503615037150381503915040150411504215043150441504515046150471504815049150501505115052150531505415055150561505715058150591506015061150621506315064150651506615067150681506915070150711507215073150741507515076150771507815079150801508115082150831508415085150861508715088150891509015091150921509315094150951509615097150981509915100151011510215103151041510515106151071510815109151101511115112151131511415115151161511715118151191512015121151221512315124151251512615127151281512915130151311513215133151341513515136151371513815139151401514115142151431514415145151461514715148151491515015151151521515315154151551515615157151581515915160151611516215163151641516515166151671516815169151701517115172151731517415175151761517715178151791518015181151821518315184151851518615187151881518915190151911519215193151941519515196151971519815199152001520115202152031520415205152061520715208152091521015211152121521315214152151521615217152181521915220152211522215223152241522515226152271522815229152301523115232152331523415235152361523715238152391524015241152421524315244152451524615247152481524915250152511525215253152541525515256152571525815259152601526115262152631526415265152661526715268152691527015271152721527315274152751527615277152781527915280152811528215283152841528515286152871528815289152901529115292152931529415295152961529715298152991530015301153021530315304153051530615307153081530915310153111531215313153141531515316153171531815319153201532115322153231532415325153261532715328153291533015331153321533315334153351533615337153381533915340153411534215343153441534515346153471534815349153501535115352153531535415355153561535715358153591536015361153621536315364153651536615367153681536915370153711537215373153741537515376153771537815379153801538115382153831538415385153861538715388153891539015391153921539315394153951539615397153981539915400154011540215403154041540515406154071540815409154101541115412154131541415415154161541715418154191542015421154221542315424154251542615427154281542915430154311543215433154341543515436154371543815439154401544115442154431544415445154461544715448154491545015451154521545315454154551545615457154581545915460154611546215463154641546515466154671546815469154701547115472154731547415475154761547715478154791548015481154821548315484154851548615487154881548915490154911549215493154941549515496154971549815499155001550115502155031550415505155061550715508155091551015511155121551315514155151551615517
  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. \end_preamble
  39. \use_default_options true
  40. \begin_modules
  41. todonotes
  42. logicalmkup
  43. \end_modules
  44. \maintain_unincluded_children false
  45. \language english
  46. \language_package default
  47. \inputencoding utf8
  48. \fontencoding default
  49. \font_roman "default" "default"
  50. \font_sans "default" "default"
  51. \font_typewriter "default" "default"
  52. \font_math "auto" "auto"
  53. \font_default_family default
  54. \use_non_tex_fonts false
  55. \font_sc false
  56. \font_osf false
  57. \font_sf_scale 100 100
  58. \font_tt_scale 100 100
  59. \use_microtype false
  60. \use_dash_ligatures true
  61. \graphics default
  62. \default_output_format pdf4
  63. \output_sync 0
  64. \bibtex_command biber
  65. \index_command default
  66. \paperfontsize 12
  67. \spacing double
  68. \use_hyperref true
  69. \pdf_bookmarks true
  70. \pdf_bookmarksnumbered false
  71. \pdf_bookmarksopen false
  72. \pdf_bookmarksopenlevel 1
  73. \pdf_breaklinks false
  74. \pdf_pdfborder false
  75. \pdf_colorlinks false
  76. \pdf_backref false
  77. \pdf_pdfusetitle true
  78. \papersize letterpaper
  79. \use_geometry true
  80. \use_package amsmath 1
  81. \use_package amssymb 1
  82. \use_package cancel 1
  83. \use_package esint 1
  84. \use_package mathdots 1
  85. \use_package mathtools 1
  86. \use_package mhchem 1
  87. \use_package stackrel 1
  88. \use_package stmaryrd 1
  89. \use_package undertilde 1
  90. \cite_engine biblatex
  91. \cite_engine_type authoryear
  92. \biblio_style plain
  93. \biblatex_bibstyle authoryear
  94. \biblatex_citestyle numeric
  95. \use_bibtopic false
  96. \use_indices false
  97. \paperorientation portrait
  98. \suppress_date false
  99. \justification true
  100. \use_refstyle 1
  101. \use_minted 0
  102. \index Index
  103. \shortcut idx
  104. \color #008000
  105. \end_index
  106. \leftmargin 1.5in
  107. \topmargin 1in
  108. \rightmargin 1in
  109. \bottommargin 1in
  110. \secnumdepth 3
  111. \tocdepth 3
  112. \paragraph_separation indent
  113. \paragraph_indentation default
  114. \is_math_indent 0
  115. \math_numbering_side default
  116. \quotes_style english
  117. \dynamic_quotes 0
  118. \papercolumns 1
  119. \papersides 1
  120. \paperpagestyle default
  121. \tracking_changes false
  122. \output_changes false
  123. \html_math_output 0
  124. \html_css_as_file 0
  125. \html_be_strict false
  126. \end_header
  127. \begin_body
  128. \begin_layout Title
  129. Bioinformatic analysis of complex, high-throughput genomic and epigenomic
  130. data in the context of immunology and transplant rejection
  131. \end_layout
  132. \begin_layout Author
  133. A thesis presented
  134. \begin_inset Newline newline
  135. \end_inset
  136. by
  137. \begin_inset Newline newline
  138. \end_inset
  139. Ryan C.
  140. Thompson
  141. \begin_inset Newline newline
  142. \end_inset
  143. to
  144. \begin_inset Newline newline
  145. \end_inset
  146. The Scripps Research Institute Graduate Program
  147. \begin_inset Newline newline
  148. \end_inset
  149. in partial fulfillment of the requirements for the degree of
  150. \begin_inset Newline newline
  151. \end_inset
  152. Doctor of Philosophy in the subject of Biology
  153. \begin_inset Newline newline
  154. \end_inset
  155. for
  156. \begin_inset Newline newline
  157. \end_inset
  158. The Scripps Research Institute
  159. \begin_inset Newline newline
  160. \end_inset
  161. La Jolla, California
  162. \end_layout
  163. \begin_layout Date
  164. October 2019
  165. \end_layout
  166. \begin_layout Standard
  167. [Copyright notice]
  168. \end_layout
  169. \begin_layout Standard
  170. [Thesis acceptance form]
  171. \end_layout
  172. \begin_layout Standard
  173. [Dedication]
  174. \end_layout
  175. \begin_layout Standard
  176. [Acknowledgements]
  177. \end_layout
  178. \begin_layout Standard
  179. \begin_inset CommandInset toc
  180. LatexCommand tableofcontents
  181. \end_inset
  182. \end_layout
  183. \begin_layout Standard
  184. \begin_inset FloatList table
  185. \end_inset
  186. \end_layout
  187. \begin_layout Standard
  188. \begin_inset FloatList figure
  189. \end_inset
  190. \end_layout
  191. \begin_layout Standard
  192. [List of Abbreviations]
  193. \end_layout
  194. \begin_layout List of TODOs
  195. \end_layout
  196. \begin_layout Standard
  197. \begin_inset Flex TODO Note (inline)
  198. status open
  199. \begin_layout Plain Layout
  200. Check all figures to make sure they fit on the page with their legends.
  201. \end_layout
  202. \end_inset
  203. \end_layout
  204. \begin_layout Standard
  205. \begin_inset Flex TODO Note (inline)
  206. status open
  207. \begin_layout Plain Layout
  208. Search and replace: naive -> naïve
  209. \end_layout
  210. \end_inset
  211. \end_layout
  212. \begin_layout Standard
  213. \begin_inset Flex TODO Note (inline)
  214. status open
  215. \begin_layout Plain Layout
  216. Look into auto-generated nomenclature list:
  217. \begin_inset CommandInset href
  218. LatexCommand href
  219. target "https://wiki.lyx.org/Tips/Nomenclature"
  220. \end_inset
  221. .
  222. Otherwise, do a manual pass for all abbreviations at the end.
  223. Do nomenclature/abbreviations independently for each chapter.
  224. \end_layout
  225. \end_inset
  226. \end_layout
  227. \begin_layout Standard
  228. \begin_inset Flex TODO Note (inline)
  229. status open
  230. \begin_layout Plain Layout
  231. Make all descriptions consistent in terms of
  232. \begin_inset Quotes eld
  233. \end_inset
  234. we did X
  235. \begin_inset Quotes erd
  236. \end_inset
  237. vs
  238. \begin_inset Quotes eld
  239. \end_inset
  240. I did X
  241. \begin_inset Quotes erd
  242. \end_inset
  243. vs
  244. \begin_inset Quotes eld
  245. \end_inset
  246. X was done
  247. \begin_inset Quotes erd
  248. \end_inset
  249. .
  250. \end_layout
  251. \end_inset
  252. \end_layout
  253. \begin_layout Chapter*
  254. Abstract
  255. \end_layout
  256. \begin_layout Standard
  257. \begin_inset Note Note
  258. status open
  259. \begin_layout Plain Layout
  260. It is included as an integral part of the thesis and should immediately
  261. precede the introduction.
  262. \end_layout
  263. \begin_layout Plain Layout
  264. Preparing your Abstract.
  265. Your abstract (a succinct description of your work) is limited to 350 words.
  266. UMI will shorten it if they must; please do not exceed the limit.
  267. \end_layout
  268. \begin_layout Itemize
  269. Include pertinent place names, names of persons (in full), and other proper
  270. nouns.
  271. These are useful in automated retrieval.
  272. \end_layout
  273. \begin_layout Itemize
  274. Display symbols, as well as foreign words and phrases, clearly and accurately.
  275. Include transliterations for characters other than Roman and Greek letters
  276. and Arabic numerals.
  277. Include accents and diacritical marks.
  278. \end_layout
  279. \begin_layout Itemize
  280. Do not include graphs, charts, tables, or illustrations in your abstract.
  281. \end_layout
  282. \end_inset
  283. \end_layout
  284. \begin_layout Standard
  285. \begin_inset Flex TODO Note (inline)
  286. status open
  287. \begin_layout Plain Layout
  288. Obviously the abstract gets written last.
  289. \end_layout
  290. \end_inset
  291. \end_layout
  292. \begin_layout Chapter*
  293. Notes to draft readers
  294. \end_layout
  295. \begin_layout Standard
  296. Thank you so much for agreeing to read my thesis and give me feedback on
  297. it.
  298. What you are currently reading is a rough draft, in need of many revisions.
  299. You can always find the latest version at
  300. \begin_inset CommandInset href
  301. LatexCommand href
  302. target "https://mneme.dedyn.io/~ryan/Thesis/thesis.pdf"
  303. literal "false"
  304. \end_inset
  305. .
  306. the PDF at this link is updated periodically with my latest revisions,
  307. but you can just download the current version and give me feedback on that.
  308. Don't worry about keeping up with the updates.
  309. \end_layout
  310. \begin_layout Standard
  311. As for what feedback I'm looking for, first of all, don't waste your time
  312. marking spelling mistakes and such.
  313. I haven't run a spell checker on it yet, so let me worry about that.
  314. Also, I'm aware that many abbreviations are not properly introduced the
  315. first time they are used, so don't worry about that either.
  316. However, if you see any glaring formatting issues, such as a figure being
  317. too large and getting cut off at the edge of the page, please note them.
  318. In addition, if any of the text in the figures is too small, please note
  319. that as well.
  320. \end_layout
  321. \begin_layout Standard
  322. Beyond that, what I'm mainly interested in is feedback on the content.
  323. For example: does the introduction flow logically, and does it provide
  324. enough background to understand the other chapters? Does each chapter make
  325. it clear what work and analyses I have done? Do the figures clearly communicate
  326. the results I'm trying to show? Do you feel that the claims in the results
  327. and discussion sections are well-supported? There's no need to suggest
  328. improvements; just note areas that you feel need improvement.
  329. Additionally, while I am well aware that Chapter 1 (the introduction) contains
  330. many un-cited claims, all the other chapters (2,3, and 4)
  331. \emph on
  332. should
  333. \emph default
  334. be fully cited.
  335. So if you notice any un-cited claims in those chapters, please flag them
  336. for my attention.
  337. Similarly, if you discover any factual errors, please note them as well.
  338. \end_layout
  339. \begin_layout Standard
  340. You can provide your feedback in whatever way is most convenient to you.
  341. You could mark up this PDF with highlights and notes, then send it back
  342. to me.
  343. Or you could collect your comments in a separate text file and send that
  344. to me, or whatever else you like.
  345. However, if you send me your feedback in a separate document, please note
  346. a section/figure/table number for each comment, and
  347. \emph on
  348. also
  349. \emph default
  350. send me the exact PDF that you read so I can reference it while reading
  351. your comments, since as mentioned above, the current version I'm working
  352. on will have changed by that point (which might include shuffling sections
  353. and figures around, changing their numbers).
  354. One last thing: you'll see a bunch of text in orange boxes throughout the
  355. PDF.
  356. These are notes to myself about things that need to be fixed later, so
  357. if you see a problem noted in an orange box, that means I'm already aware
  358. of it, and there's no need to comment on it.
  359. \end_layout
  360. \begin_layout Standard
  361. My thesis is due Thursday, October 10th, so in order to be useful to me,
  362. I'll need your feedback at least a few days before that, ideally by Monday,
  363. October 7th.
  364. If you have limited time and are unable to get through the whole thesis,
  365. please focus your efforts on Chapters 1 and 2, since those are the roughest
  366. and most in need of revision.
  367. Chapter 3 is fairly short and straightforward, and Chapter 4 is an adaptation
  368. of a paper that's already been through a few rounds of revision, so they
  369. should be a lot tighter.
  370. If you can't spare any time between now and then, or if something unexpected
  371. comes up, I understand.
  372. Just let me know.
  373. \end_layout
  374. \begin_layout Standard
  375. Thanks again for your help, and happy reading!
  376. \end_layout
  377. \begin_layout Chapter
  378. Introduction
  379. \end_layout
  380. \begin_layout Section
  381. Background & Significance
  382. \end_layout
  383. \begin_layout Subsection
  384. Biological motivation
  385. \end_layout
  386. \begin_layout Standard
  387. \begin_inset Flex TODO Note (inline)
  388. status open
  389. \begin_layout Plain Layout
  390. Rethink the subsection organization after the intro is written.
  391. \end_layout
  392. \end_inset
  393. \end_layout
  394. \begin_layout Standard
  395. \begin_inset Flex TODO Note (inline)
  396. status open
  397. \begin_layout Plain Layout
  398. Citations are needed all over the place.
  399. A lot of this is knowledge I've just absorbed from years of conversation
  400. in the Salomon lab, without ever having seen a citation for it.
  401. \end_layout
  402. \end_inset
  403. \end_layout
  404. \begin_layout Subsubsection
  405. Rejection is the major long-term threat to organ and tissue allografts
  406. \end_layout
  407. \begin_layout Standard
  408. Organ and tissue transplants are a life-saving treatment for people who
  409. have lost the function of an important organ [CITE?].
  410. In some cases, it is possible to transplant a patient's own tissue from
  411. one area of their body to another, referred to as an autograft.
  412. This is common for tissues that are distributed throughout many areas of
  413. the body, such as skin and bone.
  414. However, in cases of organ failure, there is no functional self tissue
  415. remaining, and a transplant from another person – a donor – is required.
  416. This is referred to as an allograft.
  417. \end_layout
  418. \begin_layout Standard
  419. \begin_inset Flex TODO Note (inline)
  420. status open
  421. \begin_layout Plain Layout
  422. Possible citation for degree of generic variability:
  423. \begin_inset CommandInset href
  424. LatexCommand href
  425. target "https://www.ncbi.nlm.nih.gov/pubmed/22424236?dopt=Abstract"
  426. \end_inset
  427. \end_layout
  428. \end_inset
  429. \end_layout
  430. \begin_layout Standard
  431. \begin_inset Flex TODO Note (inline)
  432. status open
  433. \begin_layout Plain Layout
  434. How much mechanistic detail is needed here? My work doesn't really go into
  435. specific rejection mechanisms, so I think it's best to keep it basic.
  436. \end_layout
  437. \end_inset
  438. \end_layout
  439. \begin_layout Standard
  440. Because an allograft comes from a different person, it is genetically distinct
  441. from the rest of the recipient's body.
  442. Some genetic variants occur in protein coding regions, resulting in protein
  443. products that differ from the equivalent proteins in the graft recipient's
  444. own tissue.
  445. As a result, without intervention, the recipient's immune system will eventuall
  446. y identify the graft as foreign tissue and begin attacking it, eventually
  447. resulting in failure and death of the graft, a process referred to as transplan
  448. t rejection.
  449. Rejection is the most significant challenge to the long-term health and
  450. survival of an allograft [CITE?].
  451. Like any adaptive immune response, graft rejection generally occurs via
  452. two broad mechanisms: cellular immunity, in which CD8+ T-cells recognizing
  453. graft-specific antigens induce apoptosis in the graft cells; and humoral
  454. immunity, in which B-cells produce antibodies that bind to graft proteins
  455. and direct an immune response against the graft [CITE?].
  456. In either case, rejection shows most of the typical hallmarks of an adaptive
  457. immune response, in particular mediation by CD4+ T-cells and formation
  458. of immune memory.
  459. \end_layout
  460. \begin_layout Subsubsection
  461. Diagnosis and treatment of allograft rejection is a major challenge
  462. \end_layout
  463. \begin_layout Standard
  464. To prevent rejection, allograft recipients are treated with immune suppressive
  465. drugs [CITE?].
  466. The goal is to achieve sufficient suppression of the immune system to prevent
  467. rejection of the graft without compromising the ability of the immune system
  468. to raise a normal response against infection.
  469. As such, a delicate balance must be struck: insufficient immune suppression
  470. may lead to rejection and ultimately loss of the graft; excessive suppression
  471. leaves the patient vulnerable to life-threatening opportunistic infections.
  472. Because every patient is different, immune suppression must be tailored
  473. for each patient.
  474. Furthermore, immune suppression must be tuned over time, as the immune
  475. system's activity is not static, nor is it held in a steady state [CITE?].
  476. In order to properly adjust the dosage of immune suppression drugs, it
  477. is necessary to monitor the health of the transplant and increase the dosage
  478. if evidence of rejection is observed.
  479. \end_layout
  480. \begin_layout Standard
  481. However, diagnosis of rejection is a significant challenge.
  482. Early diagnosis is essential in order to step up immune suppression before
  483. the immune system damages the graft beyond recovery [CITE?].
  484. The current gold standard test for graft rejection is a tissue biopsy,
  485. examined for visible signs of rejection by a trained histologist [CITE?].
  486. When a patient shows symptoms of possible rejection, a
  487. \begin_inset Quotes eld
  488. \end_inset
  489. for cause
  490. \begin_inset Quotes erd
  491. \end_inset
  492. biopsy is performed to confirm the diagnosis, and immune suppression is
  493. adjusted as necessary.
  494. However, in many cases, the early stages of rejection are asymptomatic,
  495. known as
  496. \begin_inset Quotes eld
  497. \end_inset
  498. sub-clinical
  499. \begin_inset Quotes erd
  500. \end_inset
  501. rejection [CITE?].
  502. In light of this, is is now common to perform
  503. \begin_inset Quotes eld
  504. \end_inset
  505. protocol biopsies
  506. \begin_inset Quotes erd
  507. \end_inset
  508. at specific times after transplantation of a graft, even if no symptoms
  509. of rejection are apparent, in addition to
  510. \begin_inset Quotes eld
  511. \end_inset
  512. for cause
  513. \begin_inset Quotes erd
  514. \end_inset
  515. biopsies
  516. \begin_inset CommandInset citation
  517. LatexCommand cite
  518. key "Wilkinson2006"
  519. literal "false"
  520. \end_inset
  521. .
  522. \end_layout
  523. \begin_layout Standard
  524. However, biopsies have a number of downsides that limit their effectiveness
  525. as a diagnostic tool.
  526. First, the need for manual inspection by a histologist means that diagnosis
  527. is subject to the biases of the particular histologist examining the biopsy
  528. [CITE?].
  529. In marginal cases, two different histologists may give two different diagnoses
  530. to the same biopsy.
  531. Second, a biopsy can only evaluate if rejection is occurring in the section
  532. of the graft from which the tissue was extracted.
  533. If rejection is localized to one section of the graft and the tissue is
  534. extracted from a different section, a false negative diagnosis may result.
  535. Most importantly, extraction of tissue from a graft is invasive and is
  536. treated as an injury by the body, which results in inflammation that in
  537. turn promotes increased immune system activity [CITE?].
  538. Hence, the invasiveness of biopsies severely limits the frequency with
  539. which they can safely be performed.
  540. Typically, protocol biopsies are not scheduled more than about once per
  541. month
  542. \begin_inset CommandInset citation
  543. LatexCommand cite
  544. key "Wilkinson2006"
  545. literal "false"
  546. \end_inset
  547. .
  548. A less invasive diagnostic test for rejection would bring manifold benefits.
  549. Such a test would enable more frequent testing and therefore earlier detection
  550. of rejection events.
  551. In addition, having a larger pool of historical data for a given patient
  552. would make it easier to evaluate when a given test is outside the normal
  553. parameters for that specific patient, rather than relying on normal ranges
  554. for the population as a whole.
  555. Lastly, the accumulated data from more frequent tests would be a boon to
  556. the transplant research community.
  557. Beyond simply providing more data overall, the better time granularity
  558. of the tests will enable studying the progression of a rejection event
  559. on the scale of days to weeks, rather than months.
  560. \end_layout
  561. \begin_layout Subsubsection
  562. Memory cells are resistant to immune suppression
  563. \end_layout
  564. \begin_layout Standard
  565. One of the defining features of the adaptive immune system is immune memory:
  566. the ability of the immune system to recognize a previously encountered
  567. foreign antigen and respond more quickly and more strongly to that antigen
  568. in subsequent encounters.
  569. When the immune system first encounters a new antigen, the lymphocytes
  570. that respond are known as naive cells – T-cells and B-cells that have never
  571. detected their target antigens before.
  572. Once activated by their specific antigen presented by an antigen-presenting
  573. cell in the proper co-stimulatory context, naive cells differentiate into
  574. effector cells that carry out their respective functions in targeting and
  575. destroying the source of the foreign antigen.
  576. The requirement for co-stimulation is an important feature of naive cells
  577. that limits
  578. \begin_inset Quotes eld
  579. \end_inset
  580. false positive
  581. \begin_inset Quotes erd
  582. \end_inset
  583. immune responses, because antigen-presenting cells usually only express
  584. the proper co-stimulation after detecting evidence of an infection, such
  585. as the presence of common bacterial cell components or inflamed tissue.
  586. Most effector cells die after the foreign antigen is cleared, since they
  587. are no longer needed, but some remain and differentiate into memory cells.
  588. Like naive cells, memory cells respond to detection of their specific antigen
  589. by differentiating into effector cells, ready to fight an infection.
  590. However, unlike naive cells, memory cells do not require the same degree
  591. of co-stimulatory signaling for activation, and once activated, they proliferat
  592. e and differentiate into effector cells more quickly than naive cells do.
  593. \end_layout
  594. \begin_layout Standard
  595. In the context of a pathogenic infection, immune memory is a major advantage,
  596. allowing an organism to rapidly fight off a previously encountered pathogen
  597. much more quickly and effectively than the first time it was encountered.
  598. However, if effector cells that recognize an antigen from an allograft
  599. are allowed to differentiate into memory cells, preventing rejection of
  600. the graft becomes much more difficult.
  601. Many immune suppression drugs work by interfering with the co-stimulation
  602. that naive cells require in order to mount an immune response [CITE?].
  603. Since memory cells do not require this co-stimulation, these drugs are
  604. not effective at suppressing an immune response that is mediated by memory
  605. cells.
  606. Secondly, because memory cells are able to mount a stronger and faster
  607. response to an antigen, all else being equal they require stronger immune
  608. suppression than naive cells to prevent an immune response.
  609. However, immune suppression affects the entire immune system, not just
  610. cells recognizing a specific antigen, so increasing the dosage of immune
  611. suppression drugs also increases the risk of complications from a compromised
  612. immune system, such as opportunistic infections.
  613. While the differences in cell surface markers between naive and memory
  614. cells have been fairly well characterized, the internal regulatory mechanisms
  615. that allow memory cells to respond more quickly and without co-stimulation
  616. are still poorly understood.
  617. In order to develop methods of immune suppression that either prevent the
  618. formation of memory cells or work more effectively against memory cells,
  619. the mechanisms of immune memory formation and regulation must be better
  620. understood.
  621. \end_layout
  622. \begin_layout Subsection
  623. Overview of bioinformatic analysis methods
  624. \end_layout
  625. \begin_layout Standard
  626. \begin_inset Flex TODO Note (inline)
  627. status open
  628. \begin_layout Plain Layout
  629. Also cite: R, Bioconductor, snakemake, python, pandas, bedtools, bowtie2,
  630. hisat2, STAR, samtools, sra-toolkit, picard tools
  631. \end_layout
  632. \end_inset
  633. \end_layout
  634. \begin_layout Standard
  635. The studies presented in this work all involve the analysis of high-throughput
  636. genomic and epigenomic data.
  637. These data present many unique analysis challenges, and a wide array of
  638. software tools are available to analyze them.
  639. This section presents an overview of the methods used, including what problems
  640. they solve, what assumptions they make, and a basic description of how
  641. they work.
  642. \end_layout
  643. \begin_layout Subsubsection
  644. \begin_inset Flex Code
  645. status open
  646. \begin_layout Plain Layout
  647. Limma
  648. \end_layout
  649. \end_inset
  650. : The standard linear modeling framework for genomics
  651. \end_layout
  652. \begin_layout Standard
  653. Linear models are a generalization of the
  654. \begin_inset Formula $t$
  655. \end_inset
  656. -test and ANOVA to arbitrarily complex experimental designs
  657. \begin_inset CommandInset citation
  658. LatexCommand cite
  659. key "chambers:1992"
  660. literal "false"
  661. \end_inset
  662. .
  663. In a typical linear model, there is one dependent variable observation
  664. per sample and a large number of samples.
  665. For example, in a linear model of height as a function of age and sex,
  666. there is one height measurement per person.
  667. However, when analyzing genomic data, each sample consists of observations
  668. of thousands of dependent variables.
  669. For example, in an RNA-seq experiment, the dependent variables may be the
  670. count of RNA-seq reads for each annotated gene.
  671. In abstract terms, each dependent variable being measured is referred to
  672. as a feature.
  673. The simplest approach to analyzing such data would be to fit the same model
  674. independently to each feature.
  675. However, this is undesirable for most genomics data sets.
  676. Genomics assays like high-throughput sequencing are expensive, and often
  677. the process of generating the samples is also quite expensive and time-consumin
  678. g.
  679. This expense limits the sample sizes typically employed in genomics experiments
  680. , and as a result the statistical power of the linear model for each individual
  681. feature is likewise limited.
  682. However, because thousands of features from the same samples are analyzed
  683. together, there is an opportunity to improve the statistical power of the
  684. analysis by exploiting shared patterns of variation across features.
  685. This is the core feature of
  686. \begin_inset Flex Code
  687. status open
  688. \begin_layout Plain Layout
  689. limma
  690. \end_layout
  691. \end_inset
  692. , a linear modeling framework designed for genomic data.
  693. \begin_inset Flex Code
  694. status open
  695. \begin_layout Plain Layout
  696. Limma
  697. \end_layout
  698. \end_inset
  699. is typically used to analyze expression microarray data, and more recently
  700. RNA-seq data, but it can also be used to analyze any other data for which
  701. linear modeling is appropriate.
  702. \end_layout
  703. \begin_layout Standard
  704. The central challenge when fitting a linear model is to estimate the variance
  705. of the data accurately.
  706. Out of all parameters required to evaluate statistical significance of
  707. an effect, the variance is the most difficult to estimate when sample sizes
  708. are small.
  709. A single shared variance could be estimated for all of the features together,
  710. and this estimate would be very stable, in contrast to the individual feature
  711. variance estimates.
  712. However, this would require the assumption that every feature is equally
  713. variable, which is known to be false for most genomic data sets.
  714. \begin_inset Flex Code
  715. status open
  716. \begin_layout Plain Layout
  717. limma
  718. \end_layout
  719. \end_inset
  720. offers a compromise between these two extremes by using a method called
  721. empirical Bayes moderation to
  722. \begin_inset Quotes eld
  723. \end_inset
  724. squeeze
  725. \begin_inset Quotes erd
  726. \end_inset
  727. the distribution of estimated variances toward a single common value that
  728. represents the variance of an average feature in the data
  729. \begin_inset CommandInset citation
  730. LatexCommand cite
  731. key "Smyth2004"
  732. literal "false"
  733. \end_inset
  734. .
  735. While the individual feature variance estimates are not stable, the common
  736. variance estimate for the entire data set is quite stable, so using a combinati
  737. on of the two yields a variance estimate for each feature with greater precision
  738. than the individual feature variances.
  739. The trade-off for this improvement is that squeezing each estimated variance
  740. toward the common value introduces some bias – the variance will be underestima
  741. ted for features with high variance and overestimated for features with
  742. low variance.
  743. Essentially,
  744. \begin_inset Flex Code
  745. status open
  746. \begin_layout Plain Layout
  747. limma
  748. \end_layout
  749. \end_inset
  750. assumes that extreme variances are less common than variances close to
  751. the common value.
  752. The variance estimates from this empirical Bayes procedure are shown empiricall
  753. y to yield greater statistical power than either the individual feature
  754. variances or the single common value.
  755. \end_layout
  756. \begin_layout Standard
  757. On top of this core framework,
  758. \begin_inset Flex Code
  759. status open
  760. \begin_layout Plain Layout
  761. limma
  762. \end_layout
  763. \end_inset
  764. also implements many other enhancements that, further relax the assumptions
  765. of the model and extend the scope of what kinds of data it can analyze.
  766. Instead of squeezing toward a single common variance value,
  767. \begin_inset Flex Code
  768. status open
  769. \begin_layout Plain Layout
  770. limma
  771. \end_layout
  772. \end_inset
  773. can model the common variance as a function of a covariate, such as average
  774. expression
  775. \begin_inset CommandInset citation
  776. LatexCommand cite
  777. key "Law2013"
  778. literal "false"
  779. \end_inset
  780. .
  781. This is essential for RNA-seq data, where higher gene counts yield more
  782. precise expression measurements and therefore smaller variances than low-count
  783. genes.
  784. While linear models typically assume that all samples have equal variance,
  785. \begin_inset Flex Code
  786. status open
  787. \begin_layout Plain Layout
  788. limma
  789. \end_layout
  790. \end_inset
  791. is able to relax this assumption by identifying and down-weighting samples
  792. that diverge more strongly from the linear model across many features
  793. \begin_inset CommandInset citation
  794. LatexCommand cite
  795. key "Ritchie2006,Liu2015"
  796. literal "false"
  797. \end_inset
  798. .
  799. In addition,
  800. \begin_inset Flex Code
  801. status open
  802. \begin_layout Plain Layout
  803. limma
  804. \end_layout
  805. \end_inset
  806. is also able to fit simple mixed models incorporating one random effect
  807. in addition to the fixed effects represented by an ordinary linear model
  808. \begin_inset CommandInset citation
  809. LatexCommand cite
  810. key "Smyth2005a"
  811. literal "false"
  812. \end_inset
  813. .
  814. Once again,
  815. \begin_inset Flex Code
  816. status open
  817. \begin_layout Plain Layout
  818. limma
  819. \end_layout
  820. \end_inset
  821. shares information between features to obtain a robust estimate for the
  822. random effect correlation.
  823. \end_layout
  824. \begin_layout Subsubsection
  825. edgeR provides
  826. \begin_inset Flex Code
  827. status open
  828. \begin_layout Plain Layout
  829. limma
  830. \end_layout
  831. \end_inset
  832. -like analysis features for count data
  833. \end_layout
  834. \begin_layout Standard
  835. Although
  836. \begin_inset Flex Code
  837. status open
  838. \begin_layout Plain Layout
  839. limma
  840. \end_layout
  841. \end_inset
  842. can be applied to read counts from RNA-seq data, it is less suitable for
  843. counts from ChIP-seq data, which tend to be much smaller and therefore
  844. violate the assumption of a normal distribution more severely.
  845. For all count-based data, the
  846. \begin_inset Flex Code
  847. status open
  848. \begin_layout Plain Layout
  849. edgeR
  850. \end_layout
  851. \end_inset
  852. package works similarly to
  853. \begin_inset Flex Code
  854. status open
  855. \begin_layout Plain Layout
  856. limma
  857. \end_layout
  858. \end_inset
  859. , but uses a generalized linear model instead of a linear model.
  860. The most important difference is that the GLM in
  861. \begin_inset Flex Code
  862. status open
  863. \begin_layout Plain Layout
  864. edgeR
  865. \end_layout
  866. \end_inset
  867. models the counts directly using a negative binomial distribution rather
  868. than modeling the normalized log counts using a normal distribution
  869. \begin_inset CommandInset citation
  870. LatexCommand cite
  871. key "Chen2014,McCarthy2012,Robinson2010a"
  872. literal "false"
  873. \end_inset
  874. .
  875. The negative binomial is a good fit for count data because it can be derived
  876. as a gamma-distributed mixture of Poisson distributions.
  877. The Poisson distribution accurately represents the distribution of counts
  878. expected for a given gene abundance, and the gamma distribution is then
  879. used to represent the variation in gene abundance between biological replicates.
  880. For this reason, the square root of the dispersion parameter of the negative
  881. binomial is sometimes referred to as the biological coefficient of variation,
  882. since it represents the variability that was present in the samples prior
  883. to the Poisson
  884. \begin_inset Quotes eld
  885. \end_inset
  886. noise
  887. \begin_inset Quotes erd
  888. \end_inset
  889. that was generated by the random sampling of reads in proportion to feature
  890. abundances.
  891. The choice of a gamma distribution is arbitrary and motivated by mathematical
  892. convenience, since a gamma-Poisson mixture yields the numerically tractable
  893. negative binomial distribution.
  894. Thus,
  895. \begin_inset Flex Code
  896. status open
  897. \begin_layout Plain Layout
  898. edgeR
  899. \end_layout
  900. \end_inset
  901. assumes
  902. \emph on
  903. a prioi
  904. \emph default
  905. that the variation in abundances between replicates follows a gamma distribution.
  906. For differential abundance testing,
  907. \begin_inset Flex Code
  908. status open
  909. \begin_layout Plain Layout
  910. edgeR
  911. \end_layout
  912. \end_inset
  913. offers a likelihood ratio test, but more recently recommends a quasi-likelihood
  914. test that properly factors the uncertainty in variance estimation into
  915. the statistical significance for each feature
  916. \begin_inset CommandInset citation
  917. LatexCommand cite
  918. key "Lund2012"
  919. literal "false"
  920. \end_inset
  921. .
  922. \end_layout
  923. \begin_layout Subsubsection
  924. ChIP-seq Peak calling
  925. \end_layout
  926. \begin_layout Standard
  927. Unlike RNA-seq data, in which gene annotations provide a well-defined set
  928. of discrete genomic regions in which to count reads, ChIP-seq reads can
  929. potentially occur anywhere in the genome.
  930. However, most genome regions will not contain significant ChIP-seq read
  931. coverage, and analyzing every position in the entire genome is statistically
  932. and computationally infeasible, so it is necessary to identify regions
  933. of interest inside which ChIP-seq reads will be counted and analyzed.
  934. One option is to define a set of interesting regions
  935. \emph on
  936. a priori
  937. \emph default
  938. , for example by defining a promoter region for each annotated gene.
  939. However, it is also possible to use the ChIP-seq data itself to identify
  940. regions with ChIP-seq read coverage significantly above the background
  941. level, known as peaks.
  942. \end_layout
  943. \begin_layout Standard
  944. There are generally two kinds of peaks that can be identified: narrow peaks
  945. and broadly enriched regions.
  946. Proteins like transcription factors that bind specific sites in the genome
  947. typically show most of their ChIP-seq read coverage at these specific sites
  948. and very little coverage anywhere else.
  949. Because the footprint of the protein is consistent wherever it binds, each
  950. peak has a consistent width, typically tens to hundreds of base pairs,
  951. representing the length of DNA that it binds to.
  952. Algorithms like MACS exploit this pattern to identify specific loci at
  953. which such
  954. \begin_inset Quotes eld
  955. \end_inset
  956. narrow peaks
  957. \begin_inset Quotes erd
  958. \end_inset
  959. occur by looking for the characteristic peak shape in the ChIP-seq coverage
  960. rising above the surrounding background coverage
  961. \begin_inset CommandInset citation
  962. LatexCommand cite
  963. key "Zhang2008"
  964. literal "false"
  965. \end_inset
  966. .
  967. In contrast, some proteins, chief among them histones, do not bind only
  968. at a small number of specific sites, but rather bind potentially almost
  969. everywhere in the entire genome.
  970. When looking at histone marks, adjacent histones tend to be similarly marked,
  971. and a given mark may be present on an arbitrary number of consecutive histones
  972. along the genome.
  973. Hence, there is no consistent
  974. \begin_inset Quotes eld
  975. \end_inset
  976. footprint size
  977. \begin_inset Quotes erd
  978. \end_inset
  979. for ChIP-seq peaks based on histone marks, and peaks typically span many
  980. histones.
  981. Hence, typical peaks span many hundreds or even thousands of base pairs.
  982. Instead of identifying specific loci of strong enrichment, algorithms like
  983. SICER assume that peaks are represented in the ChIP-seq data by modest
  984. enrichment above background occurring across broad regions, and they attempt
  985. to identify the extent of those regions
  986. \begin_inset CommandInset citation
  987. LatexCommand cite
  988. key "Zang2009"
  989. literal "false"
  990. \end_inset
  991. .
  992. In all cases, better results are obtained if the local background coverage
  993. level can be estimated from ChIP-seq input samples, since various biases
  994. can result in uneven background coverage.
  995. \end_layout
  996. \begin_layout Standard
  997. Regardless of the type of peak identified, it is important to identify peaks
  998. that occur consistently across biological replicates.
  999. The ENCODE project has developed a method called irreproducible discovery
  1000. rate for this purpose
  1001. \begin_inset CommandInset citation
  1002. LatexCommand cite
  1003. key "Li2006"
  1004. literal "false"
  1005. \end_inset
  1006. .
  1007. The IDR is defined as the probability that a peak identified in one biological
  1008. replicate will
  1009. \emph on
  1010. not
  1011. \emph default
  1012. also be identified in a second replicate.
  1013. Where the more familiar false discovery rate measures the degree of corresponde
  1014. nce between a data-derived ranked list and the true list of significant
  1015. features, IDR instead measures the degree of correspondence between two
  1016. ranked lists derived from different data.
  1017. IDR assumes that the highest-ranked features are
  1018. \begin_inset Quotes eld
  1019. \end_inset
  1020. signal
  1021. \begin_inset Quotes erd
  1022. \end_inset
  1023. peaks that tend to be listed in the same order in both lists, while the
  1024. lowest-ranked features are essentially noise peaks, listed in random order
  1025. with no correspondence between the lists.
  1026. IDR attempts to locate the
  1027. \begin_inset Quotes eld
  1028. \end_inset
  1029. crossover point
  1030. \begin_inset Quotes erd
  1031. \end_inset
  1032. between the signal and the noise by determining how far down the list the
  1033. correspondence between feature ranks breaks down.
  1034. \end_layout
  1035. \begin_layout Standard
  1036. In addition to other considerations, if called peaks are to be used as regions
  1037. of interest for differential abundance analysis, then care must be taken
  1038. to call peaks in a way that is blind to differential abundance between
  1039. experimental conditions, or else the statistical significance calculations
  1040. for differential abundance will overstate their confidence in the results.
  1041. The
  1042. \begin_inset Flex Code
  1043. status open
  1044. \begin_layout Plain Layout
  1045. csaw
  1046. \end_layout
  1047. \end_inset
  1048. package provides guidelines for calling peaks in this way: peaks
  1049. are called based on a combination of all ChIP-seq reads from all experimental
  1050. conditions, so that the identified peaks are based on the average abundance
  1051. across all conditions, which is independent of any differential abundance
  1052. between conditions
  1053. \begin_inset CommandInset citation
  1054. LatexCommand cite
  1055. key "Lun2015a"
  1056. literal "false"
  1057. \end_inset
  1058. .
  1059. \end_layout
  1060. \begin_layout Subsubsection
  1061. Normalization of high-throughput data is non-trivial and application-dependent
  1062. \end_layout
  1063. \begin_layout Standard
  1064. High-throughput data sets invariably require some kind of normalization
  1065. before further analysis can be conducted.
  1066. In general, the goal of normalization is to remove effects in the data
  1067. that are caused by technical factors that have nothing to do with the biology
  1068. being studied.
  1069. \end_layout
  1070. \begin_layout Standard
  1071. For Affymetrix expression arrays, the standard normalization algorithm used
  1072. in most analyses is Robust Multichip Average (RMA) [CITE].
  1073. RMA is designed with the assumption that some fraction of probes on each
  1074. array will be artifactual and takes advantage of the fact that each gene
  1075. is represented by multiple probes by implementing normalization and summarizati
  1076. on steps that are robust against outlier probes.
  1077. However, RMA uses the probe intensities of all arrays in the data set in
  1078. the normalization of each individual array, meaning that the normalized
  1079. expression values in each array depend on every array in the data set,
  1080. and will necessarily change each time an array is added or removed from
  1081. the data set.
  1082. If this is undesirable, frozen RMA implements a variant of RMA where the
  1083. relevant distributional parameters are learned from a large reference set
  1084. of diverse public array data sets and then
  1085. \begin_inset Quotes eld
  1086. \end_inset
  1087. frozen
  1088. \begin_inset Quotes erd
  1089. \end_inset
  1090. , so that each array is effectively normalized against this frozen reference
  1091. set rather than the other arrays in the data set under study [CITE].
  1092. Other array normalization methods considered include dChip, GRSN, and SCAN
  1093. [CITEx3].
  1094. \end_layout
  1095. \begin_layout Standard
  1096. In contrast, high-throughput sequencing data present very different normalizatio
  1097. n challenges.
  1098. The simplest case is RNA-seq in which read counts are obtained for a set
  1099. of gene annotations, yielding a matrix of counts with rows representing
  1100. genes and columns representing samples.
  1101. Because RNA-seq approximates a process of sampling from a population with
  1102. replacement, each gene's count is only interpretable as a fraction of the
  1103. total reads for that sample.
  1104. For that reason, RNA-seq abundances are often reported as counts per million
  1105. (CPM).
  1106. Furthermore, if the abundance of a single gene increases, then in order
  1107. for its fraction of the total reads to increase, all other genes' fractions
  1108. must decrease to accommodate it.
  1109. This effect is known as composition bias, and it is an artifact of the
  1110. read sampling process that has nothing to do with the biology of the samples
  1111. and must therefore be normalized out.
  1112. The most commonly used methods to normalize for composition bias in RNA-seq
  1113. data seek to equalize the average gene abundance across samples, under
  1114. the assumption that the average gene is likely not changing
  1115. \begin_inset CommandInset citation
  1116. LatexCommand cite
  1117. key "Robinson2010,Anders2010"
  1118. literal "false"
  1119. \end_inset
  1120. .
  1121. \end_layout
  1122. \begin_layout Standard
  1123. In ChIP-seq data, normalization is not as straightforward.
  1124. The
  1125. \begin_inset Flex Code
  1126. status open
  1127. \begin_layout Plain Layout
  1128. csaw
  1129. \end_layout
  1130. \end_inset
  1131. package implements several different normalization strategies
  1132. and provides guidance on when to use each one
  1133. \begin_inset CommandInset citation
  1134. LatexCommand cite
  1135. key "Lun2015a"
  1136. literal "false"
  1137. \end_inset
  1138. .
  1139. Briefly, a typical ChIP-seq sample has a bimodal distribution of read counts:
  1140. a low-abundance mode representing background regions and a high-abundance
  1141. mode representing signal regions.
  1142. This offers two potential normalization targets: equalizing background
  1143. coverage or equalizing signal coverage.
  1144. If the experiment is well controlled and ChIP efficiency is known to be
  1145. consistent across all samples, then normalizing the background coverage
  1146. to be equal across all samples is a reasonable strategy.
  1147. If this is not a safe assumption, then the preferred strategy is to normalize
  1148. the signal regions in a way similar to RNA-seq data by assuming that the
  1149. average signal region is not changing abundance between samples.
  1150. Beyond this, if a ChIP-seq experiment has a more complicated structure
  1151. that doesn't show the typical bimodal count distribution, it may be necessary
  1152. to implement a normalization as a smooth function of abundance.
  1153. However, this strategy makes a much stronger assumption about the data:
  1154. that the average log fold change is zero across all abundance levels.
  1155. Hence, the simpler scaling normalization based on background or signal
  1156. regions are generally preferred whenever possible.
  1157. \end_layout
  1158. \begin_layout Subsubsection
  1159. ComBat and SVA for correction of known and unknown batch effects
  1160. \end_layout
  1161. \begin_layout Standard
  1162. In addition to well-understood effects that can be easily normalized out,
  1163. a data set often contains confounding biological effects that must be accounted
  1164. for in the modeling step.
  1165. For instance, in an experiment with pre-treatment and post-treatment samples
  1166. of cells from several different donors, donor variability represents a
  1167. known batch effect.
  1168. The most straightforward correction for known batches is to estimate the
  1169. mean for each batch independently and subtract out the differences, so
  1170. that all batches have identical means for each feature.
  1171. However, as with variance estimation, estimating the differences in batch
  1172. means is not necessarily robust at the feature level, so the ComBat method
  1173. adds empirical Bayes squeezing of the batch mean differences toward a common
  1174. value, analogous to
  1175. \begin_inset Flex Code
  1176. status open
  1177. \begin_layout Plain Layout
  1178. limma
  1179. \end_layout
  1180. \end_inset
  1181. 's empirical Bayes squeezing of feature variance estimates
  1182. \begin_inset CommandInset citation
  1183. LatexCommand cite
  1184. key "Johnson2007"
  1185. literal "false"
  1186. \end_inset
  1187. .
  1188. Effectively, ComBat assumes that modest differences between batch means
  1189. are real batch effects, but extreme differences between batch means are
  1190. more likely to be the result of outlier observations that happen to line
  1191. up with the batches rather than a genuine batch effect.
  1192. The result is a batch correction that is more robust against outliers than
  1193. simple subtraction of mean differences subtraction.
  1194. \end_layout
  1195. \begin_layout Standard
  1196. In some data sets, unknown batch effects may be present due to inherent
  1197. variability in in the data, either caused by technical or biological effects.
  1198. Examples of unknown batch effects include variations in enrichment efficiency
  1199. between ChIP-seq samples, variations in populations of different cell types,
  1200. and the effects of uncontrolled environmental factors on gene expression
  1201. in humans or live animals.
  1202. In an ordinary linear model context, unknown batch effects cannot be inferred
  1203. and must be treated as random noise.
  1204. However, in high-throughput experiments, once again information can be
  1205. shared across features to identify patterns of un-modeled variation that
  1206. are repeated in many features.
  1207. One attractive strategy would be to perform singular value decomposition
  1208. (SVD) on the matrix of linear model residuals (which contain all the un-modeled
  1209. variation in the data) and take the first few singular vectors as batch
  1210. effects.
  1211. While this can be effective, it makes the unreasonable assumption that
  1212. all batch effects are uncorrelated with any of the effects being modeled.
  1213. Surrogate variable analysis (SVA) starts with this approach, but takes
  1214. some additional steps to identify batch effects in the full data that are
  1215. both highly correlated with the singular vectors in the residuals and least
  1216. correlated with the effects of interest
  1217. \begin_inset CommandInset citation
  1218. LatexCommand cite
  1219. key "Leek2007"
  1220. literal "false"
  1221. \end_inset
  1222. .
  1223. Since the final batch effects are estimated from the full data, moderate
  1224. correlations between the batch effects and effects of interest are allowed,
  1225. which gives SVA much more freedom to estimate the true extent of the batch
  1226. effects compared to simple residual SVD.
  1227. Once the surrogate variables are estimated, they can be included as coefficient
  1228. s in the linear model in a similar fashion to known batch effects in order
  1229. to subtract out their effects on each feature's abundance.
  1230. \end_layout
  1231. \begin_layout Subsubsection
  1232. Factor analysis: PCA, MDS, MOFA
  1233. \end_layout
  1234. \begin_layout Standard
  1235. \begin_inset Flex TODO Note (inline)
  1236. status open
  1237. \begin_layout Plain Layout
  1238. Not sure if this merits a subsection here.
  1239. \end_layout
  1240. \end_inset
  1241. \end_layout
  1242. \begin_layout Itemize
  1243. Batch-corrected PCA is informative, but careful application is required
  1244. to avoid bias
  1245. \end_layout
  1246. \begin_layout Section
  1247. Innovation
  1248. \end_layout
  1249. \begin_layout Standard
  1250. \begin_inset Flex TODO Note (inline)
  1251. status open
  1252. \begin_layout Plain Layout
  1253. Is this entire section redundant with the Approach sections of each chapter?
  1254. I'm not really sure what to write here.
  1255. \end_layout
  1256. \end_inset
  1257. \end_layout
  1258. \begin_layout Subsection
  1259. MSC infusion to improve transplant outcomes (prevent/delay rejection)
  1260. \end_layout
  1261. \begin_layout Standard
  1262. \begin_inset Flex TODO Note (inline)
  1263. status open
  1264. \begin_layout Plain Layout
  1265. Do I still talk about this? It's the motivation for chapter 4, but I don't
  1266. actually present any work related to MSCs.
  1267. \end_layout
  1268. \end_inset
  1269. \end_layout
  1270. \begin_layout Itemize
  1271. Demonstrated in mice, but not yet in primates
  1272. \end_layout
  1273. \begin_layout Itemize
  1274. Mechanism currently unknown, but MSC are known to be immune modulatory
  1275. \end_layout
  1276. \begin_layout Itemize
  1277. Characterize MSC response to interferon gamma
  1278. \end_layout
  1279. \begin_layout Itemize
  1280. IFN-g is thought to stimulate their function
  1281. \end_layout
  1282. \begin_layout Itemize
  1283. Test IFN-g treated MSC infusion as a therapy to delay graft rejection in
  1284. cynomolgus monkeys
  1285. \end_layout
  1286. \begin_layout Itemize
  1287. Monitor animals post-transplant using blood RNA-seq at serial time points
  1288. \end_layout
  1289. \begin_layout Subsection
  1290. Investigate dynamics of histone marks in CD4 T-cell activation and memory
  1291. \end_layout
  1292. \begin_layout Itemize
  1293. Previous studies have looked at single snapshots of histone marks
  1294. \end_layout
  1295. \begin_layout Itemize
  1296. Instead, look at changes in histone marks across activation and memory
  1297. \end_layout
  1298. \begin_layout Subsection
  1299. High-throughput sequencing and microarray technologies
  1300. \end_layout
  1301. \begin_layout Itemize
  1302. Powerful methods for assaying gene expression and epigenetics across entire
  1303. genomes
  1304. \end_layout
  1305. \begin_layout Itemize
  1306. Proper analysis requires finding and exploiting systematic genome-wide trends
  1307. \end_layout
  1308. \begin_layout Chapter
  1309. Reproducible genome-wide epigenetic analysis of H3K4 and H3K27 methylation
  1310. in naive and memory CD4 T-cell activation
  1311. \end_layout
  1312. \begin_layout Standard
  1313. \begin_inset Flex TODO Note (inline)
  1314. status open
  1315. \begin_layout Plain Layout
  1316. Chapter author list: Me, Sarah, Dan
  1317. \end_layout
  1318. \end_inset
  1319. \end_layout
  1320. \begin_layout Standard
  1321. \begin_inset Flex TODO Note (inline)
  1322. status open
  1323. \begin_layout Plain Layout
  1324. Need better section titles throughout the entire chapter
  1325. \end_layout
  1326. \end_inset
  1327. \end_layout
  1328. \begin_layout Section
  1329. Approach
  1330. \end_layout
  1331. \begin_layout Standard
  1332. \begin_inset Flex TODO Note (inline)
  1333. status open
  1334. \begin_layout Plain Layout
  1335. Check on the exact correct way to write
  1336. \begin_inset Quotes eld
  1337. \end_inset
  1338. CD4 T-cell
  1339. \begin_inset Quotes erd
  1340. \end_inset
  1341. .
  1342. I think there might be a plus sign somewhere in there now? Also, maybe
  1343. figure out a reasonable way to abbreviate
  1344. \begin_inset Quotes eld
  1345. \end_inset
  1346. naive CD4 T-cells
  1347. \begin_inset Quotes erd
  1348. \end_inset
  1349. and
  1350. \begin_inset Quotes eld
  1351. \end_inset
  1352. memory CD4 T-cells
  1353. \begin_inset Quotes erd
  1354. \end_inset
  1355. .
  1356. \end_layout
  1357. \end_inset
  1358. \end_layout
  1359. \begin_layout Standard
  1360. \begin_inset Flex TODO Note (inline)
  1361. status open
  1362. \begin_layout Plain Layout
  1363. Is it ok to just copy a bunch of citations from the intros to Sarah's papers?
  1364. That feels like cheating somehow.
  1365. \end_layout
  1366. \end_inset
  1367. \end_layout
  1368. \begin_layout Standard
  1369. CD4 T-cells are central to all adaptive immune responses, as well as immune
  1370. memory [CITE?].
  1371. After an infection is cleared, a subset of the naive CD4 T-cells that responded
  1372. to that infection differentiate into memory CD4 T-cells, which are responsible
  1373. for responding to the same pathogen in the future.
  1374. Memory CD4 T-cells are functionally distinct, able to respond to an infection
  1375. more quickly and without the co-stimulation required by naive CD4 T-cells.
  1376. However, the molecular mechanisms underlying this functional distinction
  1377. are not well-understood.
  1378. Epigenetic regulation via histone modification is thought to play an important
  1379. role, but while many studies have looked at static snapshots of histone
  1380. methylation in T-cells, few studies have looked at the dynamics of histone
  1381. regulation after T-cell activation, nor the differences in histone methylation
  1382. between naive and memory T-cells.
  1383. H3K4me2, H3K4me3 and H3K27me3 are three histone marks thought to be major
  1384. epigenetic regulators of gene expression.
  1385. The goal of the present study is to investigate the role of these histone
  1386. marks in CD4 T-cell activation kinetics and memory differentiation.
  1387. In static snapshots, H3K4me2 and H3K4me3 are often observed in the promoters
  1388. of highly transcribed genes, while H3K27me3 is more often observed in promoters
  1389. of inactive genes with little to no transcription occurring.
  1390. As a result, the two H3K4 marks have been characterized as
  1391. \begin_inset Quotes eld
  1392. \end_inset
  1393. activating
  1394. \begin_inset Quotes erd
  1395. \end_inset
  1396. marks, while H3K27me3 has been characterized as
  1397. \begin_inset Quotes eld
  1398. \end_inset
  1399. deactivating
  1400. \begin_inset Quotes erd
  1401. \end_inset
  1402. .
  1403. Despite these characterizations, the actual causal relationship between
  1404. these histone modifications and gene transcription is complex and likely
  1405. involves positive and negative feedback loops between the two.
  1406. \end_layout
  1407. \begin_layout Standard
  1408. In order to investigate the relationship between gene expression and these
  1409. histone modifications in the context of naive and memory CD4 T-cell activation,
  1410. a previously published data set of combined RNA-seq and ChIP-seq data was
  1411. re-analyzed using up-to-date methods designed to address the specific analysis
  1412. challenges posed by this data set.
  1413. The data set contains naive and memory CD4 T-cell samples in a time course
  1414. before and after activation.
  1415. Like the original analysis, this analysis looks at the dynamics of these
  1416. marks histone marks and compare them to gene expression dynamics at the
  1417. same time points during activation, as well as compare them between naive
  1418. and memory cells, in hope of discovering evidence of new mechanistic details
  1419. in the interplay between them.
  1420. The original analysis of this data treated each gene promoter as a monolithic
  1421. unit and mostly assumed that ChIP-seq reads or peaks occurring anywhere
  1422. within a promoter were equivalent, regardless of where they occurred relative
  1423. to the gene structure.
  1424. For an initial analysis of the data, this was a necessary simplifying assumptio
  1425. n.
  1426. The current analysis aims to relax this assumption, first by directly analyzing
  1427. ChIP-seq peaks for differential modification, and second by taking a more
  1428. granular look at the ChIP-seq read coverage within promoter regions to
  1429. ask whether the location of histone modifications relative to the gene's
  1430. TSS is an important factor, as opposed to simple proximity.
  1431. \end_layout
  1432. \begin_layout Section
  1433. Methods
  1434. \end_layout
  1435. \begin_layout Standard
  1436. \begin_inset Flex TODO Note (inline)
  1437. status open
  1438. \begin_layout Plain Layout
  1439. Look up some more details from the papers (e.g.
  1440. activation method).
  1441. \end_layout
  1442. \end_inset
  1443. \end_layout
  1444. \begin_layout Standard
  1445. A reproducible workflow was written to analyze the raw ChIP-seq and RNA-seq
  1446. data from previous studies
  1447. \begin_inset CommandInset citation
  1448. LatexCommand cite
  1449. key "gh-cd4-csaw,LaMere2016,LaMere2017"
  1450. literal "true"
  1451. \end_inset
  1452. .
  1453. Briefly, this data consists of RNA-seq and ChIP-seq from CD4 T-cells cultured
  1454. from 4 donors.
  1455. From each donor, naive and memory CD4 T-cells were isolated separately.
  1456. Then cultures of both cells were activated [how?], and samples were taken
  1457. at 4 time points: Day 0 (pre-activation), Day 1 (early activation), Day
  1458. 5 (peak activation), and Day 14 (post-activation).
  1459. For each combination of cell type and time point, RNA was isolated and
  1460. sequenced, and ChIP-seq was performed for each of 3 histone marks: H3K4me2,
  1461. H3K4me3, and H3K27me3.
  1462. The ChIP-seq input DNA was also sequenced for each sample.
  1463. The result was 32 samples for each assay.
  1464. \end_layout
  1465. \begin_layout Subsection
  1466. RNA-seq differential expression analysis
  1467. \end_layout
  1468. \begin_layout Standard
  1469. \begin_inset Note Note
  1470. status collapsed
  1471. \begin_layout Plain Layout
  1472. \begin_inset Float figure
  1473. wide false
  1474. sideways false
  1475. status open
  1476. \begin_layout Plain Layout
  1477. \align center
  1478. \begin_inset Float figure
  1479. wide false
  1480. sideways false
  1481. status collapsed
  1482. \begin_layout Plain Layout
  1483. \align center
  1484. \begin_inset Graphics
  1485. filename graphics/CD4-csaw/rnaseq-compare/ensmebl-vs-entrez-star-CROP.png
  1486. lyxscale 25
  1487. width 35col%
  1488. groupId rna-comp-subfig
  1489. \end_inset
  1490. \end_layout
  1491. \begin_layout Plain Layout
  1492. \begin_inset Caption Standard
  1493. \begin_layout Plain Layout
  1494. STAR quantification, Entrez vs Ensembl gene annotation
  1495. \end_layout
  1496. \end_inset
  1497. \end_layout
  1498. \end_inset
  1499. \begin_inset space \qquad{}
  1500. \end_inset
  1501. \begin_inset Float figure
  1502. wide false
  1503. sideways false
  1504. status collapsed
  1505. \begin_layout Plain Layout
  1506. \align center
  1507. \begin_inset Graphics
  1508. filename graphics/CD4-csaw/rnaseq-compare/ensmebl-vs-entrez-shoal-CROP.png
  1509. lyxscale 25
  1510. width 35col%
  1511. groupId rna-comp-subfig
  1512. \end_inset
  1513. \end_layout
  1514. \begin_layout Plain Layout
  1515. \begin_inset Caption Standard
  1516. \begin_layout Plain Layout
  1517. Salmon+Shoal quantification, Entrez vs Ensembl gene annotation
  1518. \end_layout
  1519. \end_inset
  1520. \end_layout
  1521. \end_inset
  1522. \end_layout
  1523. \begin_layout Plain Layout
  1524. \align center
  1525. \begin_inset Float figure
  1526. wide false
  1527. sideways false
  1528. status collapsed
  1529. \begin_layout Plain Layout
  1530. \align center
  1531. \begin_inset Graphics
  1532. filename graphics/CD4-csaw/rnaseq-compare/star-vs-hisat2-CROP.png
  1533. lyxscale 25
  1534. width 35col%
  1535. groupId rna-comp-subfig
  1536. \end_inset
  1537. \end_layout
  1538. \begin_layout Plain Layout
  1539. \begin_inset Caption Standard
  1540. \begin_layout Plain Layout
  1541. STAR vs HISAT2 quantification, Ensembl gene annotation
  1542. \end_layout
  1543. \end_inset
  1544. \end_layout
  1545. \end_inset
  1546. \begin_inset space \qquad{}
  1547. \end_inset
  1548. \begin_inset Float figure
  1549. wide false
  1550. sideways false
  1551. status collapsed
  1552. \begin_layout Plain Layout
  1553. \align center
  1554. \begin_inset Graphics
  1555. filename graphics/CD4-csaw/rnaseq-compare/star-vs-salmon-CROP.png
  1556. lyxscale 25
  1557. width 35col%
  1558. groupId rna-comp-subfig
  1559. \end_inset
  1560. \end_layout
  1561. \begin_layout Plain Layout
  1562. \begin_inset Caption Standard
  1563. \begin_layout Plain Layout
  1564. Salmon vs STAR quantification, Ensembl gene annotation
  1565. \end_layout
  1566. \end_inset
  1567. \end_layout
  1568. \end_inset
  1569. \end_layout
  1570. \begin_layout Plain Layout
  1571. \align center
  1572. \begin_inset Float figure
  1573. wide false
  1574. sideways false
  1575. status collapsed
  1576. \begin_layout Plain Layout
  1577. \align center
  1578. \begin_inset Graphics
  1579. filename graphics/CD4-csaw/rnaseq-compare/salmon-vs-kallisto-CROP.png
  1580. lyxscale 25
  1581. width 35col%
  1582. groupId rna-comp-subfig
  1583. \end_inset
  1584. \end_layout
  1585. \begin_layout Plain Layout
  1586. \begin_inset Caption Standard
  1587. \begin_layout Plain Layout
  1588. Salmon vs Kallisto quantification, Ensembl gene annotation
  1589. \end_layout
  1590. \end_inset
  1591. \end_layout
  1592. \end_inset
  1593. \begin_inset space \qquad{}
  1594. \end_inset
  1595. \begin_inset Float figure
  1596. wide false
  1597. sideways false
  1598. status collapsed
  1599. \begin_layout Plain Layout
  1600. \align center
  1601. \begin_inset Graphics
  1602. filename graphics/CD4-csaw/rnaseq-compare/salmon-vs-shoal-CROP.png
  1603. lyxscale 25
  1604. width 35col%
  1605. groupId rna-comp-subfig
  1606. \end_inset
  1607. \end_layout
  1608. \begin_layout Plain Layout
  1609. \begin_inset Caption Standard
  1610. \begin_layout Plain Layout
  1611. Salmon+Shoal vs Salmon alone, Ensembl gene annotation
  1612. \end_layout
  1613. \end_inset
  1614. \end_layout
  1615. \end_inset
  1616. \end_layout
  1617. \begin_layout Plain Layout
  1618. \begin_inset Caption Standard
  1619. \begin_layout Plain Layout
  1620. \begin_inset CommandInset label
  1621. LatexCommand label
  1622. name "fig:RNA-norm-comp"
  1623. \end_inset
  1624. RNA-seq comparisons
  1625. \end_layout
  1626. \end_inset
  1627. \end_layout
  1628. \end_inset
  1629. \end_layout
  1630. \end_inset
  1631. \end_layout
  1632. \begin_layout Standard
  1633. Sequence reads were retrieved from the Sequence Read Archive (SRA)
  1634. \begin_inset CommandInset citation
  1635. LatexCommand cite
  1636. key "Leinonen2011"
  1637. literal "false"
  1638. \end_inset
  1639. .
  1640. Five different alignment and quantification methods were tested for the
  1641. RNA-seq data
  1642. \begin_inset CommandInset citation
  1643. LatexCommand cite
  1644. key "Dobin2012,Kim2019,Liao2014,Pimentel2016,Patro2017,gh-shoal,gh-hg38-ref"
  1645. literal "false"
  1646. \end_inset
  1647. .
  1648. Each quantification was tested with both Ensembl transcripts and UCSC known
  1649. gene annotations [CITE? Also which versions of each?].
  1650. Comparisons of downstream results from each combination of quantification
  1651. method and reference revealed that all quantifications gave broadly similar
  1652. results for most genes, so shoal with the Ensembl annotation was chosen
  1653. as the method theoretically most likely to partially mitigate some of the
  1654. batch effect in the data.
  1655. \end_layout
  1656. \begin_layout Standard
  1657. \begin_inset Float figure
  1658. wide false
  1659. sideways false
  1660. status collapsed
  1661. \begin_layout Plain Layout
  1662. \align center
  1663. \begin_inset Float figure
  1664. wide false
  1665. sideways false
  1666. status open
  1667. \begin_layout Plain Layout
  1668. \align center
  1669. \begin_inset Graphics
  1670. filename graphics/CD4-csaw/RNA-seq/PCA-no-batchsub-CROP.png
  1671. lyxscale 25
  1672. width 75col%
  1673. groupId rna-pca-subfig
  1674. \end_inset
  1675. \end_layout
  1676. \begin_layout Plain Layout
  1677. \begin_inset Caption Standard
  1678. \begin_layout Plain Layout
  1679. \series bold
  1680. \begin_inset CommandInset label
  1681. LatexCommand label
  1682. name "fig:RNA-PCA-no-batchsub"
  1683. \end_inset
  1684. Before batch correction
  1685. \end_layout
  1686. \end_inset
  1687. \end_layout
  1688. \end_inset
  1689. \end_layout
  1690. \begin_layout Plain Layout
  1691. \align center
  1692. \begin_inset Float figure
  1693. wide false
  1694. sideways false
  1695. status open
  1696. \begin_layout Plain Layout
  1697. \align center
  1698. \begin_inset Graphics
  1699. filename graphics/CD4-csaw/RNA-seq/PCA-combat-batchsub-CROP.png
  1700. lyxscale 25
  1701. width 75col%
  1702. groupId rna-pca-subfig
  1703. \end_inset
  1704. \end_layout
  1705. \begin_layout Plain Layout
  1706. \begin_inset Caption Standard
  1707. \begin_layout Plain Layout
  1708. \series bold
  1709. \begin_inset CommandInset label
  1710. LatexCommand label
  1711. name "fig:RNA-PCA-ComBat-batchsub"
  1712. \end_inset
  1713. After batch correction with ComBat
  1714. \end_layout
  1715. \end_inset
  1716. \end_layout
  1717. \end_inset
  1718. \end_layout
  1719. \begin_layout Plain Layout
  1720. \begin_inset Caption Standard
  1721. \begin_layout Plain Layout
  1722. \series bold
  1723. \begin_inset CommandInset label
  1724. LatexCommand label
  1725. name "fig:RNA-PCA"
  1726. \end_inset
  1727. PCoA plots of RNA-seq data showing effect of batch correction.
  1728. \end_layout
  1729. \end_inset
  1730. \end_layout
  1731. \end_inset
  1732. \end_layout
  1733. \begin_layout Standard
  1734. Due to an error in sample preparation, the RNA from the samples for days
  1735. 0 and 5 were sequenced using a different kit than those for days 1 and
  1736. 14.
  1737. This induced a substantial batch effect in the data due to differences
  1738. in sequencing biases between the two kits, and this batch effect is unfortunate
  1739. ly confounded with the time point variable (Figure
  1740. \begin_inset CommandInset ref
  1741. LatexCommand ref
  1742. reference "fig:RNA-PCA-no-batchsub"
  1743. plural "false"
  1744. caps "false"
  1745. noprefix "false"
  1746. \end_inset
  1747. ).
  1748. To do the best possible analysis with this data, this batch effect was
  1749. subtracted out from the data using ComBat
  1750. \begin_inset CommandInset citation
  1751. LatexCommand cite
  1752. key "Johnson2007"
  1753. literal "false"
  1754. \end_inset
  1755. , ignoring the time point variable due to the confounding with the batch
  1756. variable.
  1757. The result is a marked improvement, but the unavoidable confounding with
  1758. time point means that certain real patterns of gene expression will be
  1759. indistinguishable from the batch effect and subtracted out as a result.
  1760. Specifically, any
  1761. \begin_inset Quotes eld
  1762. \end_inset
  1763. zig-zag
  1764. \begin_inset Quotes erd
  1765. \end_inset
  1766. pattern, such as a gene whose expression goes up on day 1, down on day
  1767. 5, and back up again on day 14, will be attenuated or eliminated entirely.
  1768. In the context of a T-cell activation time course, it is unlikely that
  1769. many genes of interest will follow such an expression pattern, so this
  1770. loss was deemed an acceptable cost for correcting the batch effect.
  1771. \end_layout
  1772. \begin_layout Standard
  1773. \begin_inset Float figure
  1774. wide false
  1775. sideways false
  1776. status collapsed
  1777. \begin_layout Plain Layout
  1778. \begin_inset Flex TODO Note (inline)
  1779. status open
  1780. \begin_layout Plain Layout
  1781. Just take the top row
  1782. \end_layout
  1783. \end_inset
  1784. \end_layout
  1785. \begin_layout Plain Layout
  1786. \align center
  1787. \begin_inset Graphics
  1788. filename graphics/CD4-csaw/RNA-seq/weights-vs-covars-CROP.png
  1789. lyxscale 25
  1790. width 100col%
  1791. groupId colwidth-raster
  1792. \end_inset
  1793. \end_layout
  1794. \begin_layout Plain Layout
  1795. \begin_inset Caption Standard
  1796. \begin_layout Plain Layout
  1797. \series bold
  1798. \begin_inset CommandInset label
  1799. LatexCommand label
  1800. name "fig:RNA-seq-weights-vs-covars"
  1801. \end_inset
  1802. RNA-seq sample weights, grouped by experimental and technical covariates.
  1803. \end_layout
  1804. \end_inset
  1805. \end_layout
  1806. \end_inset
  1807. \end_layout
  1808. \begin_layout Standard
  1809. However, removing the systematic component of the batch effect still leaves
  1810. the noise component.
  1811. The gene quantifications from the first batch are substantially noisier
  1812. than those in the second batch.
  1813. This analysis corrected for this by using
  1814. \begin_inset Flex Code
  1815. status open
  1816. \begin_layout Plain Layout
  1817. limma
  1818. \end_layout
  1819. \end_inset
  1820. 's sample weighting method to assign lower weights to the noisy samples
  1821. of batch 1
  1822. \begin_inset CommandInset citation
  1823. LatexCommand cite
  1824. key "Ritchie2006,Liu2015"
  1825. literal "false"
  1826. \end_inset
  1827. .
  1828. The resulting analysis gives an accurate assessment of statistical significance
  1829. for all comparisons, which unfortunately means a loss of statistical power
  1830. for comparisons involving samples in batch 1.
  1831. \end_layout
  1832. \begin_layout Standard
  1833. In any case, the RNA-seq counts were first normalized using trimmed mean
  1834. of M-values
  1835. \begin_inset CommandInset citation
  1836. LatexCommand cite
  1837. key "Robinson2010"
  1838. literal "false"
  1839. \end_inset
  1840. , converted to normalized logCPM with quality weights using
  1841. \begin_inset Flex Code
  1842. status open
  1843. \begin_layout Plain Layout
  1844. voomWithQualityWeights
  1845. \end_layout
  1846. \end_inset
  1847. \begin_inset CommandInset citation
  1848. LatexCommand cite
  1849. key "Law2013,Liu2015"
  1850. literal "false"
  1851. \end_inset
  1852. , and batch-corrected at this point using ComBat.
  1853. A linear model was fit to the batch-corrected, quality-weighted data for
  1854. each gene using
  1855. \begin_inset Flex Code
  1856. status open
  1857. \begin_layout Plain Layout
  1858. limma
  1859. \end_layout
  1860. \end_inset
  1861. , and each gene was tested for differential expression using
  1862. \begin_inset Flex Code
  1863. status open
  1864. \begin_layout Plain Layout
  1865. limma
  1866. \end_layout
  1867. \end_inset
  1868. 's empirical Bayes moderated
  1869. \begin_inset Formula $t$
  1870. \end_inset
  1871. -test
  1872. \begin_inset CommandInset citation
  1873. LatexCommand cite
  1874. key "Smyth2005,Law2013,Phipson2013"
  1875. literal "false"
  1876. \end_inset
  1877. .
  1878. \end_layout
  1879. \begin_layout Subsection
  1880. ChIP-seq differential modification analysis
  1881. \end_layout
  1882. \begin_layout Standard
  1883. \begin_inset Float figure
  1884. wide false
  1885. sideways false
  1886. status collapsed
  1887. \begin_layout Plain Layout
  1888. \align center
  1889. \begin_inset Float figure
  1890. wide false
  1891. sideways false
  1892. status open
  1893. \begin_layout Plain Layout
  1894. \align center
  1895. \begin_inset Graphics
  1896. filename graphics/CD4-csaw/csaw/CCF-plots-noBL-PAGE2-CROP.pdf
  1897. lyxscale 50
  1898. height 40theight%
  1899. groupId ccf-subfig
  1900. \end_inset
  1901. \end_layout
  1902. \begin_layout Plain Layout
  1903. \begin_inset Caption Standard
  1904. \begin_layout Plain Layout
  1905. \series bold
  1906. \begin_inset CommandInset label
  1907. LatexCommand label
  1908. name "fig:CCF-without-blacklist"
  1909. \end_inset
  1910. Cross-correlation plots without removing blacklisted reads.
  1911. \series default
  1912. Without blacklisting, many artifactual peaks are visible in the cross-correlatio
  1913. ns of the ChIP-seq samples, and the peak at the true fragment size (147
  1914. \begin_inset space ~
  1915. \end_inset
  1916. bp) is frequently overshadowed by the artifactual peak at the read length
  1917. (100
  1918. \begin_inset space ~
  1919. \end_inset
  1920. bp).
  1921. \end_layout
  1922. \end_inset
  1923. \end_layout
  1924. \end_inset
  1925. \end_layout
  1926. \begin_layout Plain Layout
  1927. \align center
  1928. \begin_inset Float figure
  1929. wide false
  1930. sideways false
  1931. status open
  1932. \begin_layout Plain Layout
  1933. \align center
  1934. \begin_inset Graphics
  1935. filename graphics/CD4-csaw/csaw/CCF-plots-PAGE2-CROP.pdf
  1936. lyxscale 50
  1937. height 40theight%
  1938. groupId ccf-subfig
  1939. \end_inset
  1940. \end_layout
  1941. \begin_layout Plain Layout
  1942. \begin_inset Caption Standard
  1943. \begin_layout Plain Layout
  1944. \series bold
  1945. \begin_inset CommandInset label
  1946. LatexCommand label
  1947. name "fig:CCF-with-blacklist"
  1948. \end_inset
  1949. Cross-correlation plots with blacklisted reads removed.
  1950. \series default
  1951. After blacklisting, most ChIP-seq samples have clean-looking periodic cross-cor
  1952. relation plots, with the largest peak around 147
  1953. \begin_inset space ~
  1954. \end_inset
  1955. bp, the expected size for a fragment of DNA from a single nucleosome, and
  1956. little to no peak at the read length, 100
  1957. \begin_inset space ~
  1958. \end_inset
  1959. bp.
  1960. \end_layout
  1961. \end_inset
  1962. \end_layout
  1963. \end_inset
  1964. \end_layout
  1965. \begin_layout Plain Layout
  1966. \begin_inset Caption Standard
  1967. \begin_layout Plain Layout
  1968. \series bold
  1969. \begin_inset CommandInset label
  1970. LatexCommand label
  1971. name "fig:CCF-master"
  1972. \end_inset
  1973. Strand cross-correlation plots for ChIP-seq data, before and after blacklisting.
  1974. \end_layout
  1975. \end_inset
  1976. \end_layout
  1977. \end_inset
  1978. \end_layout
  1979. \begin_layout Standard
  1980. \begin_inset Note Note
  1981. status open
  1982. \begin_layout Plain Layout
  1983. \begin_inset Float figure
  1984. wide false
  1985. sideways false
  1986. status collapsed
  1987. \begin_layout Plain Layout
  1988. \align center
  1989. \begin_inset Graphics
  1990. filename graphics/CD4-csaw/ChIP-seq/H3K4me2-sample-MAplot-bins-CROP.png
  1991. lyxscale 25
  1992. width 100col%
  1993. groupId colwidth-raster
  1994. \end_inset
  1995. \end_layout
  1996. \begin_layout Plain Layout
  1997. \begin_inset Caption Standard
  1998. \begin_layout Plain Layout
  1999. \series bold
  2000. \begin_inset CommandInset label
  2001. LatexCommand label
  2002. name "fig:MA-plot-bigbins"
  2003. \end_inset
  2004. MA plot of H3K4me2 read counts in 10kb bins for two arbitrary samples.
  2005. \end_layout
  2006. \end_inset
  2007. \end_layout
  2008. \end_inset
  2009. \end_layout
  2010. \end_inset
  2011. \end_layout
  2012. \begin_layout Standard
  2013. \begin_inset Flex TODO Note (inline)
  2014. status open
  2015. \begin_layout Plain Layout
  2016. Be consistent about use of
  2017. \begin_inset Quotes eld
  2018. \end_inset
  2019. differential binding
  2020. \begin_inset Quotes erd
  2021. \end_inset
  2022. vs
  2023. \begin_inset Quotes eld
  2024. \end_inset
  2025. differential modification
  2026. \begin_inset Quotes erd
  2027. \end_inset
  2028. throughout this chapter.
  2029. The latter is usually preferred.
  2030. \end_layout
  2031. \end_inset
  2032. \end_layout
  2033. \begin_layout Standard
  2034. Sequence reads were retrieved from SRA
  2035. \begin_inset CommandInset citation
  2036. LatexCommand cite
  2037. key "Leinonen2011"
  2038. literal "false"
  2039. \end_inset
  2040. .
  2041. ChIP-seq (and input) reads were aligned to GRCh38 genome assembly using
  2042. Bowtie 2
  2043. \begin_inset CommandInset citation
  2044. LatexCommand cite
  2045. key "Langmead2012,Schneider2017,gh-hg38-ref"
  2046. literal "false"
  2047. \end_inset
  2048. .
  2049. Artifact regions were annotated using a custom implementation of the GreyListCh
  2050. IP algorithm, and these
  2051. \begin_inset Quotes eld
  2052. \end_inset
  2053. greylists
  2054. \begin_inset Quotes erd
  2055. \end_inset
  2056. were merged with the published ENCODE blacklists
  2057. \begin_inset CommandInset citation
  2058. LatexCommand cite
  2059. key "greylistchip,Amemiya2019,Dunham2012,gh-cd4-csaw"
  2060. literal "false"
  2061. \end_inset
  2062. .
  2063. Any read or called peak overlapping one of these regions was regarded as
  2064. artifactual and excluded from downstream analyses.
  2065. Figure
  2066. \begin_inset CommandInset ref
  2067. LatexCommand ref
  2068. reference "fig:CCF-master"
  2069. plural "false"
  2070. caps "false"
  2071. noprefix "false"
  2072. \end_inset
  2073. shows the improvement after blacklisting in the strand cross-correlation
  2074. plots, a common quality control plot for ChIP-seq data.
  2075. Peaks were called using epic, an implementation of the SICER algorithm
  2076. \begin_inset CommandInset citation
  2077. LatexCommand cite
  2078. key "Zang2009,gh-epic"
  2079. literal "false"
  2080. \end_inset
  2081. .
  2082. Peaks were also called separately using MACS, but MACS was determined to
  2083. be a poor fit for the data, and these peak calls are not used in any further
  2084. analyses
  2085. \begin_inset CommandInset citation
  2086. LatexCommand cite
  2087. key "Zhang2008"
  2088. literal "false"
  2089. \end_inset
  2090. .
  2091. Consensus peaks were determined by applying the irreproducible discovery
  2092. rate (IDR) framework
  2093. \begin_inset CommandInset citation
  2094. LatexCommand cite
  2095. key "Li2006,gh-idr"
  2096. literal "false"
  2097. \end_inset
  2098. to find peaks consistently called in the same locations across all 4 donors.
  2099. \end_layout
  2100. \begin_layout Standard
  2101. Promoters were defined by computing the distance from each annotated TSS
  2102. to the nearest called peak and examining the distribution of distances,
  2103. observing that peaks for each histone mark were enriched within a certain
  2104. distance of the TSS.
  2105. For H3K4me2 and H3K4me3, this distance was about 1
  2106. \begin_inset space ~
  2107. \end_inset
  2108. kb, while for H3K27me3 it was 2.5
  2109. \begin_inset space ~
  2110. \end_inset
  2111. kb.
  2112. These distances were used as an
  2113. \begin_inset Quotes eld
  2114. \end_inset
  2115. effective promoter radius
  2116. \begin_inset Quotes erd
  2117. \end_inset
  2118. for each mark.
  2119. The promoter region for each gene was defined as the region of the genome
  2120. within this distance upstream or downstream of the gene's annotated TSS.
  2121. For genes with multiple annotated TSSs, a promoter region was defined for
  2122. each TSS individually, and any promoters that overlapped (due to multiple
  2123. TSSs being closer than 2 times the radius) were merged into one large promoter.
  2124. Thus, some genes had multiple promoters defined, which were each analyzed
  2125. separately for differential modification.
  2126. \end_layout
  2127. \begin_layout Standard
  2128. \begin_inset Float figure
  2129. wide false
  2130. sideways false
  2131. status collapsed
  2132. \begin_layout Plain Layout
  2133. \begin_inset Float figure
  2134. wide false
  2135. sideways false
  2136. status collapsed
  2137. \begin_layout Plain Layout
  2138. \align center
  2139. \begin_inset Graphics
  2140. filename graphics/CD4-csaw/ChIP-seq/H3K4me2-PCA-raw-CROP.png
  2141. lyxscale 25
  2142. width 45col%
  2143. groupId pcoa-subfig
  2144. \end_inset
  2145. \end_layout
  2146. \begin_layout Plain Layout
  2147. \begin_inset Caption Standard
  2148. \begin_layout Plain Layout
  2149. \series bold
  2150. \begin_inset CommandInset label
  2151. LatexCommand label
  2152. name "fig:PCoA-H3K4me2-bad"
  2153. \end_inset
  2154. H3K4me2, no correction
  2155. \end_layout
  2156. \end_inset
  2157. \end_layout
  2158. \end_inset
  2159. \begin_inset space \hfill{}
  2160. \end_inset
  2161. \begin_inset Float figure
  2162. wide false
  2163. sideways false
  2164. status collapsed
  2165. \begin_layout Plain Layout
  2166. \align center
  2167. \begin_inset Graphics
  2168. filename graphics/CD4-csaw/ChIP-seq/H3K4me2-PCA-SVsub-CROP.png
  2169. lyxscale 25
  2170. width 45col%
  2171. groupId pcoa-subfig
  2172. \end_inset
  2173. \end_layout
  2174. \begin_layout Plain Layout
  2175. \begin_inset Caption Standard
  2176. \begin_layout Plain Layout
  2177. \series bold
  2178. \begin_inset CommandInset label
  2179. LatexCommand label
  2180. name "fig:PCoA-H3K4me2-good"
  2181. \end_inset
  2182. H3K4me2, SVs subtracted
  2183. \end_layout
  2184. \end_inset
  2185. \end_layout
  2186. \end_inset
  2187. \end_layout
  2188. \begin_layout Plain Layout
  2189. \begin_inset Float figure
  2190. wide false
  2191. sideways false
  2192. status collapsed
  2193. \begin_layout Plain Layout
  2194. \align center
  2195. \begin_inset Graphics
  2196. filename graphics/CD4-csaw/ChIP-seq/H3K4me3-PCA-raw-CROP.png
  2197. lyxscale 25
  2198. width 45col%
  2199. groupId pcoa-subfig
  2200. \end_inset
  2201. \end_layout
  2202. \begin_layout Plain Layout
  2203. \begin_inset Caption Standard
  2204. \begin_layout Plain Layout
  2205. \series bold
  2206. \begin_inset CommandInset label
  2207. LatexCommand label
  2208. name "fig:PCoA-H3K4me3-bad"
  2209. \end_inset
  2210. H3K4me3, no correction
  2211. \end_layout
  2212. \end_inset
  2213. \end_layout
  2214. \end_inset
  2215. \begin_inset space \hfill{}
  2216. \end_inset
  2217. \begin_inset Float figure
  2218. wide false
  2219. sideways false
  2220. status collapsed
  2221. \begin_layout Plain Layout
  2222. \align center
  2223. \begin_inset Graphics
  2224. filename graphics/CD4-csaw/ChIP-seq/H3K4me3-PCA-SVsub-CROP.png
  2225. lyxscale 25
  2226. width 45col%
  2227. groupId pcoa-subfig
  2228. \end_inset
  2229. \end_layout
  2230. \begin_layout Plain Layout
  2231. \begin_inset Caption Standard
  2232. \begin_layout Plain Layout
  2233. \series bold
  2234. \begin_inset CommandInset label
  2235. LatexCommand label
  2236. name "fig:PCoA-H3K4me3-good"
  2237. \end_inset
  2238. H3K4me3, SVs subtracted
  2239. \end_layout
  2240. \end_inset
  2241. \end_layout
  2242. \end_inset
  2243. \end_layout
  2244. \begin_layout Plain Layout
  2245. \begin_inset Float figure
  2246. wide false
  2247. sideways false
  2248. status collapsed
  2249. \begin_layout Plain Layout
  2250. \align center
  2251. \begin_inset Graphics
  2252. filename graphics/CD4-csaw/ChIP-seq/H3K27me3-PCA-raw-CROP.png
  2253. lyxscale 25
  2254. width 45col%
  2255. groupId pcoa-subfig
  2256. \end_inset
  2257. \end_layout
  2258. \begin_layout Plain Layout
  2259. \begin_inset Caption Standard
  2260. \begin_layout Plain Layout
  2261. \series bold
  2262. \begin_inset CommandInset label
  2263. LatexCommand label
  2264. name "fig:PCoA-H3K27me3-bad"
  2265. \end_inset
  2266. H3K27me3, no correction
  2267. \end_layout
  2268. \end_inset
  2269. \end_layout
  2270. \end_inset
  2271. \begin_inset space \hfill{}
  2272. \end_inset
  2273. \begin_inset Float figure
  2274. wide false
  2275. sideways false
  2276. status collapsed
  2277. \begin_layout Plain Layout
  2278. \align center
  2279. \begin_inset Graphics
  2280. filename graphics/CD4-csaw/ChIP-seq/H3K27me3-PCA-SVsub-CROP.png
  2281. lyxscale 25
  2282. width 45col%
  2283. groupId pcoa-subfig
  2284. \end_inset
  2285. \end_layout
  2286. \begin_layout Plain Layout
  2287. \begin_inset Caption Standard
  2288. \begin_layout Plain Layout
  2289. \series bold
  2290. \begin_inset CommandInset label
  2291. LatexCommand label
  2292. name "fig:PCoA-H3K27me3-good"
  2293. \end_inset
  2294. H3K27me3, SVs subtracted
  2295. \end_layout
  2296. \end_inset
  2297. \end_layout
  2298. \end_inset
  2299. \end_layout
  2300. \begin_layout Plain Layout
  2301. \begin_inset Caption Standard
  2302. \begin_layout Plain Layout
  2303. \series bold
  2304. \begin_inset CommandInset label
  2305. LatexCommand label
  2306. name "fig:PCoA-ChIP"
  2307. \end_inset
  2308. PCoA plots of ChIP-seq sliding window data, before and after subtracting
  2309. surrogate variables (SVs).
  2310. \end_layout
  2311. \end_inset
  2312. \end_layout
  2313. \end_inset
  2314. \end_layout
  2315. \begin_layout Standard
  2316. Reads in promoters, peaks, and sliding windows across the genome were counted
  2317. and normalized using
  2318. \begin_inset Flex Code
  2319. status open
  2320. \begin_layout Plain Layout
  2321. csaw
  2322. \end_layout
  2323. \end_inset
  2324. and analyzed for differential modification using
  2325. \begin_inset Flex Code
  2326. status open
  2327. \begin_layout Plain Layout
  2328. edgeR
  2329. \end_layout
  2330. \end_inset
  2331. \begin_inset CommandInset citation
  2332. LatexCommand cite
  2333. key "Lun2014,Lun2015a,Lund2012,Phipson2016"
  2334. literal "false"
  2335. \end_inset
  2336. .
  2337. Unobserved confounding factors in the ChIP-seq data were corrected using
  2338. SVA
  2339. \begin_inset CommandInset citation
  2340. LatexCommand cite
  2341. key "Leek2007,Leek2014"
  2342. literal "false"
  2343. \end_inset
  2344. .
  2345. Principal coordinate plots of the promoter count data for each histone
  2346. mark before and after subtracting surrogate variable effects are shown
  2347. in Figure
  2348. \begin_inset CommandInset ref
  2349. LatexCommand ref
  2350. reference "fig:PCoA-ChIP"
  2351. plural "false"
  2352. caps "false"
  2353. noprefix "false"
  2354. \end_inset
  2355. .
  2356. \end_layout
  2357. \begin_layout Standard
  2358. To investigate whether the location of a peak within the promoter region
  2359. was important,
  2360. \begin_inset Quotes eld
  2361. \end_inset
  2362. relative coverage profiles
  2363. \begin_inset Quotes erd
  2364. \end_inset
  2365. were generated.
  2366. First, 500-bp sliding windows were tiled around each annotated TSS: one
  2367. window centered on the TSS itself, and 10 windows each upstream and downstream,
  2368. thus covering a 10.5-kb region centered on the TSS with 21 windows.
  2369. Reads in each window for each TSS were counted in each sample, and the
  2370. counts were normalized and converted to log CPM as in the differential
  2371. modification analysis.
  2372. Then, the logCPM values within each promoter were normalized to an average
  2373. of zero, such that each window's normalized abundance now represents the
  2374. relative read depth of that window compared to all other windows in the
  2375. same promoter.
  2376. The normalized abundance values for each window in a promoter are collectively
  2377. referred to as that promoter's
  2378. \begin_inset Quotes eld
  2379. \end_inset
  2380. relative coverage profile
  2381. \begin_inset Quotes erd
  2382. \end_inset
  2383. .
  2384. \end_layout
  2385. \begin_layout Subsection
  2386. MOFA recovers biologically relevant variation from blind analysis by correlating
  2387. across datasets
  2388. \end_layout
  2389. \begin_layout Standard
  2390. \begin_inset ERT
  2391. status open
  2392. \begin_layout Plain Layout
  2393. \backslash
  2394. afterpage{
  2395. \end_layout
  2396. \begin_layout Plain Layout
  2397. \backslash
  2398. begin{landscape}
  2399. \end_layout
  2400. \end_inset
  2401. \end_layout
  2402. \begin_layout Standard
  2403. \begin_inset Float figure
  2404. wide false
  2405. sideways false
  2406. status open
  2407. \begin_layout Plain Layout
  2408. \begin_inset Float figure
  2409. wide false
  2410. sideways false
  2411. status open
  2412. \begin_layout Plain Layout
  2413. \align center
  2414. \begin_inset Graphics
  2415. filename graphics/CD4-csaw/MOFA-varExplaiend-matrix-CROP.png
  2416. lyxscale 25
  2417. width 45col%
  2418. groupId mofa-subfig
  2419. \end_inset
  2420. \end_layout
  2421. \begin_layout Plain Layout
  2422. \begin_inset Caption Standard
  2423. \begin_layout Plain Layout
  2424. \series bold
  2425. \begin_inset CommandInset label
  2426. LatexCommand label
  2427. name "fig:mofa-varexplained"
  2428. \end_inset
  2429. Variance explained in each data set by each latent factor estimated by MOFA.
  2430. \series default
  2431. For each latent factor (LF) learned by MOFA, the variance explained by
  2432. that factor in each data set (
  2433. \begin_inset Quotes eld
  2434. \end_inset
  2435. view
  2436. \begin_inset Quotes erd
  2437. \end_inset
  2438. ) is shown by the shading of the cells in the lower section.
  2439. The upper section shows the total fraction of each data set's variance
  2440. that is explained by all LFs combined.
  2441. \end_layout
  2442. \end_inset
  2443. \end_layout
  2444. \end_inset
  2445. \begin_inset space \hfill{}
  2446. \end_inset
  2447. \begin_inset Float figure
  2448. wide false
  2449. sideways false
  2450. status open
  2451. \begin_layout Plain Layout
  2452. \align center
  2453. \begin_inset Graphics
  2454. filename graphics/CD4-csaw/MOFA-LF-scatter-CROP.png
  2455. lyxscale 25
  2456. width 45col%
  2457. groupId mofa-subfig
  2458. \end_inset
  2459. \end_layout
  2460. \begin_layout Plain Layout
  2461. \begin_inset Caption Standard
  2462. \begin_layout Plain Layout
  2463. \series bold
  2464. \begin_inset CommandInset label
  2465. LatexCommand label
  2466. name "fig:mofa-lf-scatter"
  2467. \end_inset
  2468. Scatter plots of specific pairs of MOFA latent factors.
  2469. \series default
  2470. LFs 1, 4, and 5 explain substantial variation in all data sets, so they
  2471. are plotted against each other in order to reveal patterns of variation
  2472. that are shared across all data sets.
  2473. \end_layout
  2474. \end_inset
  2475. \end_layout
  2476. \end_inset
  2477. \end_layout
  2478. \begin_layout Plain Layout
  2479. \begin_inset Caption Standard
  2480. \begin_layout Plain Layout
  2481. \series bold
  2482. \begin_inset CommandInset label
  2483. LatexCommand label
  2484. name "fig:MOFA-master"
  2485. \end_inset
  2486. MOFA latent factors separate technical confounders from
  2487. \end_layout
  2488. \end_inset
  2489. \end_layout
  2490. \end_inset
  2491. \end_layout
  2492. \begin_layout Standard
  2493. \begin_inset ERT
  2494. status open
  2495. \begin_layout Plain Layout
  2496. \backslash
  2497. end{landscape}
  2498. \end_layout
  2499. \begin_layout Plain Layout
  2500. }
  2501. \end_layout
  2502. \end_inset
  2503. \end_layout
  2504. \begin_layout Standard
  2505. MOFA was run on all the ChIP-seq windows overlapping consensus peaks for
  2506. each histone mark, as well as the RNA-seq data, in order to identify patterns
  2507. of coordinated variation across all data sets
  2508. \begin_inset CommandInset citation
  2509. LatexCommand cite
  2510. key "Argelaguet2018"
  2511. literal "false"
  2512. \end_inset
  2513. .
  2514. The results are summarized in Figure
  2515. \begin_inset CommandInset ref
  2516. LatexCommand ref
  2517. reference "fig:MOFA-master"
  2518. plural "false"
  2519. caps "false"
  2520. noprefix "false"
  2521. \end_inset
  2522. .
  2523. Latent factors 1, 4, and 5 were determined to explain the most variation
  2524. consistently across all data sets (Figure
  2525. \begin_inset CommandInset ref
  2526. LatexCommand ref
  2527. reference "fig:mofa-varexplained"
  2528. plural "false"
  2529. caps "false"
  2530. noprefix "false"
  2531. \end_inset
  2532. ), and scatter plots of these factors show that they also correlate best
  2533. with the experimental factors (Figure
  2534. \begin_inset CommandInset ref
  2535. LatexCommand ref
  2536. reference "fig:mofa-lf-scatter"
  2537. plural "false"
  2538. caps "false"
  2539. noprefix "false"
  2540. \end_inset
  2541. ).
  2542. Latent factor 2 captures the batch effect in the RNA-seq data.
  2543. Removing the effect of LF2 using MOFA theoretically yields a batch correction
  2544. that does not depend on knowing the experimental factors.
  2545. When this was attempted, the resulting batch correction was comparable
  2546. to ComBat (see Figure
  2547. \begin_inset CommandInset ref
  2548. LatexCommand ref
  2549. reference "fig:RNA-PCA-ComBat-batchsub"
  2550. plural "false"
  2551. caps "false"
  2552. noprefix "false"
  2553. \end_inset
  2554. ), indicating that the ComBat-based batch correction has little room for
  2555. improvement given the problems with the data set.
  2556. \end_layout
  2557. \begin_layout Standard
  2558. \begin_inset Note Note
  2559. status collapsed
  2560. \begin_layout Plain Layout
  2561. \begin_inset Float figure
  2562. wide false
  2563. sideways false
  2564. status open
  2565. \begin_layout Plain Layout
  2566. \align center
  2567. \begin_inset Graphics
  2568. filename graphics/CD4-csaw/MOFA-batch-correct-CROP.png
  2569. lyxscale 25
  2570. width 100col%
  2571. groupId colwidth-raster
  2572. \end_inset
  2573. \end_layout
  2574. \begin_layout Plain Layout
  2575. \begin_inset Caption Standard
  2576. \begin_layout Plain Layout
  2577. \series bold
  2578. \begin_inset CommandInset label
  2579. LatexCommand label
  2580. name "fig:mofa-batchsub"
  2581. \end_inset
  2582. Result of RNA-seq batch-correction using MOFA latent factors
  2583. \end_layout
  2584. \end_inset
  2585. \end_layout
  2586. \end_inset
  2587. \end_layout
  2588. \end_inset
  2589. \end_layout
  2590. \begin_layout Section
  2591. Results
  2592. \end_layout
  2593. \begin_layout Standard
  2594. \begin_inset Flex TODO Note (inline)
  2595. status open
  2596. \begin_layout Plain Layout
  2597. Focus on what hypotheses were tested, then select figures that show how
  2598. those hypotheses were tested, even if the result is a negative.
  2599. Not every interesting result needs to be in here.
  2600. Chapter should tell a story.
  2601. \end_layout
  2602. \end_inset
  2603. \end_layout
  2604. \begin_layout Standard
  2605. \begin_inset Flex TODO Note (inline)
  2606. status open
  2607. \begin_layout Plain Layout
  2608. Maybe reorder these sections to do RNA-seq, then ChIP-seq, then combined
  2609. analyses?
  2610. \end_layout
  2611. \end_inset
  2612. \end_layout
  2613. \begin_layout Subsection
  2614. Interpretation of RNA-seq analysis is limited by a major confounding factor
  2615. \end_layout
  2616. \begin_layout Standard
  2617. \begin_inset Float table
  2618. wide false
  2619. sideways false
  2620. status collapsed
  2621. \begin_layout Plain Layout
  2622. \align center
  2623. \begin_inset Tabular
  2624. <lyxtabular version="3" rows="11" columns="3">
  2625. <features tabularvalignment="middle">
  2626. <column alignment="center" valignment="top">
  2627. <column alignment="center" valignment="top">
  2628. <column alignment="center" valignment="top">
  2629. <row>
  2630. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  2631. \begin_inset Text
  2632. \begin_layout Plain Layout
  2633. Test
  2634. \end_layout
  2635. \end_inset
  2636. </cell>
  2637. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  2638. \begin_inset Text
  2639. \begin_layout Plain Layout
  2640. Est.
  2641. non-null
  2642. \end_layout
  2643. \end_inset
  2644. </cell>
  2645. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  2646. \begin_inset Text
  2647. \begin_layout Plain Layout
  2648. \begin_inset Formula $\mathrm{FDR}\le10\%$
  2649. \end_inset
  2650. \end_layout
  2651. \end_inset
  2652. </cell>
  2653. </row>
  2654. <row>
  2655. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  2656. \begin_inset Text
  2657. \begin_layout Plain Layout
  2658. Naive Day 0 vs Day 1
  2659. \end_layout
  2660. \end_inset
  2661. </cell>
  2662. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  2663. \begin_inset Text
  2664. \begin_layout Plain Layout
  2665. 5992
  2666. \end_layout
  2667. \end_inset
  2668. </cell>
  2669. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  2670. \begin_inset Text
  2671. \begin_layout Plain Layout
  2672. 1613
  2673. \end_layout
  2674. \end_inset
  2675. </cell>
  2676. </row>
  2677. <row>
  2678. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  2679. \begin_inset Text
  2680. \begin_layout Plain Layout
  2681. Naive Day 0 vs Day 5
  2682. \end_layout
  2683. \end_inset
  2684. </cell>
  2685. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  2686. \begin_inset Text
  2687. \begin_layout Plain Layout
  2688. 3038
  2689. \end_layout
  2690. \end_inset
  2691. </cell>
  2692. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  2693. \begin_inset Text
  2694. \begin_layout Plain Layout
  2695. 32
  2696. \end_layout
  2697. \end_inset
  2698. </cell>
  2699. </row>
  2700. <row>
  2701. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  2702. \begin_inset Text
  2703. \begin_layout Plain Layout
  2704. Naive Day 0 vs Day 14
  2705. \end_layout
  2706. \end_inset
  2707. </cell>
  2708. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  2709. \begin_inset Text
  2710. \begin_layout Plain Layout
  2711. 1870
  2712. \end_layout
  2713. \end_inset
  2714. </cell>
  2715. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  2716. \begin_inset Text
  2717. \begin_layout Plain Layout
  2718. 190
  2719. \end_layout
  2720. \end_inset
  2721. </cell>
  2722. </row>
  2723. <row>
  2724. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  2725. \begin_inset Text
  2726. \begin_layout Plain Layout
  2727. Memory Day 0 vs Day 1
  2728. \end_layout
  2729. \end_inset
  2730. </cell>
  2731. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  2732. \begin_inset Text
  2733. \begin_layout Plain Layout
  2734. 3195
  2735. \end_layout
  2736. \end_inset
  2737. </cell>
  2738. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  2739. \begin_inset Text
  2740. \begin_layout Plain Layout
  2741. 411
  2742. \end_layout
  2743. \end_inset
  2744. </cell>
  2745. </row>
  2746. <row>
  2747. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  2748. \begin_inset Text
  2749. \begin_layout Plain Layout
  2750. Memory Day 0 vs Day 5
  2751. \end_layout
  2752. \end_inset
  2753. </cell>
  2754. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  2755. \begin_inset Text
  2756. \begin_layout Plain Layout
  2757. 2688
  2758. \end_layout
  2759. \end_inset
  2760. </cell>
  2761. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  2762. \begin_inset Text
  2763. \begin_layout Plain Layout
  2764. 18
  2765. \end_layout
  2766. \end_inset
  2767. </cell>
  2768. </row>
  2769. <row>
  2770. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  2771. \begin_inset Text
  2772. \begin_layout Plain Layout
  2773. Memory Day 0 vs Day 14
  2774. \end_layout
  2775. \end_inset
  2776. </cell>
  2777. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  2778. \begin_inset Text
  2779. \begin_layout Plain Layout
  2780. 1911
  2781. \end_layout
  2782. \end_inset
  2783. </cell>
  2784. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  2785. \begin_inset Text
  2786. \begin_layout Plain Layout
  2787. 227
  2788. \end_layout
  2789. \end_inset
  2790. </cell>
  2791. </row>
  2792. <row>
  2793. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  2794. \begin_inset Text
  2795. \begin_layout Plain Layout
  2796. Day 0 Naive vs Memory
  2797. \end_layout
  2798. \end_inset
  2799. </cell>
  2800. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  2801. \begin_inset Text
  2802. \begin_layout Plain Layout
  2803. 0
  2804. \end_layout
  2805. \end_inset
  2806. </cell>
  2807. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  2808. \begin_inset Text
  2809. \begin_layout Plain Layout
  2810. 2
  2811. \end_layout
  2812. \end_inset
  2813. </cell>
  2814. </row>
  2815. <row>
  2816. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  2817. \begin_inset Text
  2818. \begin_layout Plain Layout
  2819. Day 1 Naive vs Memory
  2820. \end_layout
  2821. \end_inset
  2822. </cell>
  2823. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  2824. \begin_inset Text
  2825. \begin_layout Plain Layout
  2826. 9167
  2827. \end_layout
  2828. \end_inset
  2829. </cell>
  2830. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  2831. \begin_inset Text
  2832. \begin_layout Plain Layout
  2833. 5532
  2834. \end_layout
  2835. \end_inset
  2836. </cell>
  2837. </row>
  2838. <row>
  2839. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  2840. \begin_inset Text
  2841. \begin_layout Plain Layout
  2842. Day 5 Naive vs Memory
  2843. \end_layout
  2844. \end_inset
  2845. </cell>
  2846. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  2847. \begin_inset Text
  2848. \begin_layout Plain Layout
  2849. 0
  2850. \end_layout
  2851. \end_inset
  2852. </cell>
  2853. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  2854. \begin_inset Text
  2855. \begin_layout Plain Layout
  2856. 0
  2857. \end_layout
  2858. \end_inset
  2859. </cell>
  2860. </row>
  2861. <row>
  2862. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  2863. \begin_inset Text
  2864. \begin_layout Plain Layout
  2865. Day 14 Naive vs Memory
  2866. \end_layout
  2867. \end_inset
  2868. </cell>
  2869. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  2870. \begin_inset Text
  2871. \begin_layout Plain Layout
  2872. 6446
  2873. \end_layout
  2874. \end_inset
  2875. </cell>
  2876. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  2877. \begin_inset Text
  2878. \begin_layout Plain Layout
  2879. 2319
  2880. \end_layout
  2881. \end_inset
  2882. </cell>
  2883. </row>
  2884. </lyxtabular>
  2885. \end_inset
  2886. \end_layout
  2887. \begin_layout Plain Layout
  2888. \begin_inset Caption Standard
  2889. \begin_layout Plain Layout
  2890. \series bold
  2891. \begin_inset CommandInset label
  2892. LatexCommand label
  2893. name "tab:Estimated-and-detected-rnaseq"
  2894. \end_inset
  2895. Estimated and detected differentially expressed genes.
  2896. \series default
  2897. \begin_inset Quotes eld
  2898. \end_inset
  2899. Test
  2900. \begin_inset Quotes erd
  2901. \end_inset
  2902. : Which sample groups were compared;
  2903. \begin_inset Quotes eld
  2904. \end_inset
  2905. Est non-null
  2906. \begin_inset Quotes erd
  2907. \end_inset
  2908. : Estimated number of differentially expressed genes, using the method of
  2909. averaging local FDR values
  2910. \begin_inset CommandInset citation
  2911. LatexCommand cite
  2912. key "Phipson2013Thesis"
  2913. literal "false"
  2914. \end_inset
  2915. ;
  2916. \begin_inset Quotes eld
  2917. \end_inset
  2918. \begin_inset Formula $\mathrm{FDR}\le10\%$
  2919. \end_inset
  2920. \begin_inset Quotes erd
  2921. \end_inset
  2922. : Number of significantly differentially expressed genes at an FDR threshold
  2923. of 10%.
  2924. The total number of genes tested was 16707.
  2925. \end_layout
  2926. \end_inset
  2927. \end_layout
  2928. \end_inset
  2929. \end_layout
  2930. \begin_layout Standard
  2931. \begin_inset Float figure
  2932. wide false
  2933. sideways false
  2934. status collapsed
  2935. \begin_layout Plain Layout
  2936. \align center
  2937. \begin_inset Graphics
  2938. filename graphics/CD4-csaw/RNA-seq/PCA-final-12-CROP.png
  2939. lyxscale 25
  2940. width 100col%
  2941. groupId colwidth-raster
  2942. \end_inset
  2943. \end_layout
  2944. \begin_layout Plain Layout
  2945. \begin_inset Caption Standard
  2946. \begin_layout Plain Layout
  2947. \series bold
  2948. \begin_inset CommandInset label
  2949. LatexCommand label
  2950. name "fig:rna-pca-final"
  2951. \end_inset
  2952. PCoA plot of RNA-seq samples after ComBat batch correction.
  2953. \series default
  2954. Each point represents an individual sample.
  2955. Samples with the same combination of cell type and time point are encircled
  2956. with a shaded region to aid in visual identification of the sample groups.
  2957. Samples with of same cell type from the same donor are connected by lines
  2958. to indicate the
  2959. \begin_inset Quotes eld
  2960. \end_inset
  2961. trajectory
  2962. \begin_inset Quotes erd
  2963. \end_inset
  2964. of each donor's cells over time in PCoA space.
  2965. \end_layout
  2966. \end_inset
  2967. \end_layout
  2968. \begin_layout Plain Layout
  2969. \end_layout
  2970. \end_inset
  2971. \end_layout
  2972. \begin_layout Standard
  2973. Genes called present in the RNA-seq data were tested for differential expression
  2974. between all time points and cell types.
  2975. The counts of differentially expressed genes are shown in Table
  2976. \begin_inset CommandInset ref
  2977. LatexCommand ref
  2978. reference "tab:Estimated-and-detected-rnaseq"
  2979. plural "false"
  2980. caps "false"
  2981. noprefix "false"
  2982. \end_inset
  2983. .
  2984. Notably, all the results for Day 0 and Day 5 have substantially fewer genes
  2985. called differentially expressed than any of the results for other time
  2986. points.
  2987. This is an unfortunate result of the difference in sample quality between
  2988. the two batches of RNA-seq data.
  2989. All the samples in Batch 1, which includes all the samples from Days 0
  2990. and 5, have substantially more variability than the samples in Batch 2,
  2991. which includes the other time points.
  2992. This is reflected in the substantially higher weights assigned to Batch
  2993. 2 (Figure
  2994. \begin_inset CommandInset ref
  2995. LatexCommand ref
  2996. reference "fig:RNA-seq-weights-vs-covars"
  2997. plural "false"
  2998. caps "false"
  2999. noprefix "false"
  3000. \end_inset
  3001. ).
  3002. The batch effect has both a systematic component and a random noise component.
  3003. While the systematic component was subtracted out using ComBat (Figure
  3004. \begin_inset CommandInset ref
  3005. LatexCommand ref
  3006. reference "fig:RNA-PCA"
  3007. plural "false"
  3008. caps "false"
  3009. noprefix "false"
  3010. \end_inset
  3011. ), no such correction is possible for the noise component: Batch 1 simply
  3012. has substantially more random noise in it, which reduces the statistical
  3013. power for any differential expression tests involving samples in that batch.
  3014. \end_layout
  3015. \begin_layout Standard
  3016. Despite the difficulty in detecting specific differentially expressed genes,
  3017. there is still evidence that differential expression is present for these
  3018. time points.
  3019. In Figure
  3020. \begin_inset CommandInset ref
  3021. LatexCommand ref
  3022. reference "fig:rna-pca-final"
  3023. plural "false"
  3024. caps "false"
  3025. noprefix "false"
  3026. \end_inset
  3027. , there is a clear separation between naive and memory samples at Day 0,
  3028. despite the fact that only 2 genes were significantly differentially expressed
  3029. for this comparison.
  3030. Similarly, the small numbers of genes detected for the Day 0 vs Day 5 compariso
  3031. ns do not reflect the large separation between these time points in Figure
  3032. \begin_inset CommandInset ref
  3033. LatexCommand ref
  3034. reference "fig:rna-pca-final"
  3035. plural "false"
  3036. caps "false"
  3037. noprefix "false"
  3038. \end_inset
  3039. .
  3040. In addition, the MOFA latent factor plots in Figure
  3041. \begin_inset CommandInset ref
  3042. LatexCommand ref
  3043. reference "fig:mofa-lf-scatter"
  3044. plural "false"
  3045. caps "false"
  3046. noprefix "false"
  3047. \end_inset
  3048. .
  3049. This suggests that there is indeed a differential expression signal present
  3050. in the data for these comparisons, but the large variability in the Batch
  3051. 1 samples obfuscates this signal at the individual gene level.
  3052. As a result, it is impossible to make any meaningful statements about the
  3053. \begin_inset Quotes eld
  3054. \end_inset
  3055. size
  3056. \begin_inset Quotes erd
  3057. \end_inset
  3058. of the gene signature for any time point, since the number of significant
  3059. genes as well as the estimated number of differentially expressed genes
  3060. depends so strongly on the variations in sample quality in addition to
  3061. the size of the differential expression signal in the data.
  3062. Gene-set enrichment analyses are similarly impractical.
  3063. However, analyses looking at genome-wide patterns of expression are still
  3064. practical.
  3065. \end_layout
  3066. \begin_layout Subsection
  3067. H3K4 and H3K27 methylation occur in broad regions and are enriched near
  3068. promoters
  3069. \end_layout
  3070. \begin_layout Standard
  3071. \begin_inset Float table
  3072. wide false
  3073. sideways false
  3074. status collapsed
  3075. \begin_layout Plain Layout
  3076. \align center
  3077. \begin_inset Flex TODO Note (inline)
  3078. status open
  3079. \begin_layout Plain Layout
  3080. Also get
  3081. \emph on
  3082. median
  3083. \emph default
  3084. peak width and maybe other quantiles (25%, 75%)
  3085. \end_layout
  3086. \end_inset
  3087. \end_layout
  3088. \begin_layout Plain Layout
  3089. \align center
  3090. \begin_inset Tabular
  3091. <lyxtabular version="3" rows="4" columns="5">
  3092. <features tabularvalignment="middle">
  3093. <column alignment="center" valignment="top">
  3094. <column alignment="center" valignment="top">
  3095. <column alignment="center" valignment="top">
  3096. <column alignment="center" valignment="top">
  3097. <column alignment="center" valignment="top">
  3098. <row>
  3099. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  3100. \begin_inset Text
  3101. \begin_layout Plain Layout
  3102. Histone Mark
  3103. \end_layout
  3104. \end_inset
  3105. </cell>
  3106. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  3107. \begin_inset Text
  3108. \begin_layout Plain Layout
  3109. # Peaks
  3110. \end_layout
  3111. \end_inset
  3112. </cell>
  3113. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  3114. \begin_inset Text
  3115. \begin_layout Plain Layout
  3116. Mean peak width
  3117. \end_layout
  3118. \end_inset
  3119. </cell>
  3120. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  3121. \begin_inset Text
  3122. \begin_layout Plain Layout
  3123. genome coverage
  3124. \end_layout
  3125. \end_inset
  3126. </cell>
  3127. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  3128. \begin_inset Text
  3129. \begin_layout Plain Layout
  3130. FRiP
  3131. \end_layout
  3132. \end_inset
  3133. </cell>
  3134. </row>
  3135. <row>
  3136. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  3137. \begin_inset Text
  3138. \begin_layout Plain Layout
  3139. H3K4me2
  3140. \end_layout
  3141. \end_inset
  3142. </cell>
  3143. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  3144. \begin_inset Text
  3145. \begin_layout Plain Layout
  3146. 14965
  3147. \end_layout
  3148. \end_inset
  3149. </cell>
  3150. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  3151. \begin_inset Text
  3152. \begin_layout Plain Layout
  3153. 3970
  3154. \end_layout
  3155. \end_inset
  3156. </cell>
  3157. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  3158. \begin_inset Text
  3159. \begin_layout Plain Layout
  3160. 1.92%
  3161. \end_layout
  3162. \end_inset
  3163. </cell>
  3164. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  3165. \begin_inset Text
  3166. \begin_layout Plain Layout
  3167. 14.2%
  3168. \end_layout
  3169. \end_inset
  3170. </cell>
  3171. </row>
  3172. <row>
  3173. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  3174. \begin_inset Text
  3175. \begin_layout Plain Layout
  3176. H3K4me3
  3177. \end_layout
  3178. \end_inset
  3179. </cell>
  3180. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  3181. \begin_inset Text
  3182. \begin_layout Plain Layout
  3183. 6163
  3184. \end_layout
  3185. \end_inset
  3186. </cell>
  3187. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  3188. \begin_inset Text
  3189. \begin_layout Plain Layout
  3190. 2946
  3191. \end_layout
  3192. \end_inset
  3193. </cell>
  3194. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  3195. \begin_inset Text
  3196. \begin_layout Plain Layout
  3197. 0.588%
  3198. \end_layout
  3199. \end_inset
  3200. </cell>
  3201. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  3202. \begin_inset Text
  3203. \begin_layout Plain Layout
  3204. 6.57%
  3205. \end_layout
  3206. \end_inset
  3207. </cell>
  3208. </row>
  3209. <row>
  3210. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  3211. \begin_inset Text
  3212. \begin_layout Plain Layout
  3213. H3K27me3
  3214. \end_layout
  3215. \end_inset
  3216. </cell>
  3217. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  3218. \begin_inset Text
  3219. \begin_layout Plain Layout
  3220. 18139
  3221. \end_layout
  3222. \end_inset
  3223. </cell>
  3224. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  3225. \begin_inset Text
  3226. \begin_layout Plain Layout
  3227. 18967
  3228. \end_layout
  3229. \end_inset
  3230. </cell>
  3231. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  3232. \begin_inset Text
  3233. \begin_layout Plain Layout
  3234. 11.1%
  3235. \end_layout
  3236. \end_inset
  3237. </cell>
  3238. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  3239. \begin_inset Text
  3240. \begin_layout Plain Layout
  3241. 22.5%
  3242. \end_layout
  3243. \end_inset
  3244. </cell>
  3245. </row>
  3246. </lyxtabular>
  3247. \end_inset
  3248. \end_layout
  3249. \begin_layout Plain Layout
  3250. \begin_inset Caption Standard
  3251. \begin_layout Plain Layout
  3252. \series bold
  3253. \begin_inset CommandInset label
  3254. LatexCommand label
  3255. name "tab:peak-calling-summary"
  3256. \end_inset
  3257. Peak-calling summary.
  3258. \series default
  3259. For each histone mark, the number of peaks called using SICER at an IDR
  3260. threshold of ???, the mean width of those peaks, the fraction of the genome
  3261. covered by peaks, and the fraction of reads in peaks (FRiP).
  3262. \end_layout
  3263. \end_inset
  3264. \end_layout
  3265. \end_inset
  3266. \end_layout
  3267. \begin_layout Standard
  3268. Table
  3269. \begin_inset CommandInset ref
  3270. LatexCommand ref
  3271. reference "tab:peak-calling-summary"
  3272. plural "false"
  3273. caps "false"
  3274. noprefix "false"
  3275. \end_inset
  3276. gives a summary of the peak calling statistics for each histone mark.
  3277. Consistent with previous observations [CITATION NEEDED], all 3 histone
  3278. marks occur in broad regions spanning many consecutive nucleosomes, rather
  3279. than in sharp peaks as would be expected for a transcription factor or
  3280. other molecule that binds to specific sites.
  3281. This conclusion is further supported by Figure
  3282. \begin_inset CommandInset ref
  3283. LatexCommand ref
  3284. reference "fig:CCF-with-blacklist"
  3285. plural "false"
  3286. caps "false"
  3287. noprefix "false"
  3288. \end_inset
  3289. , in which a clear nucleosome-sized periodicity is visible in the cross-correlat
  3290. ion value for each sample, indicating that each time a given mark is present
  3291. on one histone, it is also likely to be found on adjacent histones as well.
  3292. H3K27me3 enrichment in particular is substantially more broad than either
  3293. H3K4 mark, with a mean peak width of almost 19,000 bp.
  3294. This is also reflected in the periodicity observed in Figure
  3295. \begin_inset CommandInset ref
  3296. LatexCommand ref
  3297. reference "fig:CCF-with-blacklist"
  3298. plural "false"
  3299. caps "false"
  3300. noprefix "false"
  3301. \end_inset
  3302. , which remains strong much farther out for H3K27me3 than the other marks,
  3303. showing H3K27me3 especially tends to be found on long runs of consecutive
  3304. histones.
  3305. \end_layout
  3306. \begin_layout Standard
  3307. \begin_inset Float figure
  3308. wide false
  3309. sideways false
  3310. status open
  3311. \begin_layout Plain Layout
  3312. \begin_inset Flex TODO Note (inline)
  3313. status open
  3314. \begin_layout Plain Layout
  3315. Ensure this figure uses the peak calls from the new analysis.
  3316. \end_layout
  3317. \end_inset
  3318. \end_layout
  3319. \begin_layout Plain Layout
  3320. \begin_inset Flex TODO Note (inline)
  3321. status open
  3322. \begin_layout Plain Layout
  3323. Need a control: shuffle all peaks and repeat, N times.
  3324. Do real vs shuffled control both in a top/bottom arrangement.
  3325. \end_layout
  3326. \end_inset
  3327. \end_layout
  3328. \begin_layout Plain Layout
  3329. \begin_inset Flex TODO Note (inline)
  3330. status open
  3331. \begin_layout Plain Layout
  3332. Consider counting TSS inside peaks as negative number indicating how far
  3333. \emph on
  3334. inside
  3335. \emph default
  3336. the peak the TSS is (i.e.
  3337. distance to nearest non-peak area).
  3338. \end_layout
  3339. \end_inset
  3340. \end_layout
  3341. \begin_layout Plain Layout
  3342. \begin_inset Flex TODO Note (inline)
  3343. status open
  3344. \begin_layout Plain Layout
  3345. The H3K4 part of this figure is included in
  3346. \begin_inset CommandInset citation
  3347. LatexCommand cite
  3348. key "LaMere2016"
  3349. literal "false"
  3350. \end_inset
  3351. as Fig.
  3352. S2.
  3353. Do I need to do anything about that?
  3354. \end_layout
  3355. \end_inset
  3356. \end_layout
  3357. \begin_layout Plain Layout
  3358. \align center
  3359. \begin_inset Graphics
  3360. filename graphics/CD4-csaw/Promoter Peak Distance Profile-PAGE1-CROP.pdf
  3361. lyxscale 50
  3362. width 80col%
  3363. \end_inset
  3364. \end_layout
  3365. \begin_layout Plain Layout
  3366. \begin_inset Caption Standard
  3367. \begin_layout Plain Layout
  3368. \series bold
  3369. \begin_inset CommandInset label
  3370. LatexCommand label
  3371. name "fig:near-promoter-peak-enrich"
  3372. \end_inset
  3373. Enrichment of peaks in promoter neighborhoods.
  3374. \series default
  3375. This plot shows the distribution of distances from each annotated transcription
  3376. start site in the genome to the nearest called peak.
  3377. Each line represents one combination of histone mark, cell type, and time
  3378. point.
  3379. Distributions are smoothed using kernel density estimation [CITE? see ggplot2
  3380. stat_density()].
  3381. Transcription start sites that occur
  3382. \emph on
  3383. within
  3384. \emph default
  3385. peaks were excluded from this plot to avoid a large spike at zero that
  3386. would overshadow the rest of the distribution.
  3387. \end_layout
  3388. \end_inset
  3389. \end_layout
  3390. \end_inset
  3391. \end_layout
  3392. \begin_layout Standard
  3393. \begin_inset Float table
  3394. wide false
  3395. sideways false
  3396. status collapsed
  3397. \begin_layout Plain Layout
  3398. \align center
  3399. \begin_inset Tabular
  3400. <lyxtabular version="3" rows="4" columns="2">
  3401. <features tabularvalignment="middle">
  3402. <column alignment="center" valignment="top">
  3403. <column alignment="center" valignment="top">
  3404. <row>
  3405. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  3406. \begin_inset Text
  3407. \begin_layout Plain Layout
  3408. Histone mark
  3409. \end_layout
  3410. \end_inset
  3411. </cell>
  3412. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  3413. \begin_inset Text
  3414. \begin_layout Plain Layout
  3415. Effective promoter radius
  3416. \end_layout
  3417. \end_inset
  3418. </cell>
  3419. </row>
  3420. <row>
  3421. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  3422. \begin_inset Text
  3423. \begin_layout Plain Layout
  3424. H3K4me2
  3425. \end_layout
  3426. \end_inset
  3427. </cell>
  3428. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  3429. \begin_inset Text
  3430. \begin_layout Plain Layout
  3431. 1 kb
  3432. \end_layout
  3433. \end_inset
  3434. </cell>
  3435. </row>
  3436. <row>
  3437. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  3438. \begin_inset Text
  3439. \begin_layout Plain Layout
  3440. H3K4me3
  3441. \end_layout
  3442. \end_inset
  3443. </cell>
  3444. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  3445. \begin_inset Text
  3446. \begin_layout Plain Layout
  3447. 1 kb
  3448. \end_layout
  3449. \end_inset
  3450. </cell>
  3451. </row>
  3452. <row>
  3453. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  3454. \begin_inset Text
  3455. \begin_layout Plain Layout
  3456. H3K27me3
  3457. \end_layout
  3458. \end_inset
  3459. </cell>
  3460. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  3461. \begin_inset Text
  3462. \begin_layout Plain Layout
  3463. 2.5 kb
  3464. \end_layout
  3465. \end_inset
  3466. </cell>
  3467. </row>
  3468. </lyxtabular>
  3469. \end_inset
  3470. \end_layout
  3471. \begin_layout Plain Layout
  3472. \begin_inset Caption Standard
  3473. \begin_layout Plain Layout
  3474. \series bold
  3475. \begin_inset CommandInset label
  3476. LatexCommand label
  3477. name "tab:effective-promoter-radius"
  3478. \end_inset
  3479. Effective promoter radius for each histone mark.
  3480. \series default
  3481. These values represent the approximate distance from transcription start
  3482. site positions within which an excess of peaks are found, as shown in Figure
  3483. \begin_inset CommandInset ref
  3484. LatexCommand ref
  3485. reference "fig:near-promoter-peak-enrich"
  3486. plural "false"
  3487. caps "false"
  3488. noprefix "false"
  3489. \end_inset
  3490. .
  3491. \end_layout
  3492. \end_inset
  3493. \end_layout
  3494. \begin_layout Plain Layout
  3495. \end_layout
  3496. \end_inset
  3497. \end_layout
  3498. \begin_layout Standard
  3499. All 3 histone marks tend to occur more often near promoter regions, as shown
  3500. in Figure
  3501. \begin_inset CommandInset ref
  3502. LatexCommand ref
  3503. reference "fig:near-promoter-peak-enrich"
  3504. plural "false"
  3505. caps "false"
  3506. noprefix "false"
  3507. \end_inset
  3508. .
  3509. The majority of each density distribution is flat, representing the background
  3510. density of peaks genome-wide.
  3511. Each distribution has a peak near zero, representing an enrichment of peaks
  3512. close transcription start site (TSS) positions relative to the remainder
  3513. of the genome.
  3514. Interestingly, the
  3515. \begin_inset Quotes eld
  3516. \end_inset
  3517. radius
  3518. \begin_inset Quotes erd
  3519. \end_inset
  3520. within which this enrichment occurs is not the same for every histone mark
  3521. (Table
  3522. \begin_inset CommandInset ref
  3523. LatexCommand ref
  3524. reference "tab:effective-promoter-radius"
  3525. plural "false"
  3526. caps "false"
  3527. noprefix "false"
  3528. \end_inset
  3529. ).
  3530. For H3K4me2 and H3K4me3, peaks are most enriched within 1
  3531. \begin_inset space ~
  3532. \end_inset
  3533. kbp of TSS positions, while for H3K27me3, enrichment is broader, extending
  3534. to 2.5
  3535. \begin_inset space ~
  3536. \end_inset
  3537. kbp.
  3538. These
  3539. \begin_inset Quotes eld
  3540. \end_inset
  3541. effective promoter radii
  3542. \begin_inset Quotes erd
  3543. \end_inset
  3544. remain approximately the same across all combinations of experimental condition
  3545. (cell type, time point, and donor), so they appear to be a property of
  3546. the histone mark itself.
  3547. Hence, these radii were used to define the promoter regions for each histone
  3548. mark in all further analyses.
  3549. \end_layout
  3550. \begin_layout Standard
  3551. \begin_inset Flex TODO Note (inline)
  3552. status open
  3553. \begin_layout Plain Layout
  3554. Consider also showing figure for distance to nearest peak center, and reference
  3555. median peak size once that is known.
  3556. \end_layout
  3557. \end_inset
  3558. \end_layout
  3559. \begin_layout Subsection
  3560. H3K4 and H3K27 promoter methylation has broadly the expected correlation
  3561. with gene expression
  3562. \end_layout
  3563. \begin_layout Standard
  3564. \begin_inset Float figure
  3565. wide false
  3566. sideways false
  3567. status collapsed
  3568. \begin_layout Plain Layout
  3569. \begin_inset Flex TODO Note (inline)
  3570. status open
  3571. \begin_layout Plain Layout
  3572. This figure is generated from the old analysis.
  3573. Either note that in some way or re-generate it from the new peak calls.
  3574. \end_layout
  3575. \end_inset
  3576. \end_layout
  3577. \begin_layout Plain Layout
  3578. \align center
  3579. \begin_inset Graphics
  3580. filename graphics/CD4-csaw/FPKM by Peak Violin Plots-CROP.pdf
  3581. lyxscale 50
  3582. width 100col%
  3583. \end_inset
  3584. \end_layout
  3585. \begin_layout Plain Layout
  3586. \begin_inset Caption Standard
  3587. \begin_layout Plain Layout
  3588. \series bold
  3589. \begin_inset CommandInset label
  3590. LatexCommand label
  3591. name "fig:fpkm-by-peak"
  3592. \end_inset
  3593. Expression distributions of genes with and without promoter peaks.
  3594. \end_layout
  3595. \end_inset
  3596. \end_layout
  3597. \end_inset
  3598. \end_layout
  3599. \begin_layout Standard
  3600. H3K4me2 and H3K4me2 have previously been reported as activating marks whose
  3601. presence in a gene's promoter is associated with higher gene expression,
  3602. while H3K27me3 has been reported as inactivating [CITE].
  3603. The data are consistent with this characterization: genes whose promoters
  3604. (as defined by the radii for each histone mark listed in
  3605. \begin_inset CommandInset ref
  3606. LatexCommand ref
  3607. reference "tab:effective-promoter-radius"
  3608. plural "false"
  3609. caps "false"
  3610. noprefix "false"
  3611. \end_inset
  3612. ) overlap with a H3K4me2 or H3K4me3 peak tend to have higher expression
  3613. than those that don't, while H3K27me3 is likewise associated with lower
  3614. gene expression, as shown in
  3615. \begin_inset CommandInset ref
  3616. LatexCommand ref
  3617. reference "fig:fpkm-by-peak"
  3618. plural "false"
  3619. caps "false"
  3620. noprefix "false"
  3621. \end_inset
  3622. .
  3623. This pattern holds across all combinations of cell type and time point
  3624. (Welch's
  3625. \emph on
  3626. t
  3627. \emph default
  3628. -test, all
  3629. \begin_inset Formula $p\mathrm{-values}\ll2.2\times10^{-16}$
  3630. \end_inset
  3631. ).
  3632. The difference in average log FPKM values when a peak overlaps the promoter
  3633. is about
  3634. \begin_inset Formula $+5.67$
  3635. \end_inset
  3636. for H3K4me2,
  3637. \begin_inset Formula $+5.76$
  3638. \end_inset
  3639. for H3K4me2, and
  3640. \begin_inset Formula $-4.00$
  3641. \end_inset
  3642. for H3K27me3.
  3643. \end_layout
  3644. \begin_layout Standard
  3645. \begin_inset Flex TODO Note (inline)
  3646. status open
  3647. \begin_layout Plain Layout
  3648. I also have some figures looking at interactions between marks (e.g.
  3649. what if a promoter has both H3K4me3 and H3K27me3), but I don't know if
  3650. that much detail is warranted here, since all the effects just seem approximate
  3651. ly additive anyway.
  3652. \end_layout
  3653. \end_inset
  3654. \end_layout
  3655. \begin_layout Subsection
  3656. Gene expression and promoter histone methylation patterns in naive and memory
  3657. show convergence at day 14
  3658. \end_layout
  3659. \begin_layout Standard
  3660. \begin_inset ERT
  3661. status open
  3662. \begin_layout Plain Layout
  3663. \backslash
  3664. afterpage{
  3665. \end_layout
  3666. \begin_layout Plain Layout
  3667. \backslash
  3668. begin{landscape}
  3669. \end_layout
  3670. \end_inset
  3671. \end_layout
  3672. \begin_layout Standard
  3673. \begin_inset Float table
  3674. wide false
  3675. sideways false
  3676. status open
  3677. \begin_layout Plain Layout
  3678. \align center
  3679. \begin_inset Tabular
  3680. <lyxtabular version="3" rows="6" columns="7">
  3681. <features tabularvalignment="middle">
  3682. <column alignment="center" valignment="top">
  3683. <column alignment="center" valignment="top">
  3684. <column alignment="center" valignment="top">
  3685. <column alignment="center" valignment="top">
  3686. <column alignment="center" valignment="top">
  3687. <column alignment="center" valignment="top">
  3688. <column alignment="center" valignment="top">
  3689. <row>
  3690. <cell alignment="center" valignment="top" usebox="none">
  3691. \begin_inset Text
  3692. \begin_layout Plain Layout
  3693. \end_layout
  3694. \end_inset
  3695. </cell>
  3696. <cell multicolumn="1" alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  3697. \begin_inset Text
  3698. \begin_layout Plain Layout
  3699. Number of significant promoters
  3700. \end_layout
  3701. \end_inset
  3702. </cell>
  3703. <cell multicolumn="2" alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  3704. \begin_inset Text
  3705. \begin_layout Plain Layout
  3706. \end_layout
  3707. \end_inset
  3708. </cell>
  3709. <cell multicolumn="2" alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  3710. \begin_inset Text
  3711. \begin_layout Plain Layout
  3712. \end_layout
  3713. \end_inset
  3714. </cell>
  3715. <cell multicolumn="1" alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  3716. \begin_inset Text
  3717. \begin_layout Plain Layout
  3718. Est.
  3719. differentially modified promoters
  3720. \end_layout
  3721. \end_inset
  3722. </cell>
  3723. <cell multicolumn="2" alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  3724. \begin_inset Text
  3725. \begin_layout Plain Layout
  3726. \end_layout
  3727. \end_inset
  3728. </cell>
  3729. <cell multicolumn="2" alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  3730. \begin_inset Text
  3731. \begin_layout Plain Layout
  3732. \end_layout
  3733. \end_inset
  3734. </cell>
  3735. </row>
  3736. <row>
  3737. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  3738. \begin_inset Text
  3739. \begin_layout Plain Layout
  3740. Time Point
  3741. \end_layout
  3742. \end_inset
  3743. </cell>
  3744. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  3745. \begin_inset Text
  3746. \begin_layout Plain Layout
  3747. H3K4me2
  3748. \end_layout
  3749. \end_inset
  3750. </cell>
  3751. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  3752. \begin_inset Text
  3753. \begin_layout Plain Layout
  3754. H3K4me3
  3755. \end_layout
  3756. \end_inset
  3757. </cell>
  3758. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  3759. \begin_inset Text
  3760. \begin_layout Plain Layout
  3761. H3K27me3
  3762. \end_layout
  3763. \end_inset
  3764. </cell>
  3765. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  3766. \begin_inset Text
  3767. \begin_layout Plain Layout
  3768. H3K4me2
  3769. \end_layout
  3770. \end_inset
  3771. </cell>
  3772. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  3773. \begin_inset Text
  3774. \begin_layout Plain Layout
  3775. H3K4me3
  3776. \end_layout
  3777. \end_inset
  3778. </cell>
  3779. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  3780. \begin_inset Text
  3781. \begin_layout Plain Layout
  3782. H3K27me3
  3783. \end_layout
  3784. \end_inset
  3785. </cell>
  3786. </row>
  3787. <row>
  3788. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  3789. \begin_inset Text
  3790. \begin_layout Plain Layout
  3791. Day 0
  3792. \end_layout
  3793. \end_inset
  3794. </cell>
  3795. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  3796. \begin_inset Text
  3797. \begin_layout Plain Layout
  3798. 4553
  3799. \end_layout
  3800. \end_inset
  3801. </cell>
  3802. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  3803. \begin_inset Text
  3804. \begin_layout Plain Layout
  3805. 927
  3806. \end_layout
  3807. \end_inset
  3808. </cell>
  3809. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  3810. \begin_inset Text
  3811. \begin_layout Plain Layout
  3812. 6
  3813. \end_layout
  3814. \end_inset
  3815. </cell>
  3816. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  3817. \begin_inset Text
  3818. \begin_layout Plain Layout
  3819. 9967
  3820. \end_layout
  3821. \end_inset
  3822. </cell>
  3823. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  3824. \begin_inset Text
  3825. \begin_layout Plain Layout
  3826. 4149
  3827. \end_layout
  3828. \end_inset
  3829. </cell>
  3830. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  3831. \begin_inset Text
  3832. \begin_layout Plain Layout
  3833. 2404
  3834. \end_layout
  3835. \end_inset
  3836. </cell>
  3837. </row>
  3838. <row>
  3839. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  3840. \begin_inset Text
  3841. \begin_layout Plain Layout
  3842. Day 1
  3843. \end_layout
  3844. \end_inset
  3845. </cell>
  3846. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  3847. \begin_inset Text
  3848. \begin_layout Plain Layout
  3849. 567
  3850. \end_layout
  3851. \end_inset
  3852. </cell>
  3853. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  3854. \begin_inset Text
  3855. \begin_layout Plain Layout
  3856. 278
  3857. \end_layout
  3858. \end_inset
  3859. </cell>
  3860. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  3861. \begin_inset Text
  3862. \begin_layout Plain Layout
  3863. 1570
  3864. \end_layout
  3865. \end_inset
  3866. </cell>
  3867. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  3868. \begin_inset Text
  3869. \begin_layout Plain Layout
  3870. 4370
  3871. \end_layout
  3872. \end_inset
  3873. </cell>
  3874. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  3875. \begin_inset Text
  3876. \begin_layout Plain Layout
  3877. 2145
  3878. \end_layout
  3879. \end_inset
  3880. </cell>
  3881. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  3882. \begin_inset Text
  3883. \begin_layout Plain Layout
  3884. 6598
  3885. \end_layout
  3886. \end_inset
  3887. </cell>
  3888. </row>
  3889. <row>
  3890. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  3891. \begin_inset Text
  3892. \begin_layout Plain Layout
  3893. Day 5
  3894. \end_layout
  3895. \end_inset
  3896. </cell>
  3897. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  3898. \begin_inset Text
  3899. \begin_layout Plain Layout
  3900. 2313
  3901. \end_layout
  3902. \end_inset
  3903. </cell>
  3904. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  3905. \begin_inset Text
  3906. \begin_layout Plain Layout
  3907. 139
  3908. \end_layout
  3909. \end_inset
  3910. </cell>
  3911. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  3912. \begin_inset Text
  3913. \begin_layout Plain Layout
  3914. 490
  3915. \end_layout
  3916. \end_inset
  3917. </cell>
  3918. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  3919. \begin_inset Text
  3920. \begin_layout Plain Layout
  3921. 9450
  3922. \end_layout
  3923. \end_inset
  3924. </cell>
  3925. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  3926. \begin_inset Text
  3927. \begin_layout Plain Layout
  3928. 1148
  3929. \end_layout
  3930. \end_inset
  3931. </cell>
  3932. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  3933. \begin_inset Text
  3934. \begin_layout Plain Layout
  3935. 4141
  3936. \end_layout
  3937. \end_inset
  3938. </cell>
  3939. </row>
  3940. <row>
  3941. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  3942. \begin_inset Text
  3943. \begin_layout Plain Layout
  3944. Day 14
  3945. \end_layout
  3946. \end_inset
  3947. </cell>
  3948. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  3949. \begin_inset Text
  3950. \begin_layout Plain Layout
  3951. 0
  3952. \end_layout
  3953. \end_inset
  3954. </cell>
  3955. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  3956. \begin_inset Text
  3957. \begin_layout Plain Layout
  3958. 0
  3959. \end_layout
  3960. \end_inset
  3961. </cell>
  3962. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  3963. \begin_inset Text
  3964. \begin_layout Plain Layout
  3965. 0
  3966. \end_layout
  3967. \end_inset
  3968. </cell>
  3969. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  3970. \begin_inset Text
  3971. \begin_layout Plain Layout
  3972. 0
  3973. \end_layout
  3974. \end_inset
  3975. </cell>
  3976. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  3977. \begin_inset Text
  3978. \begin_layout Plain Layout
  3979. 0
  3980. \end_layout
  3981. \end_inset
  3982. </cell>
  3983. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  3984. \begin_inset Text
  3985. \begin_layout Plain Layout
  3986. 0
  3987. \end_layout
  3988. \end_inset
  3989. </cell>
  3990. </row>
  3991. </lyxtabular>
  3992. \end_inset
  3993. \end_layout
  3994. \begin_layout Plain Layout
  3995. \begin_inset Caption Standard
  3996. \begin_layout Plain Layout
  3997. \series bold
  3998. \begin_inset CommandInset label
  3999. LatexCommand label
  4000. name "tab:Number-signif-promoters"
  4001. \end_inset
  4002. Number of differentially modified promoters between naive and memory cells
  4003. at each time point after activation.
  4004. \series default
  4005. This table shows both the number of differentially modified promoters detected
  4006. at a 10% FDR threshold (left half), and the total number of differentially
  4007. modified promoters as estimated using the method of
  4008. \begin_inset CommandInset citation
  4009. LatexCommand cite
  4010. key "Phipson2013"
  4011. literal "false"
  4012. \end_inset
  4013. (right half).
  4014. \end_layout
  4015. \end_inset
  4016. \end_layout
  4017. \end_inset
  4018. \end_layout
  4019. \begin_layout Standard
  4020. \begin_inset ERT
  4021. status open
  4022. \begin_layout Plain Layout
  4023. \backslash
  4024. end{landscape}
  4025. \end_layout
  4026. \begin_layout Plain Layout
  4027. }
  4028. \end_layout
  4029. \end_inset
  4030. \end_layout
  4031. \begin_layout Standard
  4032. \begin_inset Float figure
  4033. placement p
  4034. wide false
  4035. sideways false
  4036. status open
  4037. \begin_layout Plain Layout
  4038. \align center
  4039. \begin_inset Float figure
  4040. wide false
  4041. sideways false
  4042. status open
  4043. \begin_layout Plain Layout
  4044. \align center
  4045. \begin_inset Graphics
  4046. filename graphics/CD4-csaw/ChIP-seq/H3K4me2-promoter-PCA-group-CROP.png
  4047. lyxscale 25
  4048. width 45col%
  4049. groupId pcoa-prom-subfig
  4050. \end_inset
  4051. \end_layout
  4052. \begin_layout Plain Layout
  4053. \begin_inset Caption Standard
  4054. \begin_layout Plain Layout
  4055. \series bold
  4056. \begin_inset CommandInset label
  4057. LatexCommand label
  4058. name "fig:PCoA-H3K4me2-prom"
  4059. \end_inset
  4060. PCoA plot of H3K4me2 promoters, after subtracting surrogate variables
  4061. \end_layout
  4062. \end_inset
  4063. \end_layout
  4064. \end_inset
  4065. \begin_inset space \hfill{}
  4066. \end_inset
  4067. \begin_inset Float figure
  4068. wide false
  4069. sideways false
  4070. status open
  4071. \begin_layout Plain Layout
  4072. \align center
  4073. \begin_inset Graphics
  4074. filename graphics/CD4-csaw/ChIP-seq/H3K4me3-promoter-PCA-group-CROP.png
  4075. lyxscale 25
  4076. width 45col%
  4077. groupId pcoa-prom-subfig
  4078. \end_inset
  4079. \end_layout
  4080. \begin_layout Plain Layout
  4081. \begin_inset Caption Standard
  4082. \begin_layout Plain Layout
  4083. \series bold
  4084. \begin_inset CommandInset label
  4085. LatexCommand label
  4086. name "fig:PCoA-H3K4me3-prom"
  4087. \end_inset
  4088. PCoA plot of H3K4me3 promoters, after subtracting surrogate variables
  4089. \end_layout
  4090. \end_inset
  4091. \end_layout
  4092. \end_inset
  4093. \end_layout
  4094. \begin_layout Plain Layout
  4095. \align center
  4096. \begin_inset Float figure
  4097. wide false
  4098. sideways false
  4099. status collapsed
  4100. \begin_layout Plain Layout
  4101. \align center
  4102. \begin_inset Graphics
  4103. filename graphics/CD4-csaw/ChIP-seq/H3K27me3-promoter-PCA-group-CROP.png
  4104. lyxscale 25
  4105. width 45col%
  4106. groupId pcoa-prom-subfig
  4107. \end_inset
  4108. \end_layout
  4109. \begin_layout Plain Layout
  4110. \begin_inset Caption Standard
  4111. \begin_layout Plain Layout
  4112. \series bold
  4113. \begin_inset CommandInset label
  4114. LatexCommand label
  4115. name "fig:PCoA-H3K27me3-prom"
  4116. \end_inset
  4117. PCoA plot of H3K27me3 promoters, after subtracting surrogate variables
  4118. \end_layout
  4119. \end_inset
  4120. \end_layout
  4121. \end_inset
  4122. \begin_inset space \hfill{}
  4123. \end_inset
  4124. \begin_inset Float figure
  4125. wide false
  4126. sideways false
  4127. status open
  4128. \begin_layout Plain Layout
  4129. \align center
  4130. \begin_inset Graphics
  4131. filename graphics/CD4-csaw/RNA-seq/PCA-final-23-CROP.png
  4132. lyxscale 25
  4133. width 45col%
  4134. groupId pcoa-prom-subfig
  4135. \end_inset
  4136. \end_layout
  4137. \begin_layout Plain Layout
  4138. \begin_inset Caption Standard
  4139. \begin_layout Plain Layout
  4140. \series bold
  4141. \begin_inset CommandInset label
  4142. LatexCommand label
  4143. name "fig:RNA-PCA-group"
  4144. \end_inset
  4145. RNA-seq PCoA showing principal coordinates 2 and 3.
  4146. \end_layout
  4147. \end_inset
  4148. \end_layout
  4149. \end_inset
  4150. \end_layout
  4151. \begin_layout Plain Layout
  4152. \begin_inset Caption Standard
  4153. \begin_layout Plain Layout
  4154. \series bold
  4155. \begin_inset CommandInset label
  4156. LatexCommand label
  4157. name "fig:PCoA-promoters"
  4158. \end_inset
  4159. PCoA plots for promoter ChIP-seq and expression RNA-seq data
  4160. \end_layout
  4161. \end_inset
  4162. \end_layout
  4163. \end_inset
  4164. \end_layout
  4165. \begin_layout Standard
  4166. \begin_inset Flex TODO Note (inline)
  4167. status open
  4168. \begin_layout Plain Layout
  4169. Check up on figure refs in this paragraph
  4170. \end_layout
  4171. \end_inset
  4172. \end_layout
  4173. \begin_layout Standard
  4174. We hypothesized that if naive cells had differentiated into memory cells
  4175. by Day 14, then their patterns of expression and histone modification should
  4176. converge with those of memory cells at Day 14.
  4177. Figure
  4178. \begin_inset CommandInset ref
  4179. LatexCommand ref
  4180. reference "fig:PCoA-promoters"
  4181. plural "false"
  4182. caps "false"
  4183. noprefix "false"
  4184. \end_inset
  4185. shows the patterns of variation in all 3 histone marks in the promoter
  4186. regions of the genome using principal coordinate analysis.
  4187. All 3 marks show a noticeable convergence between the naive and memory
  4188. samples at day 14, visible as an overlapping of the day 14 groups on each
  4189. plot.
  4190. This is consistent with the counts of significantly differentially modified
  4191. promoters and estimates of the total numbers of differentially modified
  4192. promoters shown in Table
  4193. \begin_inset CommandInset ref
  4194. LatexCommand ref
  4195. reference "tab:Number-signif-promoters"
  4196. plural "false"
  4197. caps "false"
  4198. noprefix "false"
  4199. \end_inset
  4200. .
  4201. For all histone marks, evidence of differential modification between naive
  4202. and memory samples was detected at every time point except day 14.
  4203. The day 14 convergence pattern is also present in the RNA-seq data (Figure
  4204. \begin_inset CommandInset ref
  4205. LatexCommand ref
  4206. reference "fig:RNA-PCA-group"
  4207. plural "false"
  4208. caps "false"
  4209. noprefix "false"
  4210. \end_inset
  4211. ), albeit in the 2nd and 3rd principal coordinates, indicating that it is
  4212. not the most dominant pattern driving gene expression.
  4213. Taken together, the data show that promoter histone methylation for these
  4214. 3 histone marks and RNA expression for naive and memory cells are most
  4215. similar at day 14, the furthest time point after activation.
  4216. MOFA was also able to capture this day 14 convergence pattern in latent
  4217. factor 5 (Figure
  4218. \begin_inset CommandInset ref
  4219. LatexCommand ref
  4220. reference "fig:mofa-lf-scatter"
  4221. plural "false"
  4222. caps "false"
  4223. noprefix "false"
  4224. \end_inset
  4225. ), which accounts for shared variation across all 3 histone marks and the
  4226. RNA-seq data, confirming that this convergence is a coordinated pattern
  4227. across all 4 data sets.
  4228. While this observation does not prove that the naive cells have differentiated
  4229. into memory cells at Day 14, it is consistent with that hypothesis.
  4230. \end_layout
  4231. \begin_layout Subsection
  4232. Effect of H3K4me2 and H3K4me3 promoter coverage upstream vs downstream of
  4233. TSS
  4234. \end_layout
  4235. \begin_layout Standard
  4236. \begin_inset Flex TODO Note (inline)
  4237. status open
  4238. \begin_layout Plain Layout
  4239. Need a better section title, for this and the next one.
  4240. \end_layout
  4241. \end_inset
  4242. \end_layout
  4243. \begin_layout Standard
  4244. \begin_inset Flex TODO Note (inline)
  4245. status open
  4246. \begin_layout Plain Layout
  4247. Make sure use of coverage/abundance/whatever is consistent.
  4248. \end_layout
  4249. \end_inset
  4250. \end_layout
  4251. \begin_layout Standard
  4252. \begin_inset Flex TODO Note (inline)
  4253. status open
  4254. \begin_layout Plain Layout
  4255. For the figures in this section and the next, the group labels are arbitrary,
  4256. so if time allows, it would be good to manually reorder them in a logical
  4257. way, e.g.
  4258. most upstream to most downstream.
  4259. If this is done, make sure to update the text with the correct group labels.
  4260. \end_layout
  4261. \end_inset
  4262. \end_layout
  4263. \begin_layout Standard
  4264. \begin_inset ERT
  4265. status open
  4266. \begin_layout Plain Layout
  4267. \backslash
  4268. afterpage{
  4269. \end_layout
  4270. \begin_layout Plain Layout
  4271. \backslash
  4272. begin{landscape}
  4273. \end_layout
  4274. \end_inset
  4275. \end_layout
  4276. \begin_layout Standard
  4277. \begin_inset Float figure
  4278. wide false
  4279. sideways false
  4280. status open
  4281. \begin_layout Plain Layout
  4282. \align center
  4283. \begin_inset Float figure
  4284. wide false
  4285. sideways false
  4286. status open
  4287. \begin_layout Plain Layout
  4288. \align center
  4289. \begin_inset Graphics
  4290. filename graphics/CD4-csaw/ChIP-seq/H3K4me2-neighborhood-clusters-CROP.png
  4291. lyxscale 25
  4292. width 30col%
  4293. groupId covprof-subfig
  4294. \end_inset
  4295. \end_layout
  4296. \begin_layout Plain Layout
  4297. \begin_inset Caption Standard
  4298. \begin_layout Plain Layout
  4299. \series bold
  4300. \begin_inset CommandInset label
  4301. LatexCommand label
  4302. name "fig:H3K4me2-neighborhood-clusters"
  4303. \end_inset
  4304. Average relative coverage for each bin in each cluster
  4305. \end_layout
  4306. \end_inset
  4307. \end_layout
  4308. \end_inset
  4309. \begin_inset space \hfill{}
  4310. \end_inset
  4311. \begin_inset Float figure
  4312. wide false
  4313. sideways false
  4314. status open
  4315. \begin_layout Plain Layout
  4316. \align center
  4317. \begin_inset Graphics
  4318. filename graphics/CD4-csaw/ChIP-seq/H3K4me2-neighborhood-PCA-CROP.png
  4319. lyxscale 25
  4320. width 30col%
  4321. groupId covprof-subfig
  4322. \end_inset
  4323. \end_layout
  4324. \begin_layout Plain Layout
  4325. \begin_inset Caption Standard
  4326. \begin_layout Plain Layout
  4327. \series bold
  4328. \begin_inset CommandInset label
  4329. LatexCommand label
  4330. name "fig:H3K4me2-neighborhood-pca"
  4331. \end_inset
  4332. PCA of relative coverage depth, colored by K-means cluster membership.
  4333. \end_layout
  4334. \end_inset
  4335. \end_layout
  4336. \end_inset
  4337. \begin_inset space \hfill{}
  4338. \end_inset
  4339. \begin_inset Float figure
  4340. wide false
  4341. sideways false
  4342. status open
  4343. \begin_layout Plain Layout
  4344. \align center
  4345. \begin_inset Graphics
  4346. filename graphics/CD4-csaw/ChIP-seq/H3K4me2-neighborhood-expression-CROP.png
  4347. lyxscale 25
  4348. width 30col%
  4349. groupId covprof-subfig
  4350. \end_inset
  4351. \end_layout
  4352. \begin_layout Plain Layout
  4353. \begin_inset Caption Standard
  4354. \begin_layout Plain Layout
  4355. \series bold
  4356. \begin_inset CommandInset label
  4357. LatexCommand label
  4358. name "fig:H3K4me2-neighborhood-expression"
  4359. \end_inset
  4360. Gene expression grouped by promoter coverage clusters.
  4361. \end_layout
  4362. \end_inset
  4363. \end_layout
  4364. \end_inset
  4365. \end_layout
  4366. \begin_layout Plain Layout
  4367. \begin_inset Caption Standard
  4368. \begin_layout Plain Layout
  4369. \series bold
  4370. \begin_inset CommandInset label
  4371. LatexCommand label
  4372. name "fig:H3K4me2-neighborhood"
  4373. \end_inset
  4374. K-means clustering of promoter H3K4me2 relative coverage depth in naive
  4375. day 0 samples.
  4376. \series default
  4377. H3K4me2 ChIP-seq reads were binned into 500-bp windows tiled across each
  4378. promoter from 5
  4379. \begin_inset space ~
  4380. \end_inset
  4381. kbp upstream to 5
  4382. \begin_inset space ~
  4383. \end_inset
  4384. kbp downstream, and the logCPM values were normalized within each promoter
  4385. to an average of 0, yielding relative coverage depths.
  4386. These were then grouped using K-means clustering with
  4387. \begin_inset Formula $K=6$
  4388. \end_inset
  4389. ,
  4390. \series bold
  4391. \series default
  4392. and the average bin values were plotted for each cluster (a).
  4393. The
  4394. \begin_inset Formula $x$
  4395. \end_inset
  4396. -axis is the genomic coordinate of each bin relative to the the transcription
  4397. start site, and the
  4398. \begin_inset Formula $y$
  4399. \end_inset
  4400. -axis is the mean relative coverage depth of that bin across all promoters
  4401. in the cluster.
  4402. Each line represents the average
  4403. \begin_inset Quotes eld
  4404. \end_inset
  4405. shape
  4406. \begin_inset Quotes erd
  4407. \end_inset
  4408. of the promoter coverage for promoters in that cluster.
  4409. PCA was performed on the same data, and the first two principal components
  4410. were plotted, coloring each point by its K-means cluster identity (b).
  4411. For each cluster, the distribution of gene expression values was plotted
  4412. (c).
  4413. \end_layout
  4414. \end_inset
  4415. \end_layout
  4416. \end_inset
  4417. \end_layout
  4418. \begin_layout Standard
  4419. \begin_inset ERT
  4420. status open
  4421. \begin_layout Plain Layout
  4422. \backslash
  4423. end{landscape}
  4424. \end_layout
  4425. \begin_layout Plain Layout
  4426. }
  4427. \end_layout
  4428. \end_inset
  4429. \end_layout
  4430. \begin_layout Standard
  4431. To test whether the position of a histone mark relative to a gene's transcriptio
  4432. n start site (TSS) was important, we looked at the
  4433. \begin_inset Quotes eld
  4434. \end_inset
  4435. landscape
  4436. \begin_inset Quotes erd
  4437. \end_inset
  4438. of ChIP-seq read coverage in naive Day 0 samples within 5 kb of each gene's
  4439. TSS by binning reads into 500-bp windows tiled across each promoter LogCPM
  4440. values were calculated for the bins in each promoter and then the average
  4441. logCPM for each promoter's bins was normalized to zero, such that the values
  4442. represent coverage relative to other regions of the same promoter rather
  4443. than being proportional to absolute read count.
  4444. The promoters were then clustered based on the normalized bin abundances
  4445. using
  4446. \begin_inset Formula $k$
  4447. \end_inset
  4448. -means clustering with
  4449. \begin_inset Formula $K=6$
  4450. \end_inset
  4451. .
  4452. Different values of
  4453. \begin_inset Formula $K$
  4454. \end_inset
  4455. were also tested, but did not substantially change the interpretation of
  4456. the data.
  4457. \end_layout
  4458. \begin_layout Standard
  4459. For H3K4me2, plotting the average bin abundances for each cluster reveals
  4460. a simple pattern (Figure
  4461. \begin_inset CommandInset ref
  4462. LatexCommand ref
  4463. reference "fig:H3K4me2-neighborhood-clusters"
  4464. plural "false"
  4465. caps "false"
  4466. noprefix "false"
  4467. \end_inset
  4468. ): Cluster 5 represents a completely flat promoter coverage profile, likely
  4469. consisting of genes with no H3K4me2 methylation in the promoter.
  4470. All the other clusters represent a continuum of peak positions relative
  4471. to the TSS.
  4472. In order from must upstream to most downstream, they are Clusters 6, 4,
  4473. 3, 1, and 2.
  4474. There do not appear to be any clusters representing coverage patterns other
  4475. than lone peaks, such as coverage troughs or double peaks.
  4476. Next, all promoters were plotted in a PCA plot based on the same relative
  4477. bin abundance data, and colored based on cluster membership (Figure
  4478. \begin_inset CommandInset ref
  4479. LatexCommand ref
  4480. reference "fig:H3K4me2-neighborhood-pca"
  4481. plural "false"
  4482. caps "false"
  4483. noprefix "false"
  4484. \end_inset
  4485. ).
  4486. The PCA plot shows Cluster 5 (the
  4487. \begin_inset Quotes eld
  4488. \end_inset
  4489. no peak
  4490. \begin_inset Quotes erd
  4491. \end_inset
  4492. cluster) at the center, with the other clusters arranged in a counter-clockwise
  4493. arc around it in the order noted above, from most upstream peak to most
  4494. downstream.
  4495. Notably, the
  4496. \begin_inset Quotes eld
  4497. \end_inset
  4498. clusters
  4499. \begin_inset Quotes erd
  4500. \end_inset
  4501. form a single large
  4502. \begin_inset Quotes eld
  4503. \end_inset
  4504. cloud
  4505. \begin_inset Quotes erd
  4506. \end_inset
  4507. with no apparent separation between them, further supporting the conclusion
  4508. that these clusters represent an arbitrary partitioning of a continuous
  4509. distribution of promoter coverage landscapes.
  4510. While the clusters are a useful abstraction that aids in visualization,
  4511. they are ultimately not an accurate representation of the data.
  4512. A better representation might be something like a polar coordinate system
  4513. with the origin at the center of Cluster 5, where the radius represents
  4514. the peak height above the background and the angle represents the peak's
  4515. position upstream or downstream of the TSS.
  4516. The continuous nature of the distribution also explains why different values
  4517. of
  4518. \begin_inset Formula $K$
  4519. \end_inset
  4520. led to similar conclusions.
  4521. \end_layout
  4522. \begin_layout Standard
  4523. \begin_inset Flex TODO Note (inline)
  4524. status open
  4525. \begin_layout Plain Layout
  4526. RNA-seq values in the plots use logCPM but should really use logFPKM or
  4527. logTPM.
  4528. Fix if time allows.
  4529. \end_layout
  4530. \end_inset
  4531. \end_layout
  4532. \begin_layout Standard
  4533. \begin_inset Flex TODO Note (inline)
  4534. status open
  4535. \begin_layout Plain Layout
  4536. Should have a table of p-values on difference of means between Cluster 5
  4537. and the others.
  4538. \end_layout
  4539. \end_inset
  4540. \end_layout
  4541. \begin_layout Standard
  4542. To investigate the association between relative peak position and gene expressio
  4543. n, we plotted the Naive Day 0 expression for the genes in each cluster (Figure
  4544. \begin_inset CommandInset ref
  4545. LatexCommand ref
  4546. reference "fig:H3K4me2-neighborhood-expression"
  4547. plural "false"
  4548. caps "false"
  4549. noprefix "false"
  4550. \end_inset
  4551. ).
  4552. Most genes in Cluster 5, the
  4553. \begin_inset Quotes eld
  4554. \end_inset
  4555. no peak
  4556. \begin_inset Quotes erd
  4557. \end_inset
  4558. cluster, have low expression values.
  4559. Taking this as the
  4560. \begin_inset Quotes eld
  4561. \end_inset
  4562. baseline
  4563. \begin_inset Quotes erd
  4564. \end_inset
  4565. distribution when no H3K4me2 methylation is present, we can compare the
  4566. other clusters' distributions to determine which peak positions are associated
  4567. with elevated expression.
  4568. As might be expected, the 3 clusters representing peaks closest to the
  4569. TSS, Clusters 1, 3, and 4, show the highest average expression distributions.
  4570. Specifically, these clusters all have their highest ChIP-seq abundance
  4571. within 1kb of the TSS, consistent with the previously determined promoter
  4572. radius.
  4573. In contrast, cluster 6, which represents peaks several kb upstream of the
  4574. TSS, shows a slightly higher average expression than baseline, while Cluster
  4575. 2, which represents peaks several kb downstream, doesn't appear to show
  4576. any appreciable difference.
  4577. Interestingly, the cluster with the highest average expression is Cluster
  4578. 1, which represents peaks about 1 kb downstream of the TSS, rather than
  4579. Cluster 3, which represents peaks centered directly at the TSS.
  4580. This suggests that conceptualizing the promoter as a region centered on
  4581. the TSS with a certain
  4582. \begin_inset Quotes eld
  4583. \end_inset
  4584. radius
  4585. \begin_inset Quotes erd
  4586. \end_inset
  4587. may be an oversimplification – a peak that is a specific distance from
  4588. the TSS may have a different degree of influence depending on whether it
  4589. is upstream or downstream of the TSS.
  4590. \end_layout
  4591. \begin_layout Standard
  4592. \begin_inset ERT
  4593. status open
  4594. \begin_layout Plain Layout
  4595. \backslash
  4596. afterpage{
  4597. \end_layout
  4598. \begin_layout Plain Layout
  4599. \backslash
  4600. begin{landscape}
  4601. \end_layout
  4602. \end_inset
  4603. \end_layout
  4604. \begin_layout Standard
  4605. \begin_inset Float figure
  4606. wide false
  4607. sideways false
  4608. status open
  4609. \begin_layout Plain Layout
  4610. \align center
  4611. \begin_inset Float figure
  4612. wide false
  4613. sideways false
  4614. status open
  4615. \begin_layout Plain Layout
  4616. \align center
  4617. \begin_inset Graphics
  4618. filename graphics/CD4-csaw/ChIP-seq/H3K4me3-neighborhood-clusters-CROP.png
  4619. lyxscale 25
  4620. width 30col%
  4621. groupId covprof-subfig
  4622. \end_inset
  4623. \end_layout
  4624. \begin_layout Plain Layout
  4625. \begin_inset Caption Standard
  4626. \begin_layout Plain Layout
  4627. \series bold
  4628. \begin_inset CommandInset label
  4629. LatexCommand label
  4630. name "fig:H3K4me3-neighborhood-clusters"
  4631. \end_inset
  4632. Average relative coverage for each bin in each cluster
  4633. \end_layout
  4634. \end_inset
  4635. \end_layout
  4636. \end_inset
  4637. \begin_inset space \hfill{}
  4638. \end_inset
  4639. \begin_inset Float figure
  4640. wide false
  4641. sideways false
  4642. status open
  4643. \begin_layout Plain Layout
  4644. \align center
  4645. \begin_inset Graphics
  4646. filename graphics/CD4-csaw/ChIP-seq/H3K4me3-neighborhood-PCA-CROP.png
  4647. lyxscale 25
  4648. width 30col%
  4649. groupId covprof-subfig
  4650. \end_inset
  4651. \end_layout
  4652. \begin_layout Plain Layout
  4653. \begin_inset Caption Standard
  4654. \begin_layout Plain Layout
  4655. \series bold
  4656. \begin_inset CommandInset label
  4657. LatexCommand label
  4658. name "fig:H3K4me3-neighborhood-pca"
  4659. \end_inset
  4660. PCA of relative coverage depth, colored by K-means cluster membership.
  4661. \end_layout
  4662. \end_inset
  4663. \end_layout
  4664. \end_inset
  4665. \begin_inset space \hfill{}
  4666. \end_inset
  4667. \begin_inset Float figure
  4668. wide false
  4669. sideways false
  4670. status open
  4671. \begin_layout Plain Layout
  4672. \align center
  4673. \begin_inset Graphics
  4674. filename graphics/CD4-csaw/ChIP-seq/H3K4me3-neighborhood-expression-CROP.png
  4675. lyxscale 25
  4676. width 30col%
  4677. groupId covprof-subfig
  4678. \end_inset
  4679. \end_layout
  4680. \begin_layout Plain Layout
  4681. \begin_inset Caption Standard
  4682. \begin_layout Plain Layout
  4683. \series bold
  4684. \begin_inset CommandInset label
  4685. LatexCommand label
  4686. name "fig:H3K4me3-neighborhood-expression"
  4687. \end_inset
  4688. Gene expression grouped by promoter coverage clusters.
  4689. \end_layout
  4690. \end_inset
  4691. \end_layout
  4692. \end_inset
  4693. \end_layout
  4694. \begin_layout Plain Layout
  4695. \begin_inset Caption Standard
  4696. \begin_layout Plain Layout
  4697. \series bold
  4698. \begin_inset CommandInset label
  4699. LatexCommand label
  4700. name "fig:H3K4me3-neighborhood"
  4701. \end_inset
  4702. K-means clustering of promoter H3K4me3 relative coverage depth in naive
  4703. day 0 samples.
  4704. \series default
  4705. H3K4me2 ChIP-seq reads were binned into 500-bp windows tiled across each
  4706. promoter from 5
  4707. \begin_inset space ~
  4708. \end_inset
  4709. kbp upstream to 5
  4710. \begin_inset space ~
  4711. \end_inset
  4712. kbp downstream, and the logCPM values were normalized within each promoter
  4713. to an average of 0, yielding relative coverage depths.
  4714. These were then grouped using K-means clustering with
  4715. \begin_inset Formula $K=6$
  4716. \end_inset
  4717. ,
  4718. \series bold
  4719. \series default
  4720. and the average bin values were plotted for each cluster (a).
  4721. The
  4722. \begin_inset Formula $x$
  4723. \end_inset
  4724. -axis is the genomic coordinate of each bin relative to the the transcription
  4725. start site, and the
  4726. \begin_inset Formula $y$
  4727. \end_inset
  4728. -axis is the mean relative coverage depth of that bin across all promoters
  4729. in the cluster.
  4730. Each line represents the average
  4731. \begin_inset Quotes eld
  4732. \end_inset
  4733. shape
  4734. \begin_inset Quotes erd
  4735. \end_inset
  4736. of the promoter coverage for promoters in that cluster.
  4737. PCA was performed on the same data, and the first two principal components
  4738. were plotted, coloring each point by its K-means cluster identity (b).
  4739. For each cluster, the distribution of gene expression values was plotted
  4740. (c).
  4741. \end_layout
  4742. \end_inset
  4743. \end_layout
  4744. \end_inset
  4745. \end_layout
  4746. \begin_layout Standard
  4747. \begin_inset ERT
  4748. status open
  4749. \begin_layout Plain Layout
  4750. \backslash
  4751. end{landscape}
  4752. \end_layout
  4753. \begin_layout Plain Layout
  4754. }
  4755. \end_layout
  4756. \end_inset
  4757. \end_layout
  4758. \begin_layout Standard
  4759. \begin_inset Flex TODO Note (inline)
  4760. status open
  4761. \begin_layout Plain Layout
  4762. Is there more to say here?
  4763. \end_layout
  4764. \end_inset
  4765. \end_layout
  4766. \begin_layout Standard
  4767. All observations described above for H3K4me2 ChIP-seq also appear to hold
  4768. for H3K4me3 as well (Figure
  4769. \begin_inset CommandInset ref
  4770. LatexCommand ref
  4771. reference "fig:H3K4me3-neighborhood"
  4772. plural "false"
  4773. caps "false"
  4774. noprefix "false"
  4775. \end_inset
  4776. ).
  4777. This is expected, since there is a high correlation between the positions
  4778. where both histone marks occur.
  4779. \end_layout
  4780. \begin_layout Subsection
  4781. Promoter coverage H3K27me3
  4782. \end_layout
  4783. \begin_layout Standard
  4784. \begin_inset ERT
  4785. status open
  4786. \begin_layout Plain Layout
  4787. \backslash
  4788. afterpage{
  4789. \end_layout
  4790. \begin_layout Plain Layout
  4791. \backslash
  4792. begin{landscape}
  4793. \end_layout
  4794. \end_inset
  4795. \end_layout
  4796. \begin_layout Standard
  4797. \begin_inset Float figure
  4798. wide false
  4799. sideways false
  4800. status collapsed
  4801. \begin_layout Plain Layout
  4802. \align center
  4803. \begin_inset Float figure
  4804. wide false
  4805. sideways false
  4806. status open
  4807. \begin_layout Plain Layout
  4808. \align center
  4809. \begin_inset Graphics
  4810. filename graphics/CD4-csaw/ChIP-seq/H3K27me3-neighborhood-clusters-CROP.png
  4811. lyxscale 25
  4812. width 30col%
  4813. groupId covprof-subfig
  4814. \end_inset
  4815. \end_layout
  4816. \begin_layout Plain Layout
  4817. \begin_inset Caption Standard
  4818. \begin_layout Plain Layout
  4819. \series bold
  4820. \begin_inset CommandInset label
  4821. LatexCommand label
  4822. name "fig:H3K27me3-neighborhood-clusters"
  4823. \end_inset
  4824. Average relative coverage for each bin in each cluster
  4825. \end_layout
  4826. \end_inset
  4827. \end_layout
  4828. \end_inset
  4829. \begin_inset space \hfill{}
  4830. \end_inset
  4831. \begin_inset Float figure
  4832. wide false
  4833. sideways false
  4834. status open
  4835. \begin_layout Plain Layout
  4836. \align center
  4837. \begin_inset Graphics
  4838. filename graphics/CD4-csaw/ChIP-seq/H3K27me3-neighborhood-PCA-CROP.png
  4839. lyxscale 25
  4840. width 30col%
  4841. groupId covprof-subfig
  4842. \end_inset
  4843. \end_layout
  4844. \begin_layout Plain Layout
  4845. \begin_inset Caption Standard
  4846. \begin_layout Plain Layout
  4847. \series bold
  4848. \begin_inset CommandInset label
  4849. LatexCommand label
  4850. name "fig:H3K27me3-neighborhood-pca"
  4851. \end_inset
  4852. PCA of relative coverage depth, colored by K-means cluster membership.
  4853. \series default
  4854. Note that Cluster 6 is hidden behind all the other clusters.
  4855. \end_layout
  4856. \end_inset
  4857. \end_layout
  4858. \end_inset
  4859. \begin_inset space \hfill{}
  4860. \end_inset
  4861. \begin_inset Float figure
  4862. wide false
  4863. sideways false
  4864. status open
  4865. \begin_layout Plain Layout
  4866. \align center
  4867. \begin_inset Graphics
  4868. filename graphics/CD4-csaw/ChIP-seq/H3K27me3-neighborhood-expression-CROP.png
  4869. lyxscale 25
  4870. width 30col%
  4871. groupId covprof-subfig
  4872. \end_inset
  4873. \end_layout
  4874. \begin_layout Plain Layout
  4875. \begin_inset Caption Standard
  4876. \begin_layout Plain Layout
  4877. \series bold
  4878. \begin_inset CommandInset label
  4879. LatexCommand label
  4880. name "fig:H3K27me3-neighborhood-expression"
  4881. \end_inset
  4882. Gene expression grouped by promoter coverage clusters.
  4883. \end_layout
  4884. \end_inset
  4885. \end_layout
  4886. \end_inset
  4887. \end_layout
  4888. \begin_layout Plain Layout
  4889. \begin_inset Flex TODO Note (inline)
  4890. status open
  4891. \begin_layout Plain Layout
  4892. Repeated figure legends are kind of an issue here.
  4893. What to do?
  4894. \end_layout
  4895. \end_inset
  4896. \end_layout
  4897. \begin_layout Plain Layout
  4898. \begin_inset Caption Standard
  4899. \begin_layout Plain Layout
  4900. \series bold
  4901. \begin_inset CommandInset label
  4902. LatexCommand label
  4903. name "fig:H3K27me3-neighborhood"
  4904. \end_inset
  4905. K-means clustering of promoter H3K27me3 relative coverage depth in naive
  4906. day 0 samples.
  4907. \series default
  4908. H3K27me3 ChIP-seq reads were binned into 500-bp windows tiled across each
  4909. promoter from 5
  4910. \begin_inset space ~
  4911. \end_inset
  4912. kbp upstream to 5
  4913. \begin_inset space ~
  4914. \end_inset
  4915. kbp downstream, and the logCPM values were normalized within each promoter
  4916. to an average of 0, yielding relative coverage depths.
  4917. These were then grouped using
  4918. \begin_inset Formula $k$
  4919. \end_inset
  4920. -means clustering with
  4921. \begin_inset Formula $K=6$
  4922. \end_inset
  4923. ,
  4924. \series bold
  4925. \series default
  4926. and the average bin values were plotted for each cluster (a).
  4927. The
  4928. \begin_inset Formula $x$
  4929. \end_inset
  4930. -axis is the genomic coordinate of each bin relative to the the transcription
  4931. start site, and the
  4932. \begin_inset Formula $y$
  4933. \end_inset
  4934. -axis is the mean relative coverage depth of that bin across all promoters
  4935. in the cluster.
  4936. Each line represents the average
  4937. \begin_inset Quotes eld
  4938. \end_inset
  4939. shape
  4940. \begin_inset Quotes erd
  4941. \end_inset
  4942. of the promoter coverage for promoters in that cluster.
  4943. PCA was performed on the same data, and the first two principal components
  4944. were plotted, coloring each point by its K-means cluster identity (b).
  4945. For each cluster, the distribution of gene expression values was plotted
  4946. (c).
  4947. \end_layout
  4948. \end_inset
  4949. \end_layout
  4950. \end_inset
  4951. \end_layout
  4952. \begin_layout Standard
  4953. \begin_inset ERT
  4954. status open
  4955. \begin_layout Plain Layout
  4956. \backslash
  4957. end{landscape}
  4958. \end_layout
  4959. \begin_layout Plain Layout
  4960. }
  4961. \end_layout
  4962. \end_inset
  4963. \end_layout
  4964. \begin_layout Standard
  4965. \begin_inset Flex TODO Note (inline)
  4966. status open
  4967. \begin_layout Plain Layout
  4968. Should maybe re-explain what was done or refer back to the previous section.
  4969. \end_layout
  4970. \end_inset
  4971. \end_layout
  4972. \begin_layout Standard
  4973. Unlike both H3K4 marks, whose main patterns of variation appear directly
  4974. related to the size and position of a single peak within the promoter,
  4975. the patterns of H3K27me3 methylation in promoters are more complex (Figure
  4976. \begin_inset CommandInset ref
  4977. LatexCommand ref
  4978. reference "fig:H3K27me3-neighborhood"
  4979. plural "false"
  4980. caps "false"
  4981. noprefix "false"
  4982. \end_inset
  4983. ).
  4984. Once again looking at the relative coverage in a 500-bp wide bins in a
  4985. 5kb radius around each TSS, promoters were clustered based on the normalized
  4986. relative coverage values in each bin using
  4987. \begin_inset Formula $k$
  4988. \end_inset
  4989. -means clustering with
  4990. \begin_inset Formula $K=6$
  4991. \end_inset
  4992. (Figure
  4993. \begin_inset CommandInset ref
  4994. LatexCommand ref
  4995. reference "fig:H3K27me3-neighborhood-clusters"
  4996. plural "false"
  4997. caps "false"
  4998. noprefix "false"
  4999. \end_inset
  5000. ).
  5001. This time, 3
  5002. \begin_inset Quotes eld
  5003. \end_inset
  5004. axes
  5005. \begin_inset Quotes erd
  5006. \end_inset
  5007. of variation can be observed, each represented by 2 clusters with opposing
  5008. patterns.
  5009. The first axis is greater upstream coverage (Cluster 1) vs.
  5010. greater downstream coverage (Cluster 3); the second axis is the coverage
  5011. at the TSS itself: peak (Cluster 4) or trough (Cluster 2); lastly, the
  5012. third axis represents a trough upstream of the TSS (Cluster 5) vs.
  5013. downstream of the TSS (Cluster 6).
  5014. Referring to these opposing pairs of clusters as axes of variation is justified
  5015. , because they correspond precisely to the first 3 principal components
  5016. in the PCA plot of the relative coverage values (Figure
  5017. \begin_inset CommandInset ref
  5018. LatexCommand ref
  5019. reference "fig:H3K27me3-neighborhood-pca"
  5020. plural "false"
  5021. caps "false"
  5022. noprefix "false"
  5023. \end_inset
  5024. ).
  5025. The PCA plot reveals that as in the case of H3K4me2, all the
  5026. \begin_inset Quotes eld
  5027. \end_inset
  5028. clusters
  5029. \begin_inset Quotes erd
  5030. \end_inset
  5031. are really just sections of a single connected cloud rather than discrete
  5032. clusters.
  5033. The cloud is approximately ellipsoid-shaped, with each PC being an axis
  5034. of the ellipse, and each cluster consisting of a pyramidal section of the
  5035. ellipsoid.
  5036. \end_layout
  5037. \begin_layout Standard
  5038. In Figure
  5039. \begin_inset CommandInset ref
  5040. LatexCommand ref
  5041. reference "fig:H3K27me3-neighborhood-expression"
  5042. plural "false"
  5043. caps "false"
  5044. noprefix "false"
  5045. \end_inset
  5046. , we can see that Clusters 1 and 2 are the only clusters with higher gene
  5047. expression than the others.
  5048. For Cluster 2, this is expected, since this cluster represents genes with
  5049. depletion of H3K27me3 near the promoter.
  5050. Hence, elevated expression in cluster 2 is consistent with the conventional
  5051. view of H3K27me3 as a deactivating mark.
  5052. However, Cluster 1, the cluster with the most elevated gene expression,
  5053. represents genes with elevated coverage upstream of the TSS, or equivalently,
  5054. decreased coverage downstream, inside the gene body.
  5055. The opposite pattern, in which H3K27me3 is more abundant within the gene
  5056. body and less abundance in the upstream promoter region, does not show
  5057. any elevation in gene expression.
  5058. As with H3K4me2, this shows that the location of H3K27 trimethylation relative
  5059. to the TSS is potentially an important factor beyond simple proximity.
  5060. \end_layout
  5061. \begin_layout Standard
  5062. \begin_inset Flex TODO Note (inline)
  5063. status open
  5064. \begin_layout Plain Layout
  5065. Show the figures where the negative result ended this line of inquiry.
  5066. I need to debug some errors resulting from an R upgrade to do this.
  5067. \end_layout
  5068. \end_inset
  5069. \end_layout
  5070. \begin_layout Subsection
  5071. Defined pattern analysis
  5072. \end_layout
  5073. \begin_layout Standard
  5074. \begin_inset Flex TODO Note (inline)
  5075. status open
  5076. \begin_layout Plain Layout
  5077. This was where I defined interesting expression patterns and then looked
  5078. at initial relative promoter coverage for each expression pattern.
  5079. Negative result.
  5080. I forgot about this until recently.
  5081. Worth including? Remember to also write methods.
  5082. \end_layout
  5083. \end_inset
  5084. \end_layout
  5085. \begin_layout Subsection
  5086. Promoter CpG islands?
  5087. \end_layout
  5088. \begin_layout Standard
  5089. \begin_inset Flex TODO Note (inline)
  5090. status collapsed
  5091. \begin_layout Plain Layout
  5092. I forgot until recently about the work I did on this.
  5093. Worth including? Remember to also write methods.
  5094. \end_layout
  5095. \end_inset
  5096. \end_layout
  5097. \begin_layout Section
  5098. Discussion
  5099. \end_layout
  5100. \begin_layout Standard
  5101. \begin_inset Flex TODO Note (inline)
  5102. status open
  5103. \begin_layout Plain Layout
  5104. Write better section headers
  5105. \end_layout
  5106. \end_inset
  5107. \end_layout
  5108. \begin_layout Subsection
  5109. Effective promoter radius
  5110. \end_layout
  5111. \begin_layout Standard
  5112. Figure
  5113. \begin_inset CommandInset ref
  5114. LatexCommand ref
  5115. reference "fig:near-promoter-peak-enrich"
  5116. plural "false"
  5117. caps "false"
  5118. noprefix "false"
  5119. \end_inset
  5120. shows that H3K4me2, H3K4me3, and H3K27me3 are all enriched near promoters,
  5121. relative to the rest of the genome, consistent with their conventionally
  5122. understood role in regulating gene transcription.
  5123. Interestingly, the radius within this enrichment occurs is not the same
  5124. for each histone mark.
  5125. H3K4me2 and H3K4me3 are enriched within a 1
  5126. \begin_inset space \thinspace{}
  5127. \end_inset
  5128. kb radius, while H3K27me3 is enriched within 2.5
  5129. \begin_inset space \thinspace{}
  5130. \end_inset
  5131. kb.
  5132. Notably, the determined promoter radius was consistent across all experimental
  5133. conditions, varying only between different histone marks.
  5134. This suggests that the conventional
  5135. \begin_inset Quotes eld
  5136. \end_inset
  5137. one size fits all
  5138. \begin_inset Quotes erd
  5139. \end_inset
  5140. approach of defining a single promoter region for each gene (or each TSS)
  5141. and using that same promoter region for analyzing all types of genomic
  5142. data within an experiment may not be appropriate, and a better approach
  5143. may be to use a separate promoter radius for each kind of data, with each
  5144. radius being derived from the data itself.
  5145. Furthermore, the apparent asymmetry of upstream and downstream promoter
  5146. histone modification with respect to gene expression, seen in Figures
  5147. \begin_inset CommandInset ref
  5148. LatexCommand ref
  5149. reference "fig:H3K4me2-neighborhood"
  5150. plural "false"
  5151. caps "false"
  5152. noprefix "false"
  5153. \end_inset
  5154. ,
  5155. \begin_inset CommandInset ref
  5156. LatexCommand ref
  5157. reference "fig:H3K4me3-neighborhood"
  5158. plural "false"
  5159. caps "false"
  5160. noprefix "false"
  5161. \end_inset
  5162. , and
  5163. \begin_inset CommandInset ref
  5164. LatexCommand ref
  5165. reference "fig:H3K27me3-neighborhood"
  5166. plural "false"
  5167. caps "false"
  5168. noprefix "false"
  5169. \end_inset
  5170. , shows that even the concept of a promoter
  5171. \begin_inset Quotes eld
  5172. \end_inset
  5173. radius
  5174. \begin_inset Quotes erd
  5175. \end_inset
  5176. is likely an oversimplification.
  5177. At a minimum, nearby enrichment of peaks should be evaluated separately
  5178. for both upstream and downstream peaks, and an appropriate
  5179. \begin_inset Quotes eld
  5180. \end_inset
  5181. radius
  5182. \begin_inset Quotes erd
  5183. \end_inset
  5184. should be selected for each direction.
  5185. \end_layout
  5186. \begin_layout Standard
  5187. Figures
  5188. \begin_inset CommandInset ref
  5189. LatexCommand ref
  5190. reference "fig:H3K4me2-neighborhood"
  5191. plural "false"
  5192. caps "false"
  5193. noprefix "false"
  5194. \end_inset
  5195. and
  5196. \begin_inset CommandInset ref
  5197. LatexCommand ref
  5198. reference "fig:H3K4me3-neighborhood"
  5199. plural "false"
  5200. caps "false"
  5201. noprefix "false"
  5202. \end_inset
  5203. show that the determined promoter radius of 1
  5204. \begin_inset space ~
  5205. \end_inset
  5206. kb is approximately consistent with the distance from the TSS at which enrichmen
  5207. t of H3K4 methylation correlates with increased expression, showing that
  5208. this radius, which was determined by a simple analysis of measuring the
  5209. distance from each TSS to the nearest peak, also has functional significance.
  5210. For H3K27me3, the correlation between histone modification near the promoter
  5211. and gene expression is more complex, involving non-peak variations such
  5212. as troughs in coverage at the TSS and asymmetric coverage upstream and
  5213. downstream, so it is difficult in this case to evaluate whether the 2.5
  5214. \begin_inset space ~
  5215. \end_inset
  5216. kb radius determined from TSS-to-peak distances is functionally significant.
  5217. However, the two patterns of coverage associated with elevated expression
  5218. levels both have interesting features within this radius.
  5219. \end_layout
  5220. \begin_layout Standard
  5221. \begin_inset Flex TODO Note (inline)
  5222. status open
  5223. \begin_layout Plain Layout
  5224. My instinct is to say
  5225. \begin_inset Quotes eld
  5226. \end_inset
  5227. further study is needed
  5228. \begin_inset Quotes erd
  5229. \end_inset
  5230. here, but that goes in Chapter 5, right?
  5231. \end_layout
  5232. \end_inset
  5233. \end_layout
  5234. \begin_layout Subsection
  5235. Convergence
  5236. \end_layout
  5237. \begin_layout Standard
  5238. \begin_inset Flex TODO Note (inline)
  5239. status open
  5240. \begin_layout Plain Layout
  5241. Look up some more references for these histone marks being involved in memory
  5242. differentiation.
  5243. (Ask Sarah)
  5244. \end_layout
  5245. \end_inset
  5246. \end_layout
  5247. \begin_layout Standard
  5248. We have observed that all 3 histone marks and the gene expression data all
  5249. exhibit evidence of convergence in abundance between naive and memory cells
  5250. by day 14 after activation (Figure
  5251. \begin_inset CommandInset ref
  5252. LatexCommand ref
  5253. reference "fig:PCoA-promoters"
  5254. plural "false"
  5255. caps "false"
  5256. noprefix "false"
  5257. \end_inset
  5258. , Table
  5259. \begin_inset CommandInset ref
  5260. LatexCommand ref
  5261. reference "tab:Number-signif-promoters"
  5262. plural "false"
  5263. caps "false"
  5264. noprefix "false"
  5265. \end_inset
  5266. ).
  5267. The MOFA latent factor scatter plots (Figure
  5268. \begin_inset CommandInset ref
  5269. LatexCommand ref
  5270. reference "fig:mofa-lf-scatter"
  5271. plural "false"
  5272. caps "false"
  5273. noprefix "false"
  5274. \end_inset
  5275. ) show that this pattern of convergence is captured in latent factor 5.
  5276. Like all the latent factors in this plot, this factor explains a substantial
  5277. portion of the variance in all 4 data sets, indicating a coordinated pattern
  5278. of variation shared across all histone marks and gene expression.
  5279. This, of course, is consistent with the expectation that any naive CD4
  5280. T-cells remaining at day 14 should have differentiated into memory cells
  5281. by that time, and should therefore have a genomic state similar to memory
  5282. cells.
  5283. This convergence is evidence that these histone marks all play an important
  5284. role in the naive-to-memory differentiation process.
  5285. A histone mark that was not involved in naive-to-memory differentiation
  5286. would not be expected to converge in this way after activation.
  5287. \end_layout
  5288. \begin_layout Standard
  5289. \begin_inset Float figure
  5290. wide false
  5291. sideways false
  5292. status collapsed
  5293. \begin_layout Plain Layout
  5294. \align center
  5295. \begin_inset Graphics
  5296. filename graphics/CD4-csaw/LaMere2016_fig8.pdf
  5297. lyxscale 50
  5298. width 60col%
  5299. groupId colwidth
  5300. \end_inset
  5301. \end_layout
  5302. \begin_layout Plain Layout
  5303. \begin_inset Caption Standard
  5304. \begin_layout Plain Layout
  5305. \series bold
  5306. \begin_inset CommandInset label
  5307. LatexCommand label
  5308. name "fig:Lamere2016-Fig8"
  5309. \end_inset
  5310. Lamere 2016 Figure 8
  5311. \begin_inset CommandInset citation
  5312. LatexCommand cite
  5313. key "LaMere2016"
  5314. literal "false"
  5315. \end_inset
  5316. ,
  5317. \begin_inset Quotes eld
  5318. \end_inset
  5319. Model for the role of H3K4 methylation during CD4 T-cell activation.
  5320. \begin_inset Quotes erd
  5321. \end_inset
  5322. \series default
  5323. Reproduced with permission.
  5324. \end_layout
  5325. \end_inset
  5326. \end_layout
  5327. \end_inset
  5328. \end_layout
  5329. \begin_layout Standard
  5330. In H3K4me2, H3K4me3, and RNA-seq, this convergence appears to be in progress
  5331. already by Day 5, shown by the smaller distance between naive and memory
  5332. cells at day 5 along the
  5333. \begin_inset Formula $y$
  5334. \end_inset
  5335. -axes in Figures
  5336. \begin_inset CommandInset ref
  5337. LatexCommand ref
  5338. reference "fig:PCoA-H3K4me2-prom"
  5339. plural "false"
  5340. caps "false"
  5341. noprefix "false"
  5342. \end_inset
  5343. ,
  5344. \begin_inset CommandInset ref
  5345. LatexCommand ref
  5346. reference "fig:PCoA-H3K4me3-prom"
  5347. plural "false"
  5348. caps "false"
  5349. noprefix "false"
  5350. \end_inset
  5351. , and
  5352. \begin_inset CommandInset ref
  5353. LatexCommand ref
  5354. reference "fig:RNA-PCA-group"
  5355. plural "false"
  5356. caps "false"
  5357. noprefix "false"
  5358. \end_inset
  5359. .
  5360. This agrees with the model proposed by Sarah Lamere based on an prior analysis
  5361. of the same data, shown in Figure
  5362. \begin_inset CommandInset ref
  5363. LatexCommand ref
  5364. reference "fig:Lamere2016-Fig8"
  5365. plural "false"
  5366. caps "false"
  5367. noprefix "false"
  5368. \end_inset
  5369. , which shows the pattern of H3K4 methylation and expression for naive cells
  5370. and memory cells converging at day 5.
  5371. This model was developed without the benefit of the PCoA plots in Figure
  5372. \begin_inset CommandInset ref
  5373. LatexCommand ref
  5374. reference "fig:PCoA-promoters"
  5375. plural "false"
  5376. caps "false"
  5377. noprefix "false"
  5378. \end_inset
  5379. , which have been corrected for confounding factors by ComBat and SVA.
  5380. This shows that proper batch correction assists in extracting meaningful
  5381. patterns in the data while eliminating systematic sources of irrelevant
  5382. variation in the data, allowing simple automated procedures like PCoA to
  5383. reveal interesting behaviors in the data that were previously only detectable
  5384. by a detailed manual analysis.
  5385. \end_layout
  5386. \begin_layout Standard
  5387. While the ideal comparison to demonstrate this convergence would be naive
  5388. cells at day 14 to memory cells at day 0, this is not feasible in this
  5389. experimental system, since neither naive nor memory cells are able to fully
  5390. return to their pre-activation state, as shown by the lack of overlap between
  5391. days 0 and 14 for either naive or memory cells in Figure
  5392. \begin_inset CommandInset ref
  5393. LatexCommand ref
  5394. reference "fig:PCoA-promoters"
  5395. plural "false"
  5396. caps "false"
  5397. noprefix "false"
  5398. \end_inset
  5399. .
  5400. \end_layout
  5401. \begin_layout Subsection
  5402. Positional
  5403. \end_layout
  5404. \begin_layout Standard
  5405. When looking at patterns in the relative coverage of each histone mark near
  5406. the TSS of each gene, several interesting patterns were apparent.
  5407. For H3K4me2 and H3K4me3, the pattern was straightforward: the consistent
  5408. pattern across all promoters was a single peak a few kb wide, with the
  5409. main axis of variation being the position of this peak relative to the
  5410. TSS (Figures
  5411. \begin_inset CommandInset ref
  5412. LatexCommand ref
  5413. reference "fig:H3K4me2-neighborhood"
  5414. plural "false"
  5415. caps "false"
  5416. noprefix "false"
  5417. \end_inset
  5418. &
  5419. \begin_inset CommandInset ref
  5420. LatexCommand ref
  5421. reference "fig:H3K4me3-neighborhood"
  5422. plural "false"
  5423. caps "false"
  5424. noprefix "false"
  5425. \end_inset
  5426. ).
  5427. There were no obvious
  5428. \begin_inset Quotes eld
  5429. \end_inset
  5430. preferred
  5431. \begin_inset Quotes erd
  5432. \end_inset
  5433. positions, but rather a continuous distribution of relative positions ranging
  5434. all across the promoter region.
  5435. The association with gene expression was also straightforward: peaks closer
  5436. to the TSS were more strongly associated with elevated gene expression.
  5437. Coverage downstream of the TSS appears to be more strongly associated with
  5438. elevated expression than coverage the same distance upstream, indicating
  5439. that the
  5440. \begin_inset Quotes eld
  5441. \end_inset
  5442. effective promoter region
  5443. \begin_inset Quotes erd
  5444. \end_inset
  5445. for H3K4me2 and H3K4me3 may be centered downstream of the TSS.
  5446. \end_layout
  5447. \begin_layout Standard
  5448. The relative promoter coverage for H3K27me3 had a more complex pattern,
  5449. with two specific patterns of promoter coverage associated with elevated
  5450. expression: a sharp depletion of H3K27me3 around the TSS relative to the
  5451. surrounding area, and a depletion of H3K27me3 downstream of the TSS relative
  5452. to upstream (Figure
  5453. \begin_inset CommandInset ref
  5454. LatexCommand ref
  5455. reference "fig:H3K27me3-neighborhood"
  5456. plural "false"
  5457. caps "false"
  5458. noprefix "false"
  5459. \end_inset
  5460. ).
  5461. A previous study found that H3K27me3 depletion within the gene body was
  5462. associated with elevated gene expression in 4 different cell types in mice
  5463. \begin_inset CommandInset citation
  5464. LatexCommand cite
  5465. key "Young2011"
  5466. literal "false"
  5467. \end_inset
  5468. .
  5469. This is consistent with the second pattern described here.
  5470. This study also reported that a spike in coverage at the TSS was associated
  5471. with
  5472. \emph on
  5473. lower
  5474. \emph default
  5475. expression, which is indirectly consistent with the first pattern described
  5476. here, in the sense that it associates lower H3K27me3 levels near the TSS
  5477. with higher expression.
  5478. \end_layout
  5479. \begin_layout Subsection
  5480. Workflow
  5481. \end_layout
  5482. \begin_layout Standard
  5483. \begin_inset ERT
  5484. status open
  5485. \begin_layout Plain Layout
  5486. \backslash
  5487. afterpage{
  5488. \end_layout
  5489. \begin_layout Plain Layout
  5490. \backslash
  5491. begin{landscape}
  5492. \end_layout
  5493. \end_inset
  5494. \end_layout
  5495. \begin_layout Standard
  5496. \begin_inset Float figure
  5497. wide false
  5498. sideways false
  5499. status open
  5500. \begin_layout Plain Layout
  5501. \align center
  5502. \begin_inset Graphics
  5503. filename graphics/CD4-csaw/rulegraphs/rulegraph-all.pdf
  5504. lyxscale 50
  5505. width 100col%
  5506. height 95theight%
  5507. \end_inset
  5508. \end_layout
  5509. \begin_layout Plain Layout
  5510. \begin_inset Caption Standard
  5511. \begin_layout Plain Layout
  5512. \begin_inset CommandInset label
  5513. LatexCommand label
  5514. name "fig:rulegraph"
  5515. \end_inset
  5516. \series bold
  5517. Dependency graph of steps in reproducible workflow.
  5518. \end_layout
  5519. \end_inset
  5520. \end_layout
  5521. \end_inset
  5522. \end_layout
  5523. \begin_layout Standard
  5524. \begin_inset ERT
  5525. status open
  5526. \begin_layout Plain Layout
  5527. \backslash
  5528. end{landscape}
  5529. \end_layout
  5530. \begin_layout Plain Layout
  5531. }
  5532. \end_layout
  5533. \end_inset
  5534. \end_layout
  5535. \begin_layout Standard
  5536. The analyses described in this chapter were organized into a reproducible
  5537. workflow using the Snakemake workflow management system.
  5538. As shown in Figure
  5539. \begin_inset CommandInset ref
  5540. LatexCommand ref
  5541. reference "fig:rulegraph"
  5542. plural "false"
  5543. caps "false"
  5544. noprefix "false"
  5545. \end_inset
  5546. , the workflow includes many steps with complex dependencies between them.
  5547. For example, the step that counts the number of ChIP-seq reads in 500
  5548. \begin_inset space ~
  5549. \end_inset
  5550. bp windows in each promoter (the starting point for Figures
  5551. \begin_inset CommandInset ref
  5552. LatexCommand ref
  5553. reference "fig:H3K4me2-neighborhood"
  5554. plural "false"
  5555. caps "false"
  5556. noprefix "false"
  5557. \end_inset
  5558. ,
  5559. \begin_inset CommandInset ref
  5560. LatexCommand ref
  5561. reference "fig:H3K4me3-neighborhood"
  5562. plural "false"
  5563. caps "false"
  5564. noprefix "false"
  5565. \end_inset
  5566. , and
  5567. \begin_inset CommandInset ref
  5568. LatexCommand ref
  5569. reference "fig:H3K27me3-neighborhood"
  5570. plural "false"
  5571. caps "false"
  5572. noprefix "false"
  5573. \end_inset
  5574. ), named
  5575. \begin_inset Formula $\texttt{chipseq\_count\_tss\_neighborhoods}$
  5576. \end_inset
  5577. , depends on the RNA-seq abundance estimates in order to select the most-used
  5578. TSS for each gene, the aligned ChIP-seq reads, the index for those reads,
  5579. and the blacklist of regions to be excluded from ChIP-seq analysis.
  5580. Each step declares its inputs and outputs, and Snakemake uses these to
  5581. determine the dependencies between steps.
  5582. Each step is marked as depending on all the steps whose outputs match its
  5583. inputs, generating the workflow graph in Figure
  5584. \begin_inset CommandInset ref
  5585. LatexCommand ref
  5586. reference "fig:rulegraph"
  5587. plural "false"
  5588. caps "false"
  5589. noprefix "false"
  5590. \end_inset
  5591. , which Snakemake uses to determine order in which to execute each step
  5592. so that each step is executed only after all of the steps it depends on
  5593. have completed, thereby automating the entire workflow from start to finish.
  5594. \end_layout
  5595. \begin_layout Standard
  5596. In addition to simply making it easier to organize the steps in the analysis,
  5597. structuring the analysis as a workflow allowed for some analysis strategies
  5598. that would not have been practical otherwise.
  5599. For example, 5 different RNA-seq quantification methods were tested against
  5600. two different reference transcriptome annotations for a total of 10 different
  5601. quantifications of the same RNA-seq data.
  5602. These were then compared against each other in the exploratory data analysis
  5603. step, to determine that the results were not very sensitive to either the
  5604. choice of quantification method or the choice of annotation.
  5605. This was possible with a single script for the exploratory data analysis,
  5606. because Snakemake was able to automate running this script for every combinatio
  5607. n of method and reference.
  5608. In a similar manner, two different peak calling methods were tested against
  5609. each other, and in this case it was determined that SICER was unambiguously
  5610. superior to MACS for all histone marks studied.
  5611. By enabling these types of comparisons, structuring the analysis as an
  5612. automated workflow allowed important analysis decisions to be made in a
  5613. data-driven way, by running every reasonable option through the downstream
  5614. steps, seeing the consequences of choosing each option, and deciding accordingl
  5615. y.
  5616. \end_layout
  5617. \begin_layout Subsection
  5618. Data quality issues limit conclusions
  5619. \end_layout
  5620. \begin_layout Standard
  5621. \begin_inset Flex TODO Note (inline)
  5622. status open
  5623. \begin_layout Plain Layout
  5624. Is this needed?
  5625. \end_layout
  5626. \end_inset
  5627. \end_layout
  5628. \begin_layout Section
  5629. Future Directions
  5630. \end_layout
  5631. \begin_layout Standard
  5632. The analysis of RNA-seq and ChIP-seq in CD4 T-cells in Chapter 2 is in many
  5633. ways a preliminary study that suggests a multitude of new avenues of investigat
  5634. ion.
  5635. Here we consider a selection of such avenues.
  5636. \end_layout
  5637. \begin_layout Subsection
  5638. Negative results
  5639. \end_layout
  5640. \begin_layout Standard
  5641. Two additional analyses were conducted beyond those reported in the results.
  5642. First, we searched for evidence that the presence or absence of a CpG island
  5643. in the promoter was correlated with increases or decreases in gene expression
  5644. or any histone mark in any of the tested contrasts.
  5645. Second, we searched for evidence that the relative ChIP-seq coverage profiles
  5646. prior to activations could predict the change in expression of a gene after
  5647. activation.
  5648. Neither analysis turned up any clear positive results.
  5649. \end_layout
  5650. \begin_layout Subsection
  5651. Improve on the idea of an effective promoter radius
  5652. \end_layout
  5653. \begin_layout Standard
  5654. This study introduced the concept of an
  5655. \begin_inset Quotes eld
  5656. \end_inset
  5657. effective promoter radius
  5658. \begin_inset Quotes erd
  5659. \end_inset
  5660. specific to each histone mark based on distance from the TSS within which
  5661. an excess of peaks was called for that mark.
  5662. This concept was then used to guide further analyses throughout the study.
  5663. However, while the effective promoter radius was useful in those analyses,
  5664. it is both limited in theory and shown in practice to be a possible oversimplif
  5665. ication.
  5666. First, the effective promoter radii used in this study were chosen based
  5667. on manual inspection of the TSS-to-peak distance distributions in Figure
  5668. \begin_inset CommandInset ref
  5669. LatexCommand ref
  5670. reference "fig:near-promoter-peak-enrich"
  5671. plural "false"
  5672. caps "false"
  5673. noprefix "false"
  5674. \end_inset
  5675. , selecting round numbers of analyst convenience (Table
  5676. \begin_inset CommandInset ref
  5677. LatexCommand ref
  5678. reference "tab:effective-promoter-radius"
  5679. plural "false"
  5680. caps "false"
  5681. noprefix "false"
  5682. \end_inset
  5683. ).
  5684. It would be better to define an algorithm that selects a more precise radius
  5685. based on the features of the graph.
  5686. One possible way to do this would be to randomly rearrange the called peaks
  5687. throughout the genome many (while preserving the distribution of peak widths)
  5688. and re-generate the same plot as in Figure
  5689. \begin_inset CommandInset ref
  5690. LatexCommand ref
  5691. reference "fig:near-promoter-peak-enrich"
  5692. plural "false"
  5693. caps "false"
  5694. noprefix "false"
  5695. \end_inset
  5696. .
  5697. This would yield a better
  5698. \begin_inset Quotes eld
  5699. \end_inset
  5700. background
  5701. \begin_inset Quotes erd
  5702. \end_inset
  5703. distribution that demonstrates the degree of near-TSS enrichment that would
  5704. be expected by random chance.
  5705. The effective promoter radius could be defined as the point where the true
  5706. distribution diverges from the randomized background distribution.
  5707. \end_layout
  5708. \begin_layout Standard
  5709. Furthermore, the above definition of effective promoter radius has the significa
  5710. nt limitation of being based on the peak calling method.
  5711. It is thus very sensitive to the choice of peak caller and significance
  5712. threshold for calling peaks, as well as the degree of saturation in the
  5713. sequencing.
  5714. Calling peaks from ChIP-seq samples with insufficient coverage depth, with
  5715. the wrong peak caller, or with a different significance threshold could
  5716. give a drastically different number of called peaks, and hence a drastically
  5717. different distribution of peak-to-TSS distances.
  5718. To address this, it is desirable to develop a better method of determining
  5719. the effective promoter radius that relies only on the distribution of read
  5720. coverage around the TSS, independent of the peak calling.
  5721. Furthermore, as demonstrated by the upstream-downstream asymmetries observed
  5722. in Figures
  5723. \begin_inset CommandInset ref
  5724. LatexCommand ref
  5725. reference "fig:H3K4me2-neighborhood"
  5726. plural "false"
  5727. caps "false"
  5728. noprefix "false"
  5729. \end_inset
  5730. ,
  5731. \begin_inset CommandInset ref
  5732. LatexCommand ref
  5733. reference "fig:H3K4me3-neighborhood"
  5734. plural "false"
  5735. caps "false"
  5736. noprefix "false"
  5737. \end_inset
  5738. , and
  5739. \begin_inset CommandInset ref
  5740. LatexCommand ref
  5741. reference "fig:H3K27me3-neighborhood"
  5742. plural "false"
  5743. caps "false"
  5744. noprefix "false"
  5745. \end_inset
  5746. , this definition should determine a different radius for the upstream and
  5747. downstream directions.
  5748. At this point, it may be better to rename this concept
  5749. \begin_inset Quotes eld
  5750. \end_inset
  5751. effective promoter extent
  5752. \begin_inset Quotes erd
  5753. \end_inset
  5754. and avoid the word
  5755. \begin_inset Quotes eld
  5756. \end_inset
  5757. radius
  5758. \begin_inset Quotes erd
  5759. \end_inset
  5760. , since a radius implies a symmetry about the TSS that is not supported
  5761. by the data.
  5762. \end_layout
  5763. \begin_layout Standard
  5764. Beyond improving the definition of effective promoter extent, functional
  5765. validation is necessary to show that this measure of near-TSS enrichment
  5766. has biological meaning.
  5767. Figures
  5768. \begin_inset CommandInset ref
  5769. LatexCommand ref
  5770. reference "fig:H3K4me2-neighborhood"
  5771. plural "false"
  5772. caps "false"
  5773. noprefix "false"
  5774. \end_inset
  5775. and
  5776. \begin_inset CommandInset ref
  5777. LatexCommand ref
  5778. reference "fig:H3K4me3-neighborhood"
  5779. plural "false"
  5780. caps "false"
  5781. noprefix "false"
  5782. \end_inset
  5783. already provide a very limited functional validation of the chosen promoter
  5784. extents for H3K4me2 and H3K4me3 by showing that spikes in coverage within
  5785. this region are most strongly correlated with elevated gene expression.
  5786. However, there are other ways to show functional relevance of the promoter
  5787. extent.
  5788. For example, correlations could be computed between read counts in peaks
  5789. nearby gene promoters and the expression level of those genes, and these
  5790. correlations could be plotted against the distance of the peak upstream
  5791. or downstream of the gene's TSS.
  5792. If the promoter extent truly defines a
  5793. \begin_inset Quotes eld
  5794. \end_inset
  5795. sphere of influence
  5796. \begin_inset Quotes erd
  5797. \end_inset
  5798. within which a histone mark is involved with the regulation of a gene,
  5799. then the correlations for peaks within this extent should be significantly
  5800. higher than those further upstream or downstream.
  5801. Peaks within these extents may also be more likely to show differential
  5802. modification than those outside genic regions of the genome.
  5803. \end_layout
  5804. \begin_layout Subsection
  5805. Design experiments to focus on post-activation convergence of naive & memory
  5806. cells
  5807. \end_layout
  5808. \begin_layout Standard
  5809. In this study, a convergence between naive and memory cells was observed
  5810. in both the pattern of gene expression and in epigenetic state of the 3
  5811. histone marks studied, consistent with the hypothesis that any naive cells
  5812. remaining 14 days after activation have differentiated into memory cells,
  5813. and that both gene expression and these histone marks are involved in this
  5814. differentiation.
  5815. However, the current study was not designed with this specific hypothesis
  5816. in mind, and it therefore has some deficiencies with regard to testing
  5817. it.
  5818. The memory CD4 samples at day 14 do not resemble the memory samples at
  5819. day 0, indicating that in the specific model of activation used for this
  5820. experiment, the cells are not guaranteed to return to their original pre-activa
  5821. tion state, or perhaps this process takes substantially longer than 14 days.
  5822. This is a challenge for the convergence hypothesis because the ideal comparison
  5823. to prove that naive cells are converging to a resting memory state would
  5824. be to compare the final naive time point to the Day 0 memory samples, but
  5825. this comparison is only meaningful if memory cells generally return to
  5826. the same
  5827. \begin_inset Quotes eld
  5828. \end_inset
  5829. resting
  5830. \begin_inset Quotes erd
  5831. \end_inset
  5832. state that they started at.
  5833. \end_layout
  5834. \begin_layout Standard
  5835. To better study the convergence hypothesis, a new experiment should be designed
  5836. using a model system for T-cell activation that is known to allow cells
  5837. to return as closely as possible to their pre-activation state.
  5838. Alternatively, if it is not possible to find or design such a model system,
  5839. the same cell cultures could be activated serially multiple times, and
  5840. sequenced after each activation cycle right before the next activation.
  5841. It is likely that several activations in the same model system will settle
  5842. into a cyclical pattern, converging to a consistent
  5843. \begin_inset Quotes eld
  5844. \end_inset
  5845. resting
  5846. \begin_inset Quotes erd
  5847. \end_inset
  5848. state after each activation, even if this state is different from the initial
  5849. resting state at Day 0.
  5850. If so, it will be possible to compare the final states of both naive and
  5851. memory cells to show that they converge despite different initial conditions.
  5852. \end_layout
  5853. \begin_layout Standard
  5854. In addition, if naive-to-memory convergence is a general pattern, it should
  5855. also be detectable in other epigenetic marks, including other histone marks
  5856. and DNA methylation.
  5857. An experiment should be designed studying a large number of epigenetic
  5858. marks known or suspected to be involved in regulation of gene expression,
  5859. assaying all of these at the same pre- and post-activation time points.
  5860. Multi-dataset factor analysis methods like MOFA can then be used to identify
  5861. coordinated patterns of regulation shared across many epigenetic marks.
  5862. If possible, some
  5863. \begin_inset Quotes eld
  5864. \end_inset
  5865. negative control
  5866. \begin_inset Quotes erd
  5867. \end_inset
  5868. marks should be included that are known
  5869. \emph on
  5870. not
  5871. \emph default
  5872. to be involved in T-cell activation or memory formation.
  5873. Of course, CD4 T-cells are not the only adaptive immune cells with memory.
  5874. A similar study could be designed for CD8 T-cells, B-cells, and even specific
  5875. subsets of CD4 T-cells.
  5876. \end_layout
  5877. \begin_layout Subsection
  5878. Follow up on hints of interesting patterns in promoter relative coverage
  5879. profiles
  5880. \end_layout
  5881. \begin_layout Standard
  5882. \begin_inset Flex TODO Note (inline)
  5883. status open
  5884. \begin_layout Plain Layout
  5885. I think I might need to write up the negative results for the Promoter CpG
  5886. and defined pattern analysis before writing this section.
  5887. \end_layout
  5888. \end_inset
  5889. \end_layout
  5890. \begin_layout Itemize
  5891. Also find better normalizations: maybe borrow from MACS/SICER background
  5892. correction methods?
  5893. \end_layout
  5894. \begin_layout Itemize
  5895. For H3K4, define polar coordinates based on PC1 & 2: R = peak size, Theta
  5896. = peak position.
  5897. Then correlate with expression.
  5898. \end_layout
  5899. \begin_layout Itemize
  5900. Current analysis only at Day 0.
  5901. Need to study across time points.
  5902. \end_layout
  5903. \begin_layout Itemize
  5904. Integrating data across so many dimensions is a significant analysis challenge
  5905. \end_layout
  5906. \begin_layout Subsection
  5907. Investigate causes of high correlation between mutually exclusive histone
  5908. marks
  5909. \end_layout
  5910. \begin_layout Standard
  5911. The high correlation between coverage depth observed between H3K4me2 and
  5912. H3K4me3 is both expected and unexpected.
  5913. Since both marks are associated with elevated gene transcription, a positive
  5914. correlation between them is not surprising.
  5915. However, these two marks represent different post-translational modifications
  5916. of the
  5917. \emph on
  5918. same
  5919. \emph default
  5920. lysine residue on the histone H3 polypeptide, which means that they cannot
  5921. both be present on the same H3 subunit.
  5922. Thus, the high correlation between them has several potential explanations.
  5923. One possible reason is cell population heterogeneity: perhaps some genomic
  5924. loci are frequently marked with H3K4me2 in some cells, while in other cells
  5925. the same loci are marked with H3K4me3.
  5926. Another possibility is allele-specific modifications: the loci are marked
  5927. in each diploid cell with H3K4me2 on one allele and H3K4me3 on the other
  5928. allele.
  5929. Lastly, since each histone octamer contains 2 H3 subunits, it is possible
  5930. that having one H3K4me2 mark and one H3K4me3 mark on a given histone octamer
  5931. represents a distinct epigenetic state with a different function than either
  5932. double H3K4me2 or double H3K4me3.
  5933. \end_layout
  5934. \begin_layout Standard
  5935. These three hypotheses could be disentangled by single-cell ChIP-seq.
  5936. If the correlation between these two histone marks persists even within
  5937. the reads for each individual cell, then cell population heterogeneity
  5938. cannot explain the correlation.
  5939. Allele-specific modification can be tested for by looking at the correlation
  5940. between read coverage of the two histone marks at heterozygous loci.
  5941. If the correlation between read counts for opposite loci is low, then this
  5942. is consistent with allele-specific modification.
  5943. Finally if the modifications do not separate by either cell or allele,
  5944. the colocation of these two marks is most likely occurring at the level
  5945. of individual histones, with the heterogeneously modified histone representing
  5946. a distinct state.
  5947. \end_layout
  5948. \begin_layout Standard
  5949. However, another experiment would be required to show direct evidence of
  5950. such a heterogeneously modified state.
  5951. Specifically a
  5952. \begin_inset Quotes eld
  5953. \end_inset
  5954. double ChIP
  5955. \begin_inset Quotes erd
  5956. \end_inset
  5957. experiment would need to be performed, where the input DNA is first subjected
  5958. to an immunoprecipitation pulldown from the anti-H3K4me2 antibody, and
  5959. then the enriched material is collected, with proteins still bound, and
  5960. immunoprecipitated
  5961. \emph on
  5962. again
  5963. \emph default
  5964. using the anti-H3K4me3 antibody.
  5965. If this yields significant numbers of non-artifactual reads in the same
  5966. regions as the individual pulldowns of the two marks, this is strong evidence
  5967. that the two marks are occurring on opposite H3 subunits of the same histones.
  5968. \end_layout
  5969. \begin_layout Standard
  5970. \begin_inset Flex TODO Note (inline)
  5971. status open
  5972. \begin_layout Plain Layout
  5973. Try to see if double ChIP-seq is actually feasible, and if not, come up
  5974. with some other idea for directly detecting the mixed mod state.
  5975. Oh! Actually ChIP-seq isn't required, only double ChIP followed by quantificati
  5976. on.
  5977. That's one possible angle.
  5978. \end_layout
  5979. \end_inset
  5980. \end_layout
  5981. \begin_layout Chapter
  5982. Improving array-based diagnostics for transplant rejection by optimizing
  5983. data preprocessing
  5984. \end_layout
  5985. \begin_layout Standard
  5986. \begin_inset Note Note
  5987. status open
  5988. \begin_layout Plain Layout
  5989. Chapter author list: Me, Sunil, Tom, Padma, Dan
  5990. \end_layout
  5991. \end_inset
  5992. \end_layout
  5993. \begin_layout Section
  5994. Approach
  5995. \end_layout
  5996. \begin_layout Subsection
  5997. Proper pre-processing is essential for array data
  5998. \end_layout
  5999. \begin_layout Standard
  6000. \begin_inset Flex TODO Note (inline)
  6001. status open
  6002. \begin_layout Plain Layout
  6003. This section could probably use some citations
  6004. \end_layout
  6005. \end_inset
  6006. \end_layout
  6007. \begin_layout Standard
  6008. Microarrays, bead arrays, and similar assays produce raw data in the form
  6009. of fluorescence intensity measurements, with the each intensity measurement
  6010. proportional to the abundance of some fluorescently labelled target DNA
  6011. or RNA sequence that base pairs to a specific probe sequence.
  6012. However, these measurements for each probe are also affected my many technical
  6013. confounding factors, such as the concentration of target material, strength
  6014. of off-target binding, and the sensitivity of the imaging sensor.
  6015. Some array designs also use multiple probe sequences for each target.
  6016. Hence, extensive pre-processing of array data is necessary to normalize
  6017. out the effects of these technical factors and summarize the information
  6018. from multiple probes to arrive at a single usable estimate of abundance
  6019. or other relevant quantity, such as a ratio of two abundances, for each
  6020. target.
  6021. \end_layout
  6022. \begin_layout Standard
  6023. The choice of pre-processing algorithms used in the analysis of an array
  6024. data set can have a large effect on the results of that analysis.
  6025. However, despite their importance, these steps are often neglected or rushed
  6026. in order to get to the more scientifically interesting analysis steps involving
  6027. the actual biology of the system under study.
  6028. Hence, it is often possible to achieve substantial gains in statistical
  6029. power, model goodness-of-fit, or other relevant performance measures, by
  6030. checking the assumptions made by each preprocessing step and choosing specific
  6031. normalization methods tailored to the specific goals of the current analysis.
  6032. \end_layout
  6033. \begin_layout Subsection
  6034. Clinical diagnostic applications for microarrays require single-channel
  6035. normalization
  6036. \end_layout
  6037. \begin_layout Standard
  6038. As the cost of performing microarray assays falls, there is increasing interest
  6039. in using genomic assays for diagnostic purposes, such as distinguishing
  6040. healthy transplants (TX) from transplants undergoing acute rejection (AR)
  6041. or acute dysfunction with no rejection (ADNR).
  6042. However, the the standard normalization algorithm used for microarray data,
  6043. Robust Multi-chip Average (RMA)
  6044. \begin_inset CommandInset citation
  6045. LatexCommand cite
  6046. key "Irizarry2003a"
  6047. literal "false"
  6048. \end_inset
  6049. , is not applicable in a clinical setting.
  6050. Two of the steps in RMA, quantile normalization and probe summarization
  6051. by median polish, depend on every array in the data set being normalized.
  6052. This means that adding or removing any arrays from a data set changes the
  6053. normalized values for all arrays, and data sets that have been normalized
  6054. separately cannot be compared to each other.
  6055. Hence, when using RMA, any arrays to be analyzed together must also be
  6056. normalized together, and the set of arrays included in the data set must
  6057. be held constant throughout an analysis.
  6058. \end_layout
  6059. \begin_layout Standard
  6060. These limitations present serious impediments to the use of arrays as a
  6061. diagnostic tool.
  6062. When training a classifier, the samples to be classified must not be involved
  6063. in any step of the training process, lest their inclusion bias the training
  6064. process.
  6065. Once a classifier is deployed in a clinical setting, the samples to be
  6066. classified will not even
  6067. \emph on
  6068. exist
  6069. \emph default
  6070. at the time of training, so including them would be impossible even if
  6071. it were statistically justifiable.
  6072. Therefore, any machine learning application for microarrays demands that
  6073. the normalized expression values computed for an array must depend only
  6074. on information contained within that array.
  6075. This would ensure that each array's normalization is independent of every
  6076. other array, and that arrays normalized separately can still be compared
  6077. to each other without bias.
  6078. Such a normalization is commonly referred to as
  6079. \begin_inset Quotes eld
  6080. \end_inset
  6081. single-channel normalization
  6082. \begin_inset Quotes erd
  6083. \end_inset
  6084. .
  6085. \end_layout
  6086. \begin_layout Standard
  6087. Frozen RMA (fRMA) addresses these concerns by replacing the quantile normalizati
  6088. on and median polish with alternatives that do not introduce inter-array
  6089. dependence, allowing each array to be normalized independently of all others
  6090. \begin_inset CommandInset citation
  6091. LatexCommand cite
  6092. key "McCall2010"
  6093. literal "false"
  6094. \end_inset
  6095. .
  6096. Quantile normalization is performed against a pre-generated set of quantiles
  6097. learned from a collection of 850 publicly available arrays sampled from
  6098. a wide variety of tissues in the Gene Expression Omnibus (GEO).
  6099. Each array's probe intensity distribution is normalized against these pre-gener
  6100. ated quantiles.
  6101. The median polish step is replaced with a robust weighted average of probe
  6102. intensities, using inverse variance weights learned from the same public
  6103. GEO data.
  6104. The result is a normalization that satisfies the requirements mentioned
  6105. above: each array is normalized independently of all others, and any two
  6106. normalized arrays can be compared directly to each other.
  6107. \end_layout
  6108. \begin_layout Standard
  6109. One important limitation of fRMA is that it requires a separate reference
  6110. data set from which to learn the parameters (reference quantiles and probe
  6111. weights) that will be used to normalize each array.
  6112. These parameters are specific to a given array platform, and pre-generated
  6113. parameters are only provided for the most common platforms, such as Affymetrix
  6114. hgu133plus2.
  6115. For a less common platform, such as hthgu133pluspm, is is necessary to
  6116. learn custom parameters from in-house data before fRMA can be used to normalize
  6117. samples on that platform
  6118. \begin_inset CommandInset citation
  6119. LatexCommand cite
  6120. key "McCall2011"
  6121. literal "false"
  6122. \end_inset
  6123. .
  6124. \end_layout
  6125. \begin_layout Standard
  6126. One other option is the aptly-named Single Channel Array Normalization (SCAN),
  6127. which adapts a normalization method originally designed for tiling arrays
  6128. \begin_inset CommandInset citation
  6129. LatexCommand cite
  6130. key "Piccolo2012"
  6131. literal "false"
  6132. \end_inset
  6133. .
  6134. SCAN is truly single-channel in that it does not require a set of normalization
  6135. parameters estimated from an external set of reference samples like fRMA
  6136. does.
  6137. \end_layout
  6138. \begin_layout Subsection
  6139. Heteroskedasticity must be accounted for in methylation array data
  6140. \end_layout
  6141. \begin_layout Standard
  6142. DNA methylation arrays are a relatively new kind of assay that uses microarrays
  6143. to measure the degree of methylation on cytosines in specific regions arrayed
  6144. across the genome.
  6145. First, bisulfite treatment converts all unmethylated cytosines to uracil
  6146. (which then become thymine after amplification) while leaving methylated
  6147. cytosines unaffected.
  6148. Then, each target region is interrogated with two probes: one binds to
  6149. the original genomic sequence and interrogates the level of methylated
  6150. DNA, and the other binds to the same sequence with all cytosines replaced
  6151. by thymidines and interrogates the level of unmethylated DNA.
  6152. \end_layout
  6153. \begin_layout Standard
  6154. \begin_inset Float figure
  6155. wide false
  6156. sideways false
  6157. status collapsed
  6158. \begin_layout Plain Layout
  6159. \align center
  6160. \begin_inset Graphics
  6161. filename graphics/methylvoom/sigmoid.pdf
  6162. lyxscale 50
  6163. width 60col%
  6164. groupId colwidth
  6165. \end_inset
  6166. \end_layout
  6167. \begin_layout Plain Layout
  6168. \begin_inset Caption Standard
  6169. \begin_layout Plain Layout
  6170. \begin_inset CommandInset label
  6171. LatexCommand label
  6172. name "fig:Sigmoid-beta-m-mapping"
  6173. \end_inset
  6174. \series bold
  6175. Sigmoid shape of the mapping between β and M values
  6176. \end_layout
  6177. \end_inset
  6178. \end_layout
  6179. \end_inset
  6180. \end_layout
  6181. \begin_layout Standard
  6182. After normalization, these two probe intensities are summarized in one of
  6183. two ways, each with advantages and disadvantages.
  6184. β
  6185. \series bold
  6186. \series default
  6187. values, interpreted as fraction of DNA copies methylated, range from 0 to
  6188. 1.
  6189. β
  6190. \series bold
  6191. \series default
  6192. values are conceptually easy to interpret, but the constrained range makes
  6193. them unsuitable for linear modeling, and their error distributions are
  6194. highly non-normal, which also frustrates linear modeling.
  6195. M-values, interpreted as the log ratio of methylated to unmethylated copies,
  6196. are computed by mapping the beta values from
  6197. \begin_inset Formula $[0,1]$
  6198. \end_inset
  6199. onto
  6200. \begin_inset Formula $(-\infty,+\infty)$
  6201. \end_inset
  6202. using a sigmoid curve (Figure
  6203. \begin_inset CommandInset ref
  6204. LatexCommand ref
  6205. reference "fig:Sigmoid-beta-m-mapping"
  6206. plural "false"
  6207. caps "false"
  6208. noprefix "false"
  6209. \end_inset
  6210. ).
  6211. This transformation results in values with better statistical properties:
  6212. the unconstrained range is suitable for linear modeling, and the error
  6213. distributions are more normal.
  6214. Hence, most linear modeling and other statistical testing on methylation
  6215. arrays is performed using M-values.
  6216. \end_layout
  6217. \begin_layout Standard
  6218. However, the steep slope of the sigmoid transformation near 0 and 1 tends
  6219. to over-exaggerate small differences in β values near those extremes, which
  6220. in turn amplifies the error in those values, leading to a U-shaped trend
  6221. in the mean-variance curve: extreme values have higher variances than values
  6222. near the middle.
  6223. This mean-variance dependency must be accounted for when fitting the linear
  6224. model for differential methylation, or else the variance will be systematically
  6225. overestimated for probes with moderate M-values and underestimated for
  6226. probes with extreme M-values.
  6227. This is particularly undesirable for methylation data because the intermediate
  6228. M-values are the ones of most interest, since they are more likely to represent
  6229. areas of varying methylation, whereas extreme M-values typically represent
  6230. complete methylation or complete lack of methylation.
  6231. \end_layout
  6232. \begin_layout Standard
  6233. RNA-seq read count data are also known to show heteroskedasticity, and the
  6234. voom method was introduced for modeling this heteroskedasticity by estimating
  6235. the mean-variance trend in the data and using this trend to assign precision
  6236. weights to each observation
  6237. \begin_inset CommandInset citation
  6238. LatexCommand cite
  6239. key "Law2013"
  6240. literal "false"
  6241. \end_inset
  6242. .
  6243. While methylation array data are not derived from counts and have a very
  6244. different mean-variance relationship from that of typical RNA-seq data,
  6245. the voom method makes no specific assumptions on the shape of the mean-variance
  6246. relationship – it only assumes that the relationship can be modeled as
  6247. a smooth curve.
  6248. Hence, the method is sufficiently general to model the mean-variance relationsh
  6249. ip in methylation array data.
  6250. However, the standard implementation of voom assumes that the input is
  6251. given in raw read counts, and it must be adapted to run on methylation
  6252. M-values.
  6253. \end_layout
  6254. \begin_layout Section
  6255. Methods
  6256. \end_layout
  6257. \begin_layout Subsection
  6258. Evaluation of classifier performance with different normalization methods
  6259. \end_layout
  6260. \begin_layout Standard
  6261. For testing different expression microarray normalizations, a data set of
  6262. 157 hgu133plus2 arrays was used, consisting of blood samples from kidney
  6263. transplant patients whose grafts had been graded as TX, AR, or ADNR via
  6264. biopsy and histology (46 TX, 69 AR, 42 ADNR)
  6265. \begin_inset CommandInset citation
  6266. LatexCommand cite
  6267. key "Kurian2014"
  6268. literal "true"
  6269. \end_inset
  6270. .
  6271. Additionally, an external validation set of 75 samples was gathered from
  6272. public GEO data (37 TX, 38 AR, no ADNR).
  6273. \end_layout
  6274. \begin_layout Standard
  6275. \begin_inset Flex TODO Note (inline)
  6276. status open
  6277. \begin_layout Plain Layout
  6278. Find appropriate GEO identifiers if possible.
  6279. Kurian 2014 says GSE15296, but this seems to be different data.
  6280. I also need to look up the GEO accession for the external validation set.
  6281. \end_layout
  6282. \end_inset
  6283. \end_layout
  6284. \begin_layout Standard
  6285. To evaluate the effect of each normalization on classifier performance,
  6286. the same classifier training and validation procedure was used after each
  6287. normalization method.
  6288. The PAM package was used to train a nearest shrunken centroid classifier
  6289. on the training set and select the appropriate threshold for centroid shrinking.
  6290. Then the trained classifier was used to predict the class probabilities
  6291. of each validation sample.
  6292. From these class probabilities, ROC curves and area-under-curve (AUC) values
  6293. were generated
  6294. \begin_inset CommandInset citation
  6295. LatexCommand cite
  6296. key "Turck2011"
  6297. literal "false"
  6298. \end_inset
  6299. .
  6300. Each normalization was tested on two different sets of training and validation
  6301. samples.
  6302. For internal validation, the 115 TX and AR arrays in the internal set were
  6303. split at random into two equal sized sets, one for training and one for
  6304. validation, each containing the same numbers of TX and AR samples as the
  6305. other set.
  6306. For external validation, the full set of 115 TX and AR samples were used
  6307. as a training set, and the 75 external TX and AR samples were used as the
  6308. validation set.
  6309. Thus, 2 ROC curves and AUC values were generated for each normalization
  6310. method: one internal and one external.
  6311. Because the external validation set contains no ADNR samples, only classificati
  6312. on of TX and AR samples was considered.
  6313. The ADNR samples were included during normalization but excluded from all
  6314. classifier training and validation.
  6315. This ensures that the performance on internal and external validation sets
  6316. is directly comparable, since both are performing the same task: distinguishing
  6317. TX from AR.
  6318. \end_layout
  6319. \begin_layout Standard
  6320. \begin_inset Flex TODO Note (inline)
  6321. status open
  6322. \begin_layout Plain Layout
  6323. Summarize the get.best.threshold algorithm for PAM threshold selection, or
  6324. just put the code online?
  6325. \end_layout
  6326. \end_inset
  6327. \end_layout
  6328. \begin_layout Standard
  6329. Six different normalization strategies were evaluated.
  6330. First, 2 well-known non-single-channel normalization methods were considered:
  6331. RMA and dChip
  6332. \begin_inset CommandInset citation
  6333. LatexCommand cite
  6334. key "Li2001,Irizarry2003a"
  6335. literal "false"
  6336. \end_inset
  6337. .
  6338. Since RMA produces expression values on a log2 scale and dChip does not,
  6339. the values from dChip were log2 transformed after normalization.
  6340. Next, RMA and dChip followed by Global Rank-invariant Set Normalization
  6341. (GRSN) were tested
  6342. \begin_inset CommandInset citation
  6343. LatexCommand cite
  6344. key "Pelz2008"
  6345. literal "false"
  6346. \end_inset
  6347. .
  6348. Post-processing with GRSN does not turn RMA or dChip into single-channel
  6349. methods, but it may help mitigate batch effects and is therefore useful
  6350. as a benchmark.
  6351. Lastly, the two single-channel normalization methods, fRMA and SCAN, were
  6352. tested
  6353. \begin_inset CommandInset citation
  6354. LatexCommand cite
  6355. key "McCall2010,Piccolo2012"
  6356. literal "false"
  6357. \end_inset
  6358. .
  6359. When evaluating internal validation performance, only the 157 internal
  6360. samples were normalized; when evaluating external validation performance,
  6361. all 157 internal samples and 75 external samples were normalized together.
  6362. \end_layout
  6363. \begin_layout Standard
  6364. For demonstrating the problem with separate normalization of training and
  6365. validation data, one additional normalization was performed: the internal
  6366. and external sets were each normalized separately using RMA, and the normalized
  6367. data for each set were combined into a single set with no further attempts
  6368. at normalizing between the two sets.
  6369. The represents approximately how RMA would have to be used in a clinical
  6370. setting, where the samples to be classified are not available at the time
  6371. the classifier is trained.
  6372. \end_layout
  6373. \begin_layout Subsection
  6374. Generating custom fRMA vectors for hthgu133pluspm array platform
  6375. \end_layout
  6376. \begin_layout Standard
  6377. In order to enable fRMA normalization for the hthgu133pluspm array platform,
  6378. custom fRMA normalization vectors were trained using the
  6379. \begin_inset Flex Code
  6380. status open
  6381. \begin_layout Plain Layout
  6382. frmaTools
  6383. \end_layout
  6384. \end_inset
  6385. package
  6386. \begin_inset CommandInset citation
  6387. LatexCommand cite
  6388. key "McCall2011"
  6389. literal "false"
  6390. \end_inset
  6391. .
  6392. Separate vectors were created for two types of samples: kidney graft biopsy
  6393. samples and blood samples from graft recipients.
  6394. For training, a 341 kidney biopsy samples from 2 data sets and 965 blood
  6395. samples from 5 data sets were used as the reference set.
  6396. Arrays were groups into batches based on unique combinations of sample
  6397. type (blood or biopsy), diagnosis (TX, AR, etc.), data set, and scan date.
  6398. Thus, each batch represents arrays of the same kind that were run together
  6399. on the same day.
  6400. For estimating the probe inverse variance weights, frmaTools requires equal-siz
  6401. ed batches, which means a batch size must be chosen, and then batches smaller
  6402. than that size must be ignored, while batches larger than the chosen size
  6403. must be downsampled.
  6404. This downsampling is performed randomly, so the sampling process is repeated
  6405. 5 times and the resulting normalizations are compared to each other.
  6406. \end_layout
  6407. \begin_layout Standard
  6408. To evaluate the consistency of the generated normalization vectors, the
  6409. 5 fRMA vector sets generated from 5 random batch samplings were each used
  6410. to normalize the same 20 randomly selected samples from each tissue.
  6411. Then the normalized expression values for each probe on each array were
  6412. compared across all normalizations.
  6413. Each fRMA normalization was also compared against the normalized expression
  6414. values obtained by normalizing the same 20 samples with ordinary RMA.
  6415. \end_layout
  6416. \begin_layout Subsection
  6417. Modeling methylation array M-value heteroskedasticy in linear models with
  6418. modified voom implementation
  6419. \end_layout
  6420. \begin_layout Standard
  6421. \begin_inset Flex TODO Note (inline)
  6422. status open
  6423. \begin_layout Plain Layout
  6424. Put code on Github and reference it.
  6425. \end_layout
  6426. \end_inset
  6427. \end_layout
  6428. \begin_layout Standard
  6429. To investigate the whether DNA methylation could be used to distinguish
  6430. between healthy and dysfunctional transplants, a data set of 78 Illumina
  6431. 450k methylation arrays from human kidney graft biopsies was analyzed for
  6432. differential methylation between 4 transplant statuses: healthy transplant
  6433. (TX), transplants undergoing acute rejection (AR), acute dysfunction with
  6434. no rejection (ADNR), and chronic allograft nephropathy (CAN).
  6435. The data consisted of 33 TX, 9 AR, 8 ADNR, and 28 CAN samples.
  6436. The uneven group sizes are a result of taking the biopsy samples before
  6437. the eventual fate of the transplant was known.
  6438. Each sample was additionally annotated with a donor ID (anonymized), Sex,
  6439. Age, Ethnicity, Creatinine Level, and Diabetes diagnosis (all samples in
  6440. this data set came from patients with either Type 1 or Type 2 diabetes).
  6441. \end_layout
  6442. \begin_layout Standard
  6443. The intensity data were first normalized using subset-quantile within array
  6444. normalization (SWAN)
  6445. \begin_inset CommandInset citation
  6446. LatexCommand cite
  6447. key "Maksimovic2012"
  6448. literal "false"
  6449. \end_inset
  6450. , then converted to intensity ratios (beta values)
  6451. \begin_inset CommandInset citation
  6452. LatexCommand cite
  6453. key "Aryee2014"
  6454. literal "false"
  6455. \end_inset
  6456. .
  6457. Any probes binding to loci that overlapped annotated SNPs were dropped,
  6458. and the annotated sex of each sample was verified against the sex inferred
  6459. from the ratio of median probe intensities for the X and Y chromosomes.
  6460. Then, the ratios were transformed to M-values.
  6461. \end_layout
  6462. \begin_layout Standard
  6463. \begin_inset Float table
  6464. wide false
  6465. sideways false
  6466. status open
  6467. \begin_layout Plain Layout
  6468. \align center
  6469. \begin_inset Tabular
  6470. <lyxtabular version="3" rows="4" columns="6">
  6471. <features tabularvalignment="middle">
  6472. <column alignment="center" valignment="top">
  6473. <column alignment="center" valignment="top">
  6474. <column alignment="center" valignment="top">
  6475. <column alignment="center" valignment="top">
  6476. <column alignment="center" valignment="top">
  6477. <column alignment="center" valignment="top">
  6478. <row>
  6479. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  6480. \begin_inset Text
  6481. \begin_layout Plain Layout
  6482. Analysis
  6483. \end_layout
  6484. \end_inset
  6485. </cell>
  6486. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  6487. \begin_inset Text
  6488. \begin_layout Plain Layout
  6489. random effect
  6490. \end_layout
  6491. \end_inset
  6492. </cell>
  6493. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  6494. \begin_inset Text
  6495. \begin_layout Plain Layout
  6496. eBayes
  6497. \end_layout
  6498. \end_inset
  6499. </cell>
  6500. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  6501. \begin_inset Text
  6502. \begin_layout Plain Layout
  6503. SVA
  6504. \end_layout
  6505. \end_inset
  6506. </cell>
  6507. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  6508. \begin_inset Text
  6509. \begin_layout Plain Layout
  6510. weights
  6511. \end_layout
  6512. \end_inset
  6513. </cell>
  6514. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  6515. \begin_inset Text
  6516. \begin_layout Plain Layout
  6517. voom
  6518. \end_layout
  6519. \end_inset
  6520. </cell>
  6521. </row>
  6522. <row>
  6523. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  6524. \begin_inset Text
  6525. \begin_layout Plain Layout
  6526. A
  6527. \end_layout
  6528. \end_inset
  6529. </cell>
  6530. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  6531. \begin_inset Text
  6532. \begin_layout Plain Layout
  6533. Yes
  6534. \end_layout
  6535. \end_inset
  6536. </cell>
  6537. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  6538. \begin_inset Text
  6539. \begin_layout Plain Layout
  6540. Yes
  6541. \end_layout
  6542. \end_inset
  6543. </cell>
  6544. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  6545. \begin_inset Text
  6546. \begin_layout Plain Layout
  6547. No
  6548. \end_layout
  6549. \end_inset
  6550. </cell>
  6551. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  6552. \begin_inset Text
  6553. \begin_layout Plain Layout
  6554. No
  6555. \end_layout
  6556. \end_inset
  6557. </cell>
  6558. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  6559. \begin_inset Text
  6560. \begin_layout Plain Layout
  6561. No
  6562. \end_layout
  6563. \end_inset
  6564. </cell>
  6565. </row>
  6566. <row>
  6567. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  6568. \begin_inset Text
  6569. \begin_layout Plain Layout
  6570. B
  6571. \end_layout
  6572. \end_inset
  6573. </cell>
  6574. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  6575. \begin_inset Text
  6576. \begin_layout Plain Layout
  6577. Yes
  6578. \end_layout
  6579. \end_inset
  6580. </cell>
  6581. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  6582. \begin_inset Text
  6583. \begin_layout Plain Layout
  6584. Yes
  6585. \end_layout
  6586. \end_inset
  6587. </cell>
  6588. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  6589. \begin_inset Text
  6590. \begin_layout Plain Layout
  6591. Yes
  6592. \end_layout
  6593. \end_inset
  6594. </cell>
  6595. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  6596. \begin_inset Text
  6597. \begin_layout Plain Layout
  6598. Yes
  6599. \end_layout
  6600. \end_inset
  6601. </cell>
  6602. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  6603. \begin_inset Text
  6604. \begin_layout Plain Layout
  6605. No
  6606. \end_layout
  6607. \end_inset
  6608. </cell>
  6609. </row>
  6610. <row>
  6611. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  6612. \begin_inset Text
  6613. \begin_layout Plain Layout
  6614. C
  6615. \end_layout
  6616. \end_inset
  6617. </cell>
  6618. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  6619. \begin_inset Text
  6620. \begin_layout Plain Layout
  6621. Yes
  6622. \end_layout
  6623. \end_inset
  6624. </cell>
  6625. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  6626. \begin_inset Text
  6627. \begin_layout Plain Layout
  6628. Yes
  6629. \end_layout
  6630. \end_inset
  6631. </cell>
  6632. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  6633. \begin_inset Text
  6634. \begin_layout Plain Layout
  6635. Yes
  6636. \end_layout
  6637. \end_inset
  6638. </cell>
  6639. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  6640. \begin_inset Text
  6641. \begin_layout Plain Layout
  6642. Yes
  6643. \end_layout
  6644. \end_inset
  6645. </cell>
  6646. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  6647. \begin_inset Text
  6648. \begin_layout Plain Layout
  6649. Yes
  6650. \end_layout
  6651. \end_inset
  6652. </cell>
  6653. </row>
  6654. </lyxtabular>
  6655. \end_inset
  6656. \end_layout
  6657. \begin_layout Plain Layout
  6658. \begin_inset Caption Standard
  6659. \begin_layout Plain Layout
  6660. \series bold
  6661. \begin_inset CommandInset label
  6662. LatexCommand label
  6663. name "tab:Summary-of-meth-analysis"
  6664. \end_inset
  6665. Summary of analysis variants for methylation array data.
  6666. \series default
  6667. Each analysis included a different set of steps to adjust or account for
  6668. various systematic features of the data.
  6669. Random effect: The model included a random effect accounting for correlation
  6670. between samples from the same patient
  6671. \begin_inset CommandInset citation
  6672. LatexCommand cite
  6673. key "Smyth2005a"
  6674. literal "false"
  6675. \end_inset
  6676. ; eBayes: Empirical bayes squeezing of per-probe variances toward the mean-varia
  6677. nce trend
  6678. \begin_inset CommandInset citation
  6679. LatexCommand cite
  6680. key "Ritchie2015"
  6681. literal "false"
  6682. \end_inset
  6683. ; SVA: Surrogate variable analysis to account for unobserved confounders
  6684. \begin_inset CommandInset citation
  6685. LatexCommand cite
  6686. key "Leek2007"
  6687. literal "false"
  6688. \end_inset
  6689. ; Weights: Estimate sample weights to account for differences in sample
  6690. quality
  6691. \begin_inset CommandInset citation
  6692. LatexCommand cite
  6693. key "Liu2015,Ritchie2006"
  6694. literal "false"
  6695. \end_inset
  6696. ; voom: Use mean-variance trend to assign individual sample weights
  6697. \begin_inset CommandInset citation
  6698. LatexCommand cite
  6699. key "Law2013"
  6700. literal "false"
  6701. \end_inset
  6702. .
  6703. See the text for a more detailed explanation of each step.
  6704. \end_layout
  6705. \end_inset
  6706. \end_layout
  6707. \end_inset
  6708. \end_layout
  6709. \begin_layout Standard
  6710. From the M-values, a series of parallel analyses was performed, each adding
  6711. additional steps into the model fit to accommodate a feature of the data
  6712. (see Table
  6713. \begin_inset CommandInset ref
  6714. LatexCommand ref
  6715. reference "tab:Summary-of-meth-analysis"
  6716. plural "false"
  6717. caps "false"
  6718. noprefix "false"
  6719. \end_inset
  6720. ).
  6721. For analysis A, a
  6722. \begin_inset Quotes eld
  6723. \end_inset
  6724. basic
  6725. \begin_inset Quotes erd
  6726. \end_inset
  6727. linear modeling analysis was performed, compensating for known confounders
  6728. by including terms for the factor of interest (transplant status) as well
  6729. as the known biological confounders: sex, age, ethnicity, and diabetes.
  6730. Since some samples came from the same patients at different times, the
  6731. intra-patient correlation was modeled as a random effect, estimating a
  6732. shared correlation value across all probes
  6733. \begin_inset CommandInset citation
  6734. LatexCommand cite
  6735. key "Smyth2005a"
  6736. literal "false"
  6737. \end_inset
  6738. .
  6739. Then the linear model was fit, and the variance was modeled using empirical
  6740. Bayes squeezing toward the mean-variance trend
  6741. \begin_inset CommandInset citation
  6742. LatexCommand cite
  6743. key "Ritchie2015"
  6744. literal "false"
  6745. \end_inset
  6746. .
  6747. Finally, t-tests or F-tests were performed as appropriate for each test:
  6748. t-tests for single contrasts, and F-tests for multiple contrasts.
  6749. P-values were corrected for multiple testing using the Benjamini-Hochberg
  6750. procedure for FDR control
  6751. \begin_inset CommandInset citation
  6752. LatexCommand cite
  6753. key "Benjamini1995"
  6754. literal "false"
  6755. \end_inset
  6756. .
  6757. \end_layout
  6758. \begin_layout Standard
  6759. For the analysis B, surrogate variable analysis (SVA) was used to infer
  6760. additional unobserved sources of heterogeneity in the data
  6761. \begin_inset CommandInset citation
  6762. LatexCommand cite
  6763. key "Leek2007"
  6764. literal "false"
  6765. \end_inset
  6766. .
  6767. These surrogate variables were added to the design matrix before fitting
  6768. the linear model.
  6769. In addition, sample quality weights were estimated from the data and used
  6770. during linear modeling to down-weight the contribution of highly variable
  6771. arrays while increasing the weight to arrays with lower variability
  6772. \begin_inset CommandInset citation
  6773. LatexCommand cite
  6774. key "Ritchie2006"
  6775. literal "false"
  6776. \end_inset
  6777. .
  6778. The remainder of the analysis proceeded as in analysis A.
  6779. For analysis C, the voom method was adapted to run on methylation array
  6780. data and used to model and correct for the mean-variance trend using individual
  6781. observation weights
  6782. \begin_inset CommandInset citation
  6783. LatexCommand cite
  6784. key "Law2013"
  6785. literal "false"
  6786. \end_inset
  6787. , which were combined with the sample weights
  6788. \begin_inset CommandInset citation
  6789. LatexCommand cite
  6790. key "Liu2015,Ritchie2006"
  6791. literal "false"
  6792. \end_inset
  6793. .
  6794. Each time weights were used, they were estimated once before estimating
  6795. the random effect correlation value, and then the weights were re-estimated
  6796. taking the random effect into account.
  6797. The remainder of the analysis proceeded as in analysis B.
  6798. \end_layout
  6799. \begin_layout Section
  6800. Results
  6801. \end_layout
  6802. \begin_layout Standard
  6803. \begin_inset Flex TODO Note (inline)
  6804. status open
  6805. \begin_layout Plain Layout
  6806. Improve subsection titles in this section.
  6807. \end_layout
  6808. \end_inset
  6809. \end_layout
  6810. \begin_layout Standard
  6811. \begin_inset Flex TODO Note (inline)
  6812. status open
  6813. \begin_layout Plain Layout
  6814. Reconsider subsection organization?
  6815. \end_layout
  6816. \end_inset
  6817. \end_layout
  6818. \begin_layout Subsection
  6819. Separate normalization with RMA introduces unwanted biases in classification
  6820. \end_layout
  6821. \begin_layout Standard
  6822. \begin_inset Float figure
  6823. wide false
  6824. sideways false
  6825. status open
  6826. \begin_layout Plain Layout
  6827. \align center
  6828. \begin_inset Graphics
  6829. filename graphics/PAM/predplot.pdf
  6830. lyxscale 50
  6831. width 60col%
  6832. groupId colwidth
  6833. \end_inset
  6834. \end_layout
  6835. \begin_layout Plain Layout
  6836. \begin_inset Caption Standard
  6837. \begin_layout Plain Layout
  6838. \begin_inset CommandInset label
  6839. LatexCommand label
  6840. name "fig:Classifier-probabilities-RMA"
  6841. \end_inset
  6842. \series bold
  6843. Classifier probabilities on validation samples when normalized with RMA
  6844. together vs.
  6845. separately.
  6846. \series default
  6847. The PAM classifier algorithm was trained on the training set of arrays to
  6848. distinguish AR from TX and then used to assign class probabilities to the
  6849. validation set.
  6850. The process was performed after normalizing all samples together and after
  6851. normalizing the training and test sets separately, and the class probabilities
  6852. assigned to each sample in the validation set were plotted against each
  6853. other (PP(AR), posterior probability of being AR).
  6854. The color of each point indicates the true classification of that sample.
  6855. \end_layout
  6856. \end_inset
  6857. \end_layout
  6858. \end_inset
  6859. \end_layout
  6860. \begin_layout Standard
  6861. To demonstrate the problem with non-single-channel normalization methods,
  6862. we considered the problem of training a classifier to distinguish TX from
  6863. AR using the samples from the internal set as training data, evaluating
  6864. performance on the external set.
  6865. First, training and evaluation were performed after normalizing all array
  6866. samples together as a single set using RMA, and second, the internal samples
  6867. were normalized separately from the external samples and the training and
  6868. evaluation were repeated.
  6869. For each sample in the validation set, the classifier probabilities from
  6870. both classifiers were plotted against each other (Fig.
  6871. \begin_inset CommandInset ref
  6872. LatexCommand ref
  6873. reference "fig:Classifier-probabilities-RMA"
  6874. plural "false"
  6875. caps "false"
  6876. noprefix "false"
  6877. \end_inset
  6878. ).
  6879. As expected, separate normalization biases the classifier probabilities,
  6880. resulting in several misclassifications.
  6881. In this case, the bias from separate normalization causes the classifier
  6882. to assign a lower probability of AR to every sample.
  6883. \end_layout
  6884. \begin_layout Subsection
  6885. fRMA and SCAN maintain classification performance while eliminating dependence
  6886. on normalization strategy
  6887. \end_layout
  6888. \begin_layout Standard
  6889. \begin_inset Float figure
  6890. wide false
  6891. sideways false
  6892. status open
  6893. \begin_layout Plain Layout
  6894. \align center
  6895. \begin_inset Float figure
  6896. placement tb
  6897. wide false
  6898. sideways false
  6899. status open
  6900. \begin_layout Plain Layout
  6901. \align center
  6902. \begin_inset Graphics
  6903. filename graphics/PAM/ROC-TXvsAR-internal.pdf
  6904. lyxscale 50
  6905. height 40theight%
  6906. groupId roc-pam
  6907. \end_inset
  6908. \end_layout
  6909. \begin_layout Plain Layout
  6910. \begin_inset Caption Standard
  6911. \begin_layout Plain Layout
  6912. \begin_inset CommandInset label
  6913. LatexCommand label
  6914. name "fig:ROC-PAM-int"
  6915. \end_inset
  6916. ROC curves for PAM on internal validation data
  6917. \end_layout
  6918. \end_inset
  6919. \end_layout
  6920. \end_inset
  6921. \end_layout
  6922. \begin_layout Plain Layout
  6923. \align center
  6924. \begin_inset Float figure
  6925. placement tb
  6926. wide false
  6927. sideways false
  6928. status open
  6929. \begin_layout Plain Layout
  6930. \align center
  6931. \begin_inset Graphics
  6932. filename graphics/PAM/ROC-TXvsAR-external.pdf
  6933. lyxscale 50
  6934. height 40theight%
  6935. groupId roc-pam
  6936. \end_inset
  6937. \end_layout
  6938. \begin_layout Plain Layout
  6939. \begin_inset Caption Standard
  6940. \begin_layout Plain Layout
  6941. \begin_inset CommandInset label
  6942. LatexCommand label
  6943. name "fig:ROC-PAM-ext"
  6944. \end_inset
  6945. ROC curves for PAM on external validation data
  6946. \end_layout
  6947. \end_inset
  6948. \end_layout
  6949. \end_inset
  6950. \end_layout
  6951. \begin_layout Plain Layout
  6952. \begin_inset Caption Standard
  6953. \begin_layout Plain Layout
  6954. \series bold
  6955. \begin_inset CommandInset label
  6956. LatexCommand label
  6957. name "fig:ROC-PAM-main"
  6958. \end_inset
  6959. ROC curves for PAM using different normalization strategies.
  6960. \series default
  6961. ROC curves were generated for PAM classification of AR vs TX after 6 different
  6962. normalization strategies applied to the same data sets.
  6963. Only fRMA and SCAN are single-channel normalizations.
  6964. The other normalizations are for comparison.
  6965. \end_layout
  6966. \end_inset
  6967. \end_layout
  6968. \end_inset
  6969. \end_layout
  6970. \begin_layout Standard
  6971. \begin_inset Float table
  6972. wide false
  6973. sideways false
  6974. status open
  6975. \begin_layout Plain Layout
  6976. \align center
  6977. \begin_inset Tabular
  6978. <lyxtabular version="3" rows="7" columns="4">
  6979. <features tabularvalignment="middle">
  6980. <column alignment="center" valignment="top">
  6981. <column alignment="center" valignment="top">
  6982. <column alignment="center" valignment="top">
  6983. <column alignment="center" valignment="top">
  6984. <row>
  6985. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  6986. \begin_inset Text
  6987. \begin_layout Plain Layout
  6988. \family roman
  6989. \series medium
  6990. \shape up
  6991. \size normal
  6992. \emph off
  6993. \bar no
  6994. \strikeout off
  6995. \xout off
  6996. \uuline off
  6997. \uwave off
  6998. \noun off
  6999. \color none
  7000. Normalization
  7001. \end_layout
  7002. \end_inset
  7003. </cell>
  7004. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  7005. \begin_inset Text
  7006. \begin_layout Plain Layout
  7007. Single-channel?
  7008. \end_layout
  7009. \end_inset
  7010. </cell>
  7011. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  7012. \begin_inset Text
  7013. \begin_layout Plain Layout
  7014. \family roman
  7015. \series medium
  7016. \shape up
  7017. \size normal
  7018. \emph off
  7019. \bar no
  7020. \strikeout off
  7021. \xout off
  7022. \uuline off
  7023. \uwave off
  7024. \noun off
  7025. \color none
  7026. Internal Val.
  7027. AUC
  7028. \end_layout
  7029. \end_inset
  7030. </cell>
  7031. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  7032. \begin_inset Text
  7033. \begin_layout Plain Layout
  7034. External Val.
  7035. AUC
  7036. \end_layout
  7037. \end_inset
  7038. </cell>
  7039. </row>
  7040. <row>
  7041. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  7042. \begin_inset Text
  7043. \begin_layout Plain Layout
  7044. \family roman
  7045. \series medium
  7046. \shape up
  7047. \size normal
  7048. \emph off
  7049. \bar no
  7050. \strikeout off
  7051. \xout off
  7052. \uuline off
  7053. \uwave off
  7054. \noun off
  7055. \color none
  7056. RMA
  7057. \end_layout
  7058. \end_inset
  7059. </cell>
  7060. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  7061. \begin_inset Text
  7062. \begin_layout Plain Layout
  7063. No
  7064. \end_layout
  7065. \end_inset
  7066. </cell>
  7067. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  7068. \begin_inset Text
  7069. \begin_layout Plain Layout
  7070. \family roman
  7071. \series medium
  7072. \shape up
  7073. \size normal
  7074. \emph off
  7075. \bar no
  7076. \strikeout off
  7077. \xout off
  7078. \uuline off
  7079. \uwave off
  7080. \noun off
  7081. \color none
  7082. 0.852
  7083. \end_layout
  7084. \end_inset
  7085. </cell>
  7086. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  7087. \begin_inset Text
  7088. \begin_layout Plain Layout
  7089. \family roman
  7090. \series medium
  7091. \shape up
  7092. \size normal
  7093. \emph off
  7094. \bar no
  7095. \strikeout off
  7096. \xout off
  7097. \uuline off
  7098. \uwave off
  7099. \noun off
  7100. \color none
  7101. 0.713
  7102. \end_layout
  7103. \end_inset
  7104. </cell>
  7105. </row>
  7106. <row>
  7107. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  7108. \begin_inset Text
  7109. \begin_layout Plain Layout
  7110. \family roman
  7111. \series medium
  7112. \shape up
  7113. \size normal
  7114. \emph off
  7115. \bar no
  7116. \strikeout off
  7117. \xout off
  7118. \uuline off
  7119. \uwave off
  7120. \noun off
  7121. \color none
  7122. dChip
  7123. \end_layout
  7124. \end_inset
  7125. </cell>
  7126. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  7127. \begin_inset Text
  7128. \begin_layout Plain Layout
  7129. No
  7130. \end_layout
  7131. \end_inset
  7132. </cell>
  7133. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  7134. \begin_inset Text
  7135. \begin_layout Plain Layout
  7136. \family roman
  7137. \series medium
  7138. \shape up
  7139. \size normal
  7140. \emph off
  7141. \bar no
  7142. \strikeout off
  7143. \xout off
  7144. \uuline off
  7145. \uwave off
  7146. \noun off
  7147. \color none
  7148. 0.891
  7149. \end_layout
  7150. \end_inset
  7151. </cell>
  7152. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  7153. \begin_inset Text
  7154. \begin_layout Plain Layout
  7155. \family roman
  7156. \series medium
  7157. \shape up
  7158. \size normal
  7159. \emph off
  7160. \bar no
  7161. \strikeout off
  7162. \xout off
  7163. \uuline off
  7164. \uwave off
  7165. \noun off
  7166. \color none
  7167. 0.657
  7168. \end_layout
  7169. \end_inset
  7170. </cell>
  7171. </row>
  7172. <row>
  7173. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  7174. \begin_inset Text
  7175. \begin_layout Plain Layout
  7176. \family roman
  7177. \series medium
  7178. \shape up
  7179. \size normal
  7180. \emph off
  7181. \bar no
  7182. \strikeout off
  7183. \xout off
  7184. \uuline off
  7185. \uwave off
  7186. \noun off
  7187. \color none
  7188. RMA + GRSN
  7189. \end_layout
  7190. \end_inset
  7191. </cell>
  7192. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  7193. \begin_inset Text
  7194. \begin_layout Plain Layout
  7195. No
  7196. \end_layout
  7197. \end_inset
  7198. </cell>
  7199. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  7200. \begin_inset Text
  7201. \begin_layout Plain Layout
  7202. \family roman
  7203. \series medium
  7204. \shape up
  7205. \size normal
  7206. \emph off
  7207. \bar no
  7208. \strikeout off
  7209. \xout off
  7210. \uuline off
  7211. \uwave off
  7212. \noun off
  7213. \color none
  7214. 0.816
  7215. \end_layout
  7216. \end_inset
  7217. </cell>
  7218. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  7219. \begin_inset Text
  7220. \begin_layout Plain Layout
  7221. \family roman
  7222. \series medium
  7223. \shape up
  7224. \size normal
  7225. \emph off
  7226. \bar no
  7227. \strikeout off
  7228. \xout off
  7229. \uuline off
  7230. \uwave off
  7231. \noun off
  7232. \color none
  7233. 0.750
  7234. \end_layout
  7235. \end_inset
  7236. </cell>
  7237. </row>
  7238. <row>
  7239. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  7240. \begin_inset Text
  7241. \begin_layout Plain Layout
  7242. \family roman
  7243. \series medium
  7244. \shape up
  7245. \size normal
  7246. \emph off
  7247. \bar no
  7248. \strikeout off
  7249. \xout off
  7250. \uuline off
  7251. \uwave off
  7252. \noun off
  7253. \color none
  7254. dChip + GRSN
  7255. \end_layout
  7256. \end_inset
  7257. </cell>
  7258. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  7259. \begin_inset Text
  7260. \begin_layout Plain Layout
  7261. No
  7262. \end_layout
  7263. \end_inset
  7264. </cell>
  7265. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  7266. \begin_inset Text
  7267. \begin_layout Plain Layout
  7268. \family roman
  7269. \series medium
  7270. \shape up
  7271. \size normal
  7272. \emph off
  7273. \bar no
  7274. \strikeout off
  7275. \xout off
  7276. \uuline off
  7277. \uwave off
  7278. \noun off
  7279. \color none
  7280. 0.875
  7281. \end_layout
  7282. \end_inset
  7283. </cell>
  7284. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  7285. \begin_inset Text
  7286. \begin_layout Plain Layout
  7287. \family roman
  7288. \series medium
  7289. \shape up
  7290. \size normal
  7291. \emph off
  7292. \bar no
  7293. \strikeout off
  7294. \xout off
  7295. \uuline off
  7296. \uwave off
  7297. \noun off
  7298. \color none
  7299. 0.642
  7300. \end_layout
  7301. \end_inset
  7302. </cell>
  7303. </row>
  7304. <row>
  7305. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  7306. \begin_inset Text
  7307. \begin_layout Plain Layout
  7308. \family roman
  7309. \series medium
  7310. \shape up
  7311. \size normal
  7312. \emph off
  7313. \bar no
  7314. \strikeout off
  7315. \xout off
  7316. \uuline off
  7317. \uwave off
  7318. \noun off
  7319. \color none
  7320. fRMA
  7321. \end_layout
  7322. \end_inset
  7323. </cell>
  7324. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  7325. \begin_inset Text
  7326. \begin_layout Plain Layout
  7327. Yes
  7328. \end_layout
  7329. \end_inset
  7330. </cell>
  7331. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  7332. \begin_inset Text
  7333. \begin_layout Plain Layout
  7334. \family roman
  7335. \series medium
  7336. \shape up
  7337. \size normal
  7338. \emph off
  7339. \bar no
  7340. \strikeout off
  7341. \xout off
  7342. \uuline off
  7343. \uwave off
  7344. \noun off
  7345. \color none
  7346. 0.863
  7347. \end_layout
  7348. \end_inset
  7349. </cell>
  7350. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  7351. \begin_inset Text
  7352. \begin_layout Plain Layout
  7353. \family roman
  7354. \series medium
  7355. \shape up
  7356. \size normal
  7357. \emph off
  7358. \bar no
  7359. \strikeout off
  7360. \xout off
  7361. \uuline off
  7362. \uwave off
  7363. \noun off
  7364. \color none
  7365. 0.718
  7366. \end_layout
  7367. \end_inset
  7368. </cell>
  7369. </row>
  7370. <row>
  7371. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  7372. \begin_inset Text
  7373. \begin_layout Plain Layout
  7374. \family roman
  7375. \series medium
  7376. \shape up
  7377. \size normal
  7378. \emph off
  7379. \bar no
  7380. \strikeout off
  7381. \xout off
  7382. \uuline off
  7383. \uwave off
  7384. \noun off
  7385. \color none
  7386. SCAN
  7387. \end_layout
  7388. \end_inset
  7389. </cell>
  7390. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  7391. \begin_inset Text
  7392. \begin_layout Plain Layout
  7393. Yes
  7394. \end_layout
  7395. \end_inset
  7396. </cell>
  7397. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  7398. \begin_inset Text
  7399. \begin_layout Plain Layout
  7400. \family roman
  7401. \series medium
  7402. \shape up
  7403. \size normal
  7404. \emph off
  7405. \bar no
  7406. \strikeout off
  7407. \xout off
  7408. \uuline off
  7409. \uwave off
  7410. \noun off
  7411. \color none
  7412. 0.853
  7413. \end_layout
  7414. \end_inset
  7415. </cell>
  7416. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  7417. \begin_inset Text
  7418. \begin_layout Plain Layout
  7419. \family roman
  7420. \series medium
  7421. \shape up
  7422. \size normal
  7423. \emph off
  7424. \bar no
  7425. \strikeout off
  7426. \xout off
  7427. \uuline off
  7428. \uwave off
  7429. \noun off
  7430. \color none
  7431. 0.689
  7432. \end_layout
  7433. \end_inset
  7434. </cell>
  7435. </row>
  7436. </lyxtabular>
  7437. \end_inset
  7438. \end_layout
  7439. \begin_layout Plain Layout
  7440. \begin_inset Caption Standard
  7441. \begin_layout Plain Layout
  7442. \begin_inset CommandInset label
  7443. LatexCommand label
  7444. name "tab:AUC-PAM"
  7445. \end_inset
  7446. \series bold
  7447. ROC curve AUC values for internal and external validation with 6 different
  7448. normalization strategies.
  7449. \series default
  7450. These AUC values correspond to the ROC curves in Figure
  7451. \begin_inset CommandInset ref
  7452. LatexCommand ref
  7453. reference "fig:ROC-PAM-main"
  7454. plural "false"
  7455. caps "false"
  7456. noprefix "false"
  7457. \end_inset
  7458. .
  7459. \end_layout
  7460. \end_inset
  7461. \end_layout
  7462. \end_inset
  7463. \end_layout
  7464. \begin_layout Standard
  7465. For internal validation, the 6 methods' AUC values ranged from 0.816 to 0.891,
  7466. as shown in Table
  7467. \begin_inset CommandInset ref
  7468. LatexCommand ref
  7469. reference "tab:AUC-PAM"
  7470. plural "false"
  7471. caps "false"
  7472. noprefix "false"
  7473. \end_inset
  7474. .
  7475. Among the non-single-channel normalizations, dChip outperformed RMA, while
  7476. GRSN reduced the AUC values for both dChip and RMA.
  7477. Both single-channel methods, fRMA and SCAN, slightly outperformed RMA,
  7478. with fRMA ahead of SCAN.
  7479. However, the difference between RMA and fRMA is still quite small.
  7480. Figure
  7481. \begin_inset CommandInset ref
  7482. LatexCommand ref
  7483. reference "fig:ROC-PAM-int"
  7484. plural "false"
  7485. caps "false"
  7486. noprefix "false"
  7487. \end_inset
  7488. shows that the ROC curves for RMA, dChip, and fRMA look very similar and
  7489. relatively smooth, while both GRSN curves and the curve for SCAN have a
  7490. more jagged appearance.
  7491. \end_layout
  7492. \begin_layout Standard
  7493. For external validation, as expected, all the AUC values are lower than
  7494. the internal validations, ranging from 0.642 to 0.750 (Table
  7495. \begin_inset CommandInset ref
  7496. LatexCommand ref
  7497. reference "tab:AUC-PAM"
  7498. plural "false"
  7499. caps "false"
  7500. noprefix "false"
  7501. \end_inset
  7502. ).
  7503. With or without GRSN, RMA shows its dominance over dChip in this more challengi
  7504. ng test.
  7505. Unlike in the internal validation, GRSN actually improves the classifier
  7506. performance for RMA, although it does not for dChip.
  7507. Once again, both single-channel methods perform about on par with RMA,
  7508. with fRMA performing slightly better and SCAN performing a bit worse.
  7509. Figure
  7510. \begin_inset CommandInset ref
  7511. LatexCommand ref
  7512. reference "fig:ROC-PAM-ext"
  7513. plural "false"
  7514. caps "false"
  7515. noprefix "false"
  7516. \end_inset
  7517. shows the ROC curves for the external validation test.
  7518. As expected, none of them are as clean-looking as the internal validation
  7519. ROC curves.
  7520. The curves for RMA, RMA+GRSN, and fRMA all look similar, while the other
  7521. curves look more divergent.
  7522. \end_layout
  7523. \begin_layout Subsection
  7524. fRMA with custom-generated vectors enables single-channel normalization
  7525. on hthgu133pluspm platform
  7526. \end_layout
  7527. \begin_layout Standard
  7528. \begin_inset Float figure
  7529. wide false
  7530. sideways false
  7531. status open
  7532. \begin_layout Plain Layout
  7533. \align center
  7534. \begin_inset Float figure
  7535. placement tb
  7536. wide false
  7537. sideways false
  7538. status collapsed
  7539. \begin_layout Plain Layout
  7540. \align center
  7541. \begin_inset Graphics
  7542. filename graphics/frma-pax-bx/batchsize_batches.pdf
  7543. lyxscale 50
  7544. height 35theight%
  7545. groupId frmatools-subfig
  7546. \end_inset
  7547. \end_layout
  7548. \begin_layout Plain Layout
  7549. \begin_inset Caption Standard
  7550. \begin_layout Plain Layout
  7551. \begin_inset CommandInset label
  7552. LatexCommand label
  7553. name "fig:batch-size-batches"
  7554. \end_inset
  7555. \series bold
  7556. Number of batches usable in fRMA probe weight learning as a function of
  7557. batch size.
  7558. \end_layout
  7559. \end_inset
  7560. \end_layout
  7561. \end_inset
  7562. \end_layout
  7563. \begin_layout Plain Layout
  7564. \align center
  7565. \begin_inset Float figure
  7566. placement tb
  7567. wide false
  7568. sideways false
  7569. status collapsed
  7570. \begin_layout Plain Layout
  7571. \align center
  7572. \begin_inset Graphics
  7573. filename graphics/frma-pax-bx/batchsize_samples.pdf
  7574. lyxscale 50
  7575. height 35theight%
  7576. groupId frmatools-subfig
  7577. \end_inset
  7578. \end_layout
  7579. \begin_layout Plain Layout
  7580. \begin_inset Caption Standard
  7581. \begin_layout Plain Layout
  7582. \begin_inset CommandInset label
  7583. LatexCommand label
  7584. name "fig:batch-size-samples"
  7585. \end_inset
  7586. \series bold
  7587. Number of samples usable in fRMA probe weight learning as a function of
  7588. batch size.
  7589. \end_layout
  7590. \end_inset
  7591. \end_layout
  7592. \end_inset
  7593. \end_layout
  7594. \begin_layout Plain Layout
  7595. \begin_inset Caption Standard
  7596. \begin_layout Plain Layout
  7597. \series bold
  7598. \begin_inset CommandInset label
  7599. LatexCommand label
  7600. name "fig:frmatools-batch-size"
  7601. \end_inset
  7602. Effect of batch size selection on number of batches and number of samples
  7603. included in fRMA probe weight learning.
  7604. \series default
  7605. For batch sizes ranging from 3 to 15, the number of batches (a) and samples
  7606. (b) included in probe weight training were plotted for biopsy (BX) and
  7607. blood (PAX) samples.
  7608. The selected batch size, 5, is marked with a dotted vertical line.
  7609. \end_layout
  7610. \end_inset
  7611. \end_layout
  7612. \end_inset
  7613. \end_layout
  7614. \begin_layout Standard
  7615. In order to enable use of fRMA to normalize hthgu133pluspm, a custom set
  7616. of fRMA vectors was created.
  7617. First, an appropriate batch size was chosen by looking at the number of
  7618. batches and number of samples included as a function of batch size (Figure
  7619. \begin_inset CommandInset ref
  7620. LatexCommand ref
  7621. reference "fig:frmatools-batch-size"
  7622. plural "false"
  7623. caps "false"
  7624. noprefix "false"
  7625. \end_inset
  7626. ).
  7627. For a given batch size, all batches with fewer samples that the chosen
  7628. size must be ignored during training, while larger batches must be randomly
  7629. downsampled to the chosen size.
  7630. Hence, the number of samples included for a given batch size equals the
  7631. batch size times the number of batches with at least that many samples.
  7632. From Figure
  7633. \begin_inset CommandInset ref
  7634. LatexCommand ref
  7635. reference "fig:batch-size-samples"
  7636. plural "false"
  7637. caps "false"
  7638. noprefix "false"
  7639. \end_inset
  7640. , it is apparent that that a batch size of 8 maximizes the number of samples
  7641. included in training.
  7642. Increasing the batch size beyond this causes too many smaller batches to
  7643. be excluded, reducing the total number of samples for both tissue types.
  7644. However, a batch size of 8 is not necessarily optimal.
  7645. The article introducing frmaTools concluded that it was highly advantageous
  7646. to use a smaller batch size in order to include more batches, even at the
  7647. expense of including fewer total samples in training
  7648. \begin_inset CommandInset citation
  7649. LatexCommand cite
  7650. key "McCall2011"
  7651. literal "false"
  7652. \end_inset
  7653. .
  7654. To strike an appropriate balance between more batches and more samples,
  7655. a batch size of 5 was chosen.
  7656. For both blood and biopsy samples, this increased the number of batches
  7657. included by 10, with only a modest reduction in the number of samples compared
  7658. to a batch size of 8.
  7659. With a batch size of 5, 26 batches of biopsy samples and 46 batches of
  7660. blood samples were available.
  7661. \end_layout
  7662. \begin_layout Standard
  7663. \begin_inset Float figure
  7664. wide false
  7665. sideways false
  7666. status collapsed
  7667. \begin_layout Plain Layout
  7668. \begin_inset Float figure
  7669. wide false
  7670. sideways false
  7671. status open
  7672. \begin_layout Plain Layout
  7673. \align center
  7674. \begin_inset Graphics
  7675. filename graphics/frma-pax-bx/M-BX-violin.pdf
  7676. lyxscale 40
  7677. width 45col%
  7678. groupId m-violin
  7679. \end_inset
  7680. \end_layout
  7681. \begin_layout Plain Layout
  7682. \begin_inset Caption Standard
  7683. \begin_layout Plain Layout
  7684. \begin_inset CommandInset label
  7685. LatexCommand label
  7686. name "fig:m-bx-violin"
  7687. \end_inset
  7688. \series bold
  7689. Violin plot of inter-normalization log ratios for biopsy samples.
  7690. \end_layout
  7691. \end_inset
  7692. \end_layout
  7693. \end_inset
  7694. \begin_inset space \hfill{}
  7695. \end_inset
  7696. \begin_inset Float figure
  7697. wide false
  7698. sideways false
  7699. status collapsed
  7700. \begin_layout Plain Layout
  7701. \align center
  7702. \begin_inset Graphics
  7703. filename graphics/frma-pax-bx/M-PAX-violin.pdf
  7704. lyxscale 40
  7705. width 45col%
  7706. groupId m-violin
  7707. \end_inset
  7708. \end_layout
  7709. \begin_layout Plain Layout
  7710. \begin_inset Caption Standard
  7711. \begin_layout Plain Layout
  7712. \begin_inset CommandInset label
  7713. LatexCommand label
  7714. name "fig:m-pax-violin"
  7715. \end_inset
  7716. \series bold
  7717. Violin plot of inter-normalization log ratios for blood samples.
  7718. \end_layout
  7719. \end_inset
  7720. \end_layout
  7721. \end_inset
  7722. \end_layout
  7723. \begin_layout Plain Layout
  7724. \begin_inset Caption Standard
  7725. \begin_layout Plain Layout
  7726. \begin_inset CommandInset label
  7727. LatexCommand label
  7728. name "fig:frma-violin"
  7729. \end_inset
  7730. \series bold
  7731. Violin plot of log ratios between normalizations for 20 biopsy samples.
  7732. \series default
  7733. Each of 20 randomly selected samples was normalized with RMA and with 5
  7734. different sets of fRMA vectors.
  7735. The distribution of log ratios between normalized expression values, aggregated
  7736. across all 20 arrays, was plotted for each pair of normalizations.
  7737. \end_layout
  7738. \end_inset
  7739. \end_layout
  7740. \end_inset
  7741. \end_layout
  7742. \begin_layout Standard
  7743. Since fRMA training requires equal-size batches, larger batches are downsampled
  7744. randomly.
  7745. This introduces a nondeterministic step in the generation of normalization
  7746. vectors.
  7747. To show that this randomness does not substantially change the outcome,
  7748. the random downsampling and subsequent vector learning was repeated 5 times,
  7749. with a different random seed each time.
  7750. 20 samples were selected at random as a test set and normalized with each
  7751. of the 5 sets of fRMA normalization vectors as well as ordinary RMA, and
  7752. the normalized expression values were compared across normalizations.
  7753. Figure
  7754. \begin_inset CommandInset ref
  7755. LatexCommand ref
  7756. reference "fig:m-bx-violin"
  7757. plural "false"
  7758. caps "false"
  7759. noprefix "false"
  7760. \end_inset
  7761. shows a summary of these comparisons for biopsy samples.
  7762. Comparing RMA to each of the 5 fRMA normalizations, the distribution of
  7763. log ratios is somewhat wide, indicating that the normalizations disagree
  7764. on the expression values of a fair number of probe sets.
  7765. In contrast, comparisons of fRMA against fRMA, the vast majority of probe
  7766. sets have very small log ratios, indicating a very high agreement between
  7767. the normalized values generated by the two normalizations.
  7768. This shows that the fRMA normalization's behavior is not very sensitive
  7769. to the random downsampling of larger batches during training.
  7770. \end_layout
  7771. \begin_layout Standard
  7772. \begin_inset Float figure
  7773. wide false
  7774. sideways false
  7775. status open
  7776. \begin_layout Plain Layout
  7777. \align center
  7778. \begin_inset Float figure
  7779. wide false
  7780. sideways false
  7781. status collapsed
  7782. \begin_layout Plain Layout
  7783. \align center
  7784. \begin_inset Graphics
  7785. filename graphics/frma-pax-bx/MA-BX-RMA.fRMA-RASTER.png
  7786. lyxscale 10
  7787. width 45col%
  7788. groupId ma-frma
  7789. \end_inset
  7790. \end_layout
  7791. \begin_layout Plain Layout
  7792. \begin_inset Caption Standard
  7793. \begin_layout Plain Layout
  7794. \begin_inset CommandInset label
  7795. LatexCommand label
  7796. name "fig:ma-bx-rma-frma"
  7797. \end_inset
  7798. RMA vs.
  7799. fRMA for biopsy samples.
  7800. \end_layout
  7801. \end_inset
  7802. \end_layout
  7803. \end_inset
  7804. \begin_inset space \hfill{}
  7805. \end_inset
  7806. \begin_inset Float figure
  7807. wide false
  7808. sideways false
  7809. status collapsed
  7810. \begin_layout Plain Layout
  7811. \align center
  7812. \begin_inset Graphics
  7813. filename graphics/frma-pax-bx/MA-BX-fRMA.fRMA-RASTER.png
  7814. lyxscale 10
  7815. width 45col%
  7816. groupId ma-frma
  7817. \end_inset
  7818. \end_layout
  7819. \begin_layout Plain Layout
  7820. \begin_inset Caption Standard
  7821. \begin_layout Plain Layout
  7822. \begin_inset CommandInset label
  7823. LatexCommand label
  7824. name "fig:ma-bx-frma-frma"
  7825. \end_inset
  7826. fRMA vs fRMA for biopsy samples.
  7827. \end_layout
  7828. \end_inset
  7829. \end_layout
  7830. \end_inset
  7831. \end_layout
  7832. \begin_layout Plain Layout
  7833. \align center
  7834. \begin_inset Float figure
  7835. wide false
  7836. sideways false
  7837. status collapsed
  7838. \begin_layout Plain Layout
  7839. \align center
  7840. \begin_inset Graphics
  7841. filename graphics/frma-pax-bx/MA-PAX-RMA.fRMA-RASTER.png
  7842. lyxscale 10
  7843. width 45col%
  7844. groupId ma-frma
  7845. \end_inset
  7846. \end_layout
  7847. \begin_layout Plain Layout
  7848. \begin_inset Caption Standard
  7849. \begin_layout Plain Layout
  7850. \begin_inset CommandInset label
  7851. LatexCommand label
  7852. name "fig:MA-PAX-rma-frma"
  7853. \end_inset
  7854. RMA vs.
  7855. fRMA for blood samples.
  7856. \end_layout
  7857. \end_inset
  7858. \end_layout
  7859. \end_inset
  7860. \begin_inset space \hfill{}
  7861. \end_inset
  7862. \begin_inset Float figure
  7863. wide false
  7864. sideways false
  7865. status collapsed
  7866. \begin_layout Plain Layout
  7867. \align center
  7868. \begin_inset Graphics
  7869. filename graphics/frma-pax-bx/MA-PAX-fRMA.fRMA-RASTER.png
  7870. lyxscale 10
  7871. width 45col%
  7872. groupId ma-frma
  7873. \end_inset
  7874. \end_layout
  7875. \begin_layout Plain Layout
  7876. \begin_inset Caption Standard
  7877. \begin_layout Plain Layout
  7878. \begin_inset CommandInset label
  7879. LatexCommand label
  7880. name "fig:MA-PAX-frma-frma"
  7881. \end_inset
  7882. fRMA vs fRMA for blood samples.
  7883. \end_layout
  7884. \end_inset
  7885. \end_layout
  7886. \end_inset
  7887. \end_layout
  7888. \begin_layout Plain Layout
  7889. \begin_inset Caption Standard
  7890. \begin_layout Plain Layout
  7891. \series bold
  7892. \begin_inset CommandInset label
  7893. LatexCommand label
  7894. name "fig:Representative-MA-plots"
  7895. \end_inset
  7896. Representative MA plots comparing RMA and custom fRMA normalizations.
  7897. \series default
  7898. For each plot, 20 samples were normalized using 2 different normalizations,
  7899. and then averages (A) and log ratios (M) were plotted between the two different
  7900. normalizations for every probe.
  7901. For the
  7902. \begin_inset Quotes eld
  7903. \end_inset
  7904. fRMA vs fRMA
  7905. \begin_inset Quotes erd
  7906. \end_inset
  7907. plots (b & d), two different fRMA normalizations using vectors from two
  7908. independent batch samplings were compared.
  7909. Density of points is represented by blue shading, and individual outlier
  7910. points are plotted.
  7911. \end_layout
  7912. \end_inset
  7913. \end_layout
  7914. \end_inset
  7915. \end_layout
  7916. \begin_layout Standard
  7917. Figure
  7918. \begin_inset CommandInset ref
  7919. LatexCommand ref
  7920. reference "fig:ma-bx-rma-frma"
  7921. plural "false"
  7922. caps "false"
  7923. noprefix "false"
  7924. \end_inset
  7925. shows an MA plot of the RMA-normalized values against the fRMA-normalized
  7926. values for the same probe sets and arrays, corresponding to the first row
  7927. of Figure
  7928. \begin_inset CommandInset ref
  7929. LatexCommand ref
  7930. reference "fig:m-bx-violin"
  7931. plural "false"
  7932. caps "false"
  7933. noprefix "false"
  7934. \end_inset
  7935. .
  7936. This MA plot shows that not only is there a wide distribution of M-values,
  7937. but the trend of M-values is dependent on the average normalized intensity.
  7938. This is expected, since the overall trend represents the differences in
  7939. the quantile normalization step.
  7940. When running RMA, only the quantiles for these specific 20 arrays are used,
  7941. while for fRMA the quantile distribution is taking from all arrays used
  7942. in training.
  7943. Figure
  7944. \begin_inset CommandInset ref
  7945. LatexCommand ref
  7946. reference "fig:ma-bx-frma-frma"
  7947. plural "false"
  7948. caps "false"
  7949. noprefix "false"
  7950. \end_inset
  7951. shows a similar MA plot comparing 2 different fRMA normalizations, correspondin
  7952. g to the 6th row of Figure
  7953. \begin_inset CommandInset ref
  7954. LatexCommand ref
  7955. reference "fig:m-bx-violin"
  7956. plural "false"
  7957. caps "false"
  7958. noprefix "false"
  7959. \end_inset
  7960. .
  7961. The MA plot is very tightly centered around zero with no visible trend.
  7962. Figures
  7963. \begin_inset CommandInset ref
  7964. LatexCommand ref
  7965. reference "fig:m-pax-violin"
  7966. plural "false"
  7967. caps "false"
  7968. noprefix "false"
  7969. \end_inset
  7970. ,
  7971. \begin_inset CommandInset ref
  7972. LatexCommand ref
  7973. reference "fig:MA-PAX-rma-frma"
  7974. plural "false"
  7975. caps "false"
  7976. noprefix "false"
  7977. \end_inset
  7978. , and
  7979. \begin_inset CommandInset ref
  7980. LatexCommand ref
  7981. reference "fig:ma-bx-frma-frma"
  7982. plural "false"
  7983. caps "false"
  7984. noprefix "false"
  7985. \end_inset
  7986. show exactly the same information for the blood samples, once again comparing
  7987. the normalized expression values between normalizations for all probe sets
  7988. across 20 randomly selected test arrays.
  7989. Once again, there is a wider distribution of log ratios between RMA-normalized
  7990. values and fRMA-normalized, and a much tighter distribution when comparing
  7991. different fRMA normalizations to each other, indicating that the fRMA training
  7992. process is robust to random batch downsampling for the blood samples as
  7993. well.
  7994. \end_layout
  7995. \begin_layout Subsection
  7996. SVA, voom, and array weights improve model fit for methylation array data
  7997. \end_layout
  7998. \begin_layout Standard
  7999. \begin_inset ERT
  8000. status open
  8001. \begin_layout Plain Layout
  8002. \backslash
  8003. afterpage{
  8004. \end_layout
  8005. \begin_layout Plain Layout
  8006. \backslash
  8007. begin{landscape}
  8008. \end_layout
  8009. \end_inset
  8010. \end_layout
  8011. \begin_layout Standard
  8012. \begin_inset Float figure
  8013. wide false
  8014. sideways false
  8015. status open
  8016. \begin_layout Plain Layout
  8017. \begin_inset Flex TODO Note (inline)
  8018. status open
  8019. \begin_layout Plain Layout
  8020. Fix axis labels:
  8021. \begin_inset Quotes eld
  8022. \end_inset
  8023. log2 M-value
  8024. \begin_inset Quotes erd
  8025. \end_inset
  8026. is redundant because M-values are already log scale
  8027. \end_layout
  8028. \end_inset
  8029. \end_layout
  8030. \begin_layout Plain Layout
  8031. \begin_inset Float figure
  8032. wide false
  8033. sideways false
  8034. status collapsed
  8035. \begin_layout Plain Layout
  8036. \align center
  8037. \begin_inset Graphics
  8038. filename graphics/methylvoom/unadj.dupcor/meanvar-trends-PAGE1-CROP-RASTER.png
  8039. lyxscale 15
  8040. width 30col%
  8041. groupId voomaw-subfig
  8042. \end_inset
  8043. \end_layout
  8044. \begin_layout Plain Layout
  8045. \begin_inset Caption Standard
  8046. \begin_layout Plain Layout
  8047. \begin_inset CommandInset label
  8048. LatexCommand label
  8049. name "fig:meanvar-basic"
  8050. \end_inset
  8051. Mean-variance trend for analysis A.
  8052. \end_layout
  8053. \end_inset
  8054. \end_layout
  8055. \end_inset
  8056. \begin_inset space \hfill{}
  8057. \end_inset
  8058. \begin_inset Float figure
  8059. wide false
  8060. sideways false
  8061. status collapsed
  8062. \begin_layout Plain Layout
  8063. \align center
  8064. \begin_inset Graphics
  8065. filename graphics/methylvoom/unadj.dupcor.sva.aw/meanvar-trends-PAGE1-CROP-RASTER.png
  8066. lyxscale 15
  8067. width 30col%
  8068. groupId voomaw-subfig
  8069. \end_inset
  8070. \end_layout
  8071. \begin_layout Plain Layout
  8072. \begin_inset Caption Standard
  8073. \begin_layout Plain Layout
  8074. \begin_inset CommandInset label
  8075. LatexCommand label
  8076. name "fig:meanvar-sva-aw"
  8077. \end_inset
  8078. Mean-variance trend for analysis B.
  8079. \end_layout
  8080. \end_inset
  8081. \end_layout
  8082. \end_inset
  8083. \begin_inset space \hfill{}
  8084. \end_inset
  8085. \begin_inset Float figure
  8086. wide false
  8087. sideways false
  8088. status collapsed
  8089. \begin_layout Plain Layout
  8090. \align center
  8091. \begin_inset Graphics
  8092. filename graphics/methylvoom/unadj.dupcor.sva.voomaw/meanvar-trends-PAGE2-CROP-RASTER.png
  8093. lyxscale 15
  8094. width 30col%
  8095. groupId voomaw-subfig
  8096. \end_inset
  8097. \end_layout
  8098. \begin_layout Plain Layout
  8099. \begin_inset Caption Standard
  8100. \begin_layout Plain Layout
  8101. \begin_inset CommandInset label
  8102. LatexCommand label
  8103. name "fig:meanvar-sva-voomaw"
  8104. \end_inset
  8105. Mean-variance trend after voom modeling in analysis C.
  8106. \end_layout
  8107. \end_inset
  8108. \end_layout
  8109. \end_inset
  8110. \end_layout
  8111. \begin_layout Plain Layout
  8112. \begin_inset Caption Standard
  8113. \begin_layout Plain Layout
  8114. \series bold
  8115. Mean-variance trend modeling in methylation array data.
  8116. \series default
  8117. The estimated log2(standard deviation) for each probe is plotted against
  8118. the probe's average M-value across all samples as a black point, with some
  8119. transparency to make over-plotting more visible, since there are about
  8120. 450,000 points.
  8121. Density of points is also indicated by the dark blue contour lines.
  8122. The prior variance trend estimated by eBayes is shown in light blue, while
  8123. the lowess trend of the points is shown in red.
  8124. \end_layout
  8125. \end_inset
  8126. \end_layout
  8127. \end_inset
  8128. \end_layout
  8129. \begin_layout Standard
  8130. \begin_inset ERT
  8131. status open
  8132. \begin_layout Plain Layout
  8133. \backslash
  8134. end{landscape}
  8135. \end_layout
  8136. \begin_layout Plain Layout
  8137. }
  8138. \end_layout
  8139. \end_inset
  8140. \end_layout
  8141. \begin_layout Standard
  8142. Figure
  8143. \begin_inset CommandInset ref
  8144. LatexCommand ref
  8145. reference "fig:meanvar-basic"
  8146. plural "false"
  8147. caps "false"
  8148. noprefix "false"
  8149. \end_inset
  8150. shows the relationship between the mean M-value and the standard deviation
  8151. calculated for each probe in the methylation array data set.
  8152. A few features of the data are apparent.
  8153. First, the data are very strongly bimodal, with peaks in the density around
  8154. M-values of +4 and -4.
  8155. These modes correspond to methylation sites that are nearly 100% methylated
  8156. and nearly 100% unmethylated, respectively.
  8157. The strong bimodality indicates that a majority of probes interrogate sites
  8158. that fall into one of these two categories.
  8159. The points in between these modes represent sites that are either partially
  8160. methylated in many samples, or are fully methylated in some samples and
  8161. fully unmethylated in other samples, or some combination.
  8162. The next visible feature of the data is the W-shaped variance trend.
  8163. The upticks in the variance trend on either side are expected, based on
  8164. the sigmoid transformation exaggerating small differences at extreme M-values
  8165. (Figure
  8166. \begin_inset CommandInset ref
  8167. LatexCommand ref
  8168. reference "fig:Sigmoid-beta-m-mapping"
  8169. plural "false"
  8170. caps "false"
  8171. noprefix "false"
  8172. \end_inset
  8173. ).
  8174. However, the uptick in the center is interesting: it indicates that sites
  8175. that are not constitutively methylated or unmethylated have a higher variance.
  8176. This could be a genuine biological effect, or it could be spurious noise
  8177. that is only observable at sites with varying methylation.
  8178. \end_layout
  8179. \begin_layout Standard
  8180. In Figure
  8181. \begin_inset CommandInset ref
  8182. LatexCommand ref
  8183. reference "fig:meanvar-sva-aw"
  8184. plural "false"
  8185. caps "false"
  8186. noprefix "false"
  8187. \end_inset
  8188. , we see the mean-variance trend for the same methylation array data, this
  8189. time with surrogate variables and sample quality weights estimated from
  8190. the data and included in the model.
  8191. As expected, the overall average variance is smaller, since the surrogate
  8192. variables account for some of the variance.
  8193. In addition, the uptick in variance in the middle of the M-value range
  8194. has disappeared, turning the W shape into a wide U shape.
  8195. This indicates that the excess variance in the probes with intermediate
  8196. M-values was explained by systematic variations not correlated with known
  8197. covariates, and these variations were modeled by the surrogate variables.
  8198. The result is a nearly flat variance trend for the entire intermediate
  8199. M-value range from about -3 to +3.
  8200. Note that this corresponds closely to the range within which the M-value
  8201. transformation shown in Figure
  8202. \begin_inset CommandInset ref
  8203. LatexCommand ref
  8204. reference "fig:Sigmoid-beta-m-mapping"
  8205. plural "false"
  8206. caps "false"
  8207. noprefix "false"
  8208. \end_inset
  8209. is nearly linear.
  8210. In contrast, the excess variance at the extremes (greater than +3 and less
  8211. than -3) was not
  8212. \begin_inset Quotes eld
  8213. \end_inset
  8214. absorbed
  8215. \begin_inset Quotes erd
  8216. \end_inset
  8217. by the surrogate variables and remains in the plot, indicating that this
  8218. variation has no systematic component: probes with extreme M-values are
  8219. uniformly more variable across all samples, as expected.
  8220. \end_layout
  8221. \begin_layout Standard
  8222. Figure
  8223. \begin_inset CommandInset ref
  8224. LatexCommand ref
  8225. reference "fig:meanvar-sva-voomaw"
  8226. plural "false"
  8227. caps "false"
  8228. noprefix "false"
  8229. \end_inset
  8230. shows the mean-variance trend after fitting the model with the observation
  8231. weights assigned by voom based on the mean-variance trend shown in Figure
  8232. \begin_inset CommandInset ref
  8233. LatexCommand ref
  8234. reference "fig:meanvar-sva-aw"
  8235. plural "false"
  8236. caps "false"
  8237. noprefix "false"
  8238. \end_inset
  8239. .
  8240. As expected, the weights exactly counteract the trend in the data, resulting
  8241. in a nearly flat trend centered vertically at 1 (i.e.
  8242. 0 on the log scale).
  8243. This shows that the observations with extreme M-values have been appropriately
  8244. down-weighted to account for the fact that the noise in those observations
  8245. has been amplified by the non-linear M-value transformation.
  8246. In turn, this gives relatively more weight to observations in the middle
  8247. region, which are more likely to correspond to probes measuring interesting
  8248. biology (not constitutively methylated or unmethylated).
  8249. \end_layout
  8250. \begin_layout Standard
  8251. \begin_inset Float table
  8252. wide false
  8253. sideways false
  8254. status open
  8255. \begin_layout Plain Layout
  8256. \align center
  8257. \begin_inset Tabular
  8258. <lyxtabular version="3" rows="5" columns="3">
  8259. <features tabularvalignment="middle">
  8260. <column alignment="center" valignment="top">
  8261. <column alignment="center" valignment="top">
  8262. <column alignment="center" valignment="top">
  8263. <row>
  8264. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  8265. \begin_inset Text
  8266. \begin_layout Plain Layout
  8267. Covariate
  8268. \end_layout
  8269. \end_inset
  8270. </cell>
  8271. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  8272. \begin_inset Text
  8273. \begin_layout Plain Layout
  8274. Test used
  8275. \end_layout
  8276. \end_inset
  8277. </cell>
  8278. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  8279. \begin_inset Text
  8280. \begin_layout Plain Layout
  8281. p-value
  8282. \end_layout
  8283. \end_inset
  8284. </cell>
  8285. </row>
  8286. <row>
  8287. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  8288. \begin_inset Text
  8289. \begin_layout Plain Layout
  8290. Transplant Status
  8291. \end_layout
  8292. \end_inset
  8293. </cell>
  8294. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  8295. \begin_inset Text
  8296. \begin_layout Plain Layout
  8297. F-test
  8298. \end_layout
  8299. \end_inset
  8300. </cell>
  8301. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  8302. \begin_inset Text
  8303. \begin_layout Plain Layout
  8304. 0.404
  8305. \end_layout
  8306. \end_inset
  8307. </cell>
  8308. </row>
  8309. <row>
  8310. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  8311. \begin_inset Text
  8312. \begin_layout Plain Layout
  8313. Diabetes Diagnosis
  8314. \end_layout
  8315. \end_inset
  8316. </cell>
  8317. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  8318. \begin_inset Text
  8319. \begin_layout Plain Layout
  8320. \emph on
  8321. t
  8322. \emph default
  8323. -test
  8324. \end_layout
  8325. \end_inset
  8326. </cell>
  8327. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  8328. \begin_inset Text
  8329. \begin_layout Plain Layout
  8330. 0.00106
  8331. \end_layout
  8332. \end_inset
  8333. </cell>
  8334. </row>
  8335. <row>
  8336. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  8337. \begin_inset Text
  8338. \begin_layout Plain Layout
  8339. Sex
  8340. \end_layout
  8341. \end_inset
  8342. </cell>
  8343. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  8344. \begin_inset Text
  8345. \begin_layout Plain Layout
  8346. \emph on
  8347. t
  8348. \emph default
  8349. -test
  8350. \end_layout
  8351. \end_inset
  8352. </cell>
  8353. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  8354. \begin_inset Text
  8355. \begin_layout Plain Layout
  8356. 0.148
  8357. \end_layout
  8358. \end_inset
  8359. </cell>
  8360. </row>
  8361. <row>
  8362. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  8363. \begin_inset Text
  8364. \begin_layout Plain Layout
  8365. Age
  8366. \end_layout
  8367. \end_inset
  8368. </cell>
  8369. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  8370. \begin_inset Text
  8371. \begin_layout Plain Layout
  8372. linear regression
  8373. \end_layout
  8374. \end_inset
  8375. </cell>
  8376. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  8377. \begin_inset Text
  8378. \begin_layout Plain Layout
  8379. 0.212
  8380. \end_layout
  8381. \end_inset
  8382. </cell>
  8383. </row>
  8384. </lyxtabular>
  8385. \end_inset
  8386. \end_layout
  8387. \begin_layout Plain Layout
  8388. \begin_inset Caption Standard
  8389. \begin_layout Plain Layout
  8390. \series bold
  8391. \begin_inset CommandInset label
  8392. LatexCommand label
  8393. name "tab:weight-covariate-tests"
  8394. \end_inset
  8395. Association of sample weights with clinical covariates in methylation array
  8396. data.
  8397. \series default
  8398. Computed sample quality log weights were tested for significant association
  8399. with each of the variables in the model (1st column).
  8400. An appropriate test was selected for each variable based on whether the
  8401. variable had 2 categories (
  8402. \emph on
  8403. t
  8404. \emph default
  8405. -test), had more than 2 categories (F-test), or was numeric (linear regression).
  8406. The test selected is shown in the 2nd column.
  8407. P-values for association with the log weights are shown in the 3rd column.
  8408. No multiple testing adjustment was performed for these p-values.
  8409. \end_layout
  8410. \end_inset
  8411. \end_layout
  8412. \end_inset
  8413. \end_layout
  8414. \begin_layout Standard
  8415. \begin_inset Float figure
  8416. wide false
  8417. sideways false
  8418. status open
  8419. \begin_layout Plain Layout
  8420. \begin_inset Flex TODO Note (inline)
  8421. status open
  8422. \begin_layout Plain Layout
  8423. Redo the sample weight boxplot with notches, and remove fill colors
  8424. \end_layout
  8425. \end_inset
  8426. \end_layout
  8427. \begin_layout Plain Layout
  8428. \align center
  8429. \begin_inset Graphics
  8430. filename graphics/methylvoom/unadj.dupcor.sva.voomaw/sample-weights-PAGE3-CROP.pdf
  8431. lyxscale 50
  8432. width 60col%
  8433. groupId colwidth
  8434. \end_inset
  8435. \end_layout
  8436. \begin_layout Plain Layout
  8437. \begin_inset Caption Standard
  8438. \begin_layout Plain Layout
  8439. \begin_inset CommandInset label
  8440. LatexCommand label
  8441. name "fig:diabetes-sample-weights"
  8442. \end_inset
  8443. \series bold
  8444. Box-and-whiskers plot of sample quality weights grouped by diabetes diagnosis.
  8445. \series default
  8446. Samples were grouped based on diabetes diagnosis, and the distribution of
  8447. sample quality weights for each diagnosis was plotted as a box-and-whiskers
  8448. plot
  8449. \begin_inset CommandInset citation
  8450. LatexCommand cite
  8451. key "McGill1978"
  8452. literal "false"
  8453. \end_inset
  8454. .
  8455. \end_layout
  8456. \end_inset
  8457. \end_layout
  8458. \begin_layout Plain Layout
  8459. \end_layout
  8460. \end_inset
  8461. \end_layout
  8462. \begin_layout Standard
  8463. To determine whether any of the known experimental factors had an impact
  8464. on data quality, the sample quality weights estimated from the data were
  8465. tested for association with each of the experimental factors (Table
  8466. \begin_inset CommandInset ref
  8467. LatexCommand ref
  8468. reference "tab:weight-covariate-tests"
  8469. plural "false"
  8470. caps "false"
  8471. noprefix "false"
  8472. \end_inset
  8473. ).
  8474. Diabetes diagnosis was found to have a potentially significant association
  8475. with the sample weights, with a t-test p-value of
  8476. \begin_inset Formula $1.06\times10^{-3}$
  8477. \end_inset
  8478. .
  8479. Figure
  8480. \begin_inset CommandInset ref
  8481. LatexCommand ref
  8482. reference "fig:diabetes-sample-weights"
  8483. plural "false"
  8484. caps "false"
  8485. noprefix "false"
  8486. \end_inset
  8487. shows the distribution of sample weights grouped by diabetes diagnosis.
  8488. The samples from patients with Type 2 diabetes were assigned significantly
  8489. lower weights than those from patients with Type 1 diabetes.
  8490. This indicates that the type 2 diabetes samples had an overall higher variance
  8491. on average across all probes.
  8492. \end_layout
  8493. \begin_layout Standard
  8494. \begin_inset Float table
  8495. wide false
  8496. sideways false
  8497. status open
  8498. \begin_layout Plain Layout
  8499. \align center
  8500. \begin_inset Flex TODO Note (inline)
  8501. status open
  8502. \begin_layout Plain Layout
  8503. Consider transposing these tables
  8504. \end_layout
  8505. \end_inset
  8506. \end_layout
  8507. \begin_layout Plain Layout
  8508. \begin_inset Float table
  8509. wide false
  8510. sideways false
  8511. status open
  8512. \begin_layout Plain Layout
  8513. \align center
  8514. \begin_inset Tabular
  8515. <lyxtabular version="3" rows="5" columns="4">
  8516. <features tabularvalignment="middle">
  8517. <column alignment="center" valignment="top">
  8518. <column alignment="center" valignment="top">
  8519. <column alignment="center" valignment="top">
  8520. <column alignment="center" valignment="top">
  8521. <row>
  8522. <cell alignment="center" valignment="top" usebox="none">
  8523. \begin_inset Text
  8524. \begin_layout Plain Layout
  8525. \end_layout
  8526. \end_inset
  8527. </cell>
  8528. <cell multicolumn="1" alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  8529. \begin_inset Text
  8530. \begin_layout Plain Layout
  8531. Analysis
  8532. \end_layout
  8533. \end_inset
  8534. </cell>
  8535. <cell multicolumn="2" alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  8536. \begin_inset Text
  8537. \begin_layout Plain Layout
  8538. \end_layout
  8539. \end_inset
  8540. </cell>
  8541. <cell multicolumn="2" alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  8542. \begin_inset Text
  8543. \begin_layout Plain Layout
  8544. \end_layout
  8545. \end_inset
  8546. </cell>
  8547. </row>
  8548. <row>
  8549. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  8550. \begin_inset Text
  8551. \begin_layout Plain Layout
  8552. Contrast
  8553. \end_layout
  8554. \end_inset
  8555. </cell>
  8556. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  8557. \begin_inset Text
  8558. \begin_layout Plain Layout
  8559. A
  8560. \end_layout
  8561. \end_inset
  8562. </cell>
  8563. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  8564. \begin_inset Text
  8565. \begin_layout Plain Layout
  8566. B
  8567. \end_layout
  8568. \end_inset
  8569. </cell>
  8570. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  8571. \begin_inset Text
  8572. \begin_layout Plain Layout
  8573. C
  8574. \end_layout
  8575. \end_inset
  8576. </cell>
  8577. </row>
  8578. <row>
  8579. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  8580. \begin_inset Text
  8581. \begin_layout Plain Layout
  8582. TX vs AR
  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. 0
  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. 25
  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. 22
  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. TX vs ADNR
  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. 7
  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. 338
  8627. \end_layout
  8628. \end_inset
  8629. </cell>
  8630. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  8631. \begin_inset Text
  8632. \begin_layout Plain Layout
  8633. 369
  8634. \end_layout
  8635. \end_inset
  8636. </cell>
  8637. </row>
  8638. <row>
  8639. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  8640. \begin_inset Text
  8641. \begin_layout Plain Layout
  8642. TX vs CAN
  8643. \end_layout
  8644. \end_inset
  8645. </cell>
  8646. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  8647. \begin_inset Text
  8648. \begin_layout Plain Layout
  8649. 0
  8650. \end_layout
  8651. \end_inset
  8652. </cell>
  8653. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  8654. \begin_inset Text
  8655. \begin_layout Plain Layout
  8656. 231
  8657. \end_layout
  8658. \end_inset
  8659. </cell>
  8660. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  8661. \begin_inset Text
  8662. \begin_layout Plain Layout
  8663. 278
  8664. \end_layout
  8665. \end_inset
  8666. </cell>
  8667. </row>
  8668. </lyxtabular>
  8669. \end_inset
  8670. \end_layout
  8671. \begin_layout Plain Layout
  8672. \begin_inset Caption Standard
  8673. \begin_layout Plain Layout
  8674. \begin_inset CommandInset label
  8675. LatexCommand label
  8676. name "tab:methyl-num-signif"
  8677. \end_inset
  8678. Number of probes significant at 10% FDR.
  8679. \end_layout
  8680. \end_inset
  8681. \end_layout
  8682. \end_inset
  8683. \begin_inset space \hfill{}
  8684. \end_inset
  8685. \begin_inset Float table
  8686. wide false
  8687. sideways false
  8688. status open
  8689. \begin_layout Plain Layout
  8690. \align center
  8691. \begin_inset Tabular
  8692. <lyxtabular version="3" rows="5" columns="4">
  8693. <features tabularvalignment="middle">
  8694. <column alignment="center" valignment="top">
  8695. <column alignment="center" valignment="top">
  8696. <column alignment="center" valignment="top">
  8697. <column alignment="center" valignment="top">
  8698. <row>
  8699. <cell alignment="center" valignment="top" usebox="none">
  8700. \begin_inset Text
  8701. \begin_layout Plain Layout
  8702. \end_layout
  8703. \end_inset
  8704. </cell>
  8705. <cell multicolumn="1" alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  8706. \begin_inset Text
  8707. \begin_layout Plain Layout
  8708. Analysis
  8709. \end_layout
  8710. \end_inset
  8711. </cell>
  8712. <cell multicolumn="2" alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  8713. \begin_inset Text
  8714. \begin_layout Plain Layout
  8715. \end_layout
  8716. \end_inset
  8717. </cell>
  8718. <cell multicolumn="2" alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  8719. \begin_inset Text
  8720. \begin_layout Plain Layout
  8721. \end_layout
  8722. \end_inset
  8723. </cell>
  8724. </row>
  8725. <row>
  8726. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  8727. \begin_inset Text
  8728. \begin_layout Plain Layout
  8729. Contrast
  8730. \end_layout
  8731. \end_inset
  8732. </cell>
  8733. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  8734. \begin_inset Text
  8735. \begin_layout Plain Layout
  8736. A
  8737. \end_layout
  8738. \end_inset
  8739. </cell>
  8740. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  8741. \begin_inset Text
  8742. \begin_layout Plain Layout
  8743. B
  8744. \end_layout
  8745. \end_inset
  8746. </cell>
  8747. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  8748. \begin_inset Text
  8749. \begin_layout Plain Layout
  8750. C
  8751. \end_layout
  8752. \end_inset
  8753. </cell>
  8754. </row>
  8755. <row>
  8756. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  8757. \begin_inset Text
  8758. \begin_layout Plain Layout
  8759. TX vs AR
  8760. \end_layout
  8761. \end_inset
  8762. </cell>
  8763. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  8764. \begin_inset Text
  8765. \begin_layout Plain Layout
  8766. 0
  8767. \end_layout
  8768. \end_inset
  8769. </cell>
  8770. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  8771. \begin_inset Text
  8772. \begin_layout Plain Layout
  8773. 10,063
  8774. \end_layout
  8775. \end_inset
  8776. </cell>
  8777. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  8778. \begin_inset Text
  8779. \begin_layout Plain Layout
  8780. 11,225
  8781. \end_layout
  8782. \end_inset
  8783. </cell>
  8784. </row>
  8785. <row>
  8786. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  8787. \begin_inset Text
  8788. \begin_layout Plain Layout
  8789. TX vs ADNR
  8790. \end_layout
  8791. \end_inset
  8792. </cell>
  8793. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  8794. \begin_inset Text
  8795. \begin_layout Plain Layout
  8796. 27
  8797. \end_layout
  8798. \end_inset
  8799. </cell>
  8800. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  8801. \begin_inset Text
  8802. \begin_layout Plain Layout
  8803. 12,674
  8804. \end_layout
  8805. \end_inset
  8806. </cell>
  8807. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  8808. \begin_inset Text
  8809. \begin_layout Plain Layout
  8810. 13,086
  8811. \end_layout
  8812. \end_inset
  8813. </cell>
  8814. </row>
  8815. <row>
  8816. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  8817. \begin_inset Text
  8818. \begin_layout Plain Layout
  8819. TX vs CAN
  8820. \end_layout
  8821. \end_inset
  8822. </cell>
  8823. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  8824. \begin_inset Text
  8825. \begin_layout Plain Layout
  8826. 966
  8827. \end_layout
  8828. \end_inset
  8829. </cell>
  8830. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  8831. \begin_inset Text
  8832. \begin_layout Plain Layout
  8833. 20,039
  8834. \end_layout
  8835. \end_inset
  8836. </cell>
  8837. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  8838. \begin_inset Text
  8839. \begin_layout Plain Layout
  8840. 20,955
  8841. \end_layout
  8842. \end_inset
  8843. </cell>
  8844. </row>
  8845. </lyxtabular>
  8846. \end_inset
  8847. \end_layout
  8848. \begin_layout Plain Layout
  8849. \begin_inset Caption Standard
  8850. \begin_layout Plain Layout
  8851. \begin_inset CommandInset label
  8852. LatexCommand label
  8853. name "tab:methyl-est-nonnull"
  8854. \end_inset
  8855. Estimated number of non-null tests, using the method of averaging local
  8856. FDR values
  8857. \begin_inset CommandInset citation
  8858. LatexCommand cite
  8859. key "Phipson2013Thesis"
  8860. literal "false"
  8861. \end_inset
  8862. .
  8863. \end_layout
  8864. \end_inset
  8865. \end_layout
  8866. \end_inset
  8867. \end_layout
  8868. \begin_layout Plain Layout
  8869. \begin_inset Caption Standard
  8870. \begin_layout Plain Layout
  8871. \series bold
  8872. Estimates of degree of differential methylation in for each contrast in
  8873. each analysis.
  8874. \series default
  8875. For each of the analyses in Table
  8876. \begin_inset CommandInset ref
  8877. LatexCommand ref
  8878. reference "tab:Summary-of-meth-analysis"
  8879. plural "false"
  8880. caps "false"
  8881. noprefix "false"
  8882. \end_inset
  8883. , these tables show the number of probes called significantly differentially
  8884. methylated at a threshold of 10% FDR for each comparison between TX and
  8885. the other 3 transplant statuses (a) and the estimated total number of probes
  8886. that are differentially methylated (b).
  8887. \end_layout
  8888. \end_inset
  8889. \end_layout
  8890. \end_inset
  8891. \end_layout
  8892. \begin_layout Standard
  8893. \begin_inset Float figure
  8894. wide false
  8895. sideways false
  8896. status open
  8897. \begin_layout Plain Layout
  8898. \align center
  8899. \series bold
  8900. \begin_inset Float figure
  8901. wide false
  8902. sideways false
  8903. status collapsed
  8904. \begin_layout Plain Layout
  8905. \align center
  8906. \begin_inset Graphics
  8907. filename graphics/methylvoom/unadj.dupcor/pval-histograms-PAGE1.pdf
  8908. lyxscale 33
  8909. width 30col%
  8910. groupId meth-pval-hist
  8911. \end_inset
  8912. \end_layout
  8913. \begin_layout Plain Layout
  8914. \series bold
  8915. \begin_inset Caption Standard
  8916. \begin_layout Plain Layout
  8917. AR vs.
  8918. TX, Analysis A
  8919. \end_layout
  8920. \end_inset
  8921. \end_layout
  8922. \begin_layout Plain Layout
  8923. \end_layout
  8924. \end_inset
  8925. \begin_inset space \hfill{}
  8926. \end_inset
  8927. \begin_inset Float figure
  8928. wide false
  8929. sideways false
  8930. status collapsed
  8931. \begin_layout Plain Layout
  8932. \align center
  8933. \begin_inset Graphics
  8934. filename graphics/methylvoom/unadj.dupcor/pval-histograms-PAGE2.pdf
  8935. lyxscale 33
  8936. width 30col%
  8937. groupId meth-pval-hist
  8938. \end_inset
  8939. \end_layout
  8940. \begin_layout Plain Layout
  8941. \series bold
  8942. \begin_inset Caption Standard
  8943. \begin_layout Plain Layout
  8944. ADNR vs.
  8945. TX, Analysis A
  8946. \end_layout
  8947. \end_inset
  8948. \end_layout
  8949. \end_inset
  8950. \begin_inset space \hfill{}
  8951. \end_inset
  8952. \begin_inset Float figure
  8953. wide false
  8954. sideways false
  8955. status collapsed
  8956. \begin_layout Plain Layout
  8957. \align center
  8958. \begin_inset Graphics
  8959. filename graphics/methylvoom/unadj.dupcor/pval-histograms-PAGE3.pdf
  8960. lyxscale 33
  8961. width 30col%
  8962. groupId meth-pval-hist
  8963. \end_inset
  8964. \end_layout
  8965. \begin_layout Plain Layout
  8966. \series bold
  8967. \begin_inset Caption Standard
  8968. \begin_layout Plain Layout
  8969. CAN vs.
  8970. TX, Analysis A
  8971. \end_layout
  8972. \end_inset
  8973. \end_layout
  8974. \end_inset
  8975. \end_layout
  8976. \begin_layout Plain Layout
  8977. \align center
  8978. \series bold
  8979. \begin_inset Float figure
  8980. wide false
  8981. sideways false
  8982. status collapsed
  8983. \begin_layout Plain Layout
  8984. \align center
  8985. \begin_inset Graphics
  8986. filename graphics/methylvoom/unadj.dupcor.sva.aw/pval-histograms-PAGE1.pdf
  8987. lyxscale 33
  8988. width 30col%
  8989. groupId meth-pval-hist
  8990. \end_inset
  8991. \end_layout
  8992. \begin_layout Plain Layout
  8993. \series bold
  8994. \begin_inset Caption Standard
  8995. \begin_layout Plain Layout
  8996. AR vs.
  8997. TX, Analysis B
  8998. \end_layout
  8999. \end_inset
  9000. \end_layout
  9001. \end_inset
  9002. \begin_inset space \hfill{}
  9003. \end_inset
  9004. \begin_inset Float figure
  9005. wide false
  9006. sideways false
  9007. status collapsed
  9008. \begin_layout Plain Layout
  9009. \align center
  9010. \begin_inset Graphics
  9011. filename graphics/methylvoom/unadj.dupcor.sva.aw/pval-histograms-PAGE2.pdf
  9012. lyxscale 33
  9013. width 30col%
  9014. groupId meth-pval-hist
  9015. \end_inset
  9016. \end_layout
  9017. \begin_layout Plain Layout
  9018. \series bold
  9019. \begin_inset Caption Standard
  9020. \begin_layout Plain Layout
  9021. ADNR vs.
  9022. TX, Analysis B
  9023. \end_layout
  9024. \end_inset
  9025. \end_layout
  9026. \end_inset
  9027. \begin_inset space \hfill{}
  9028. \end_inset
  9029. \begin_inset Float figure
  9030. wide false
  9031. sideways false
  9032. status collapsed
  9033. \begin_layout Plain Layout
  9034. \align center
  9035. \begin_inset Graphics
  9036. filename graphics/methylvoom/unadj.dupcor.sva.aw/pval-histograms-PAGE3.pdf
  9037. lyxscale 33
  9038. width 30col%
  9039. groupId meth-pval-hist
  9040. \end_inset
  9041. \end_layout
  9042. \begin_layout Plain Layout
  9043. \series bold
  9044. \begin_inset Caption Standard
  9045. \begin_layout Plain Layout
  9046. CAN vs.
  9047. TX, Analysis B
  9048. \end_layout
  9049. \end_inset
  9050. \end_layout
  9051. \end_inset
  9052. \end_layout
  9053. \begin_layout Plain Layout
  9054. \align center
  9055. \series bold
  9056. \begin_inset Float figure
  9057. wide false
  9058. sideways false
  9059. status collapsed
  9060. \begin_layout Plain Layout
  9061. \align center
  9062. \begin_inset Graphics
  9063. filename graphics/methylvoom/unadj.dupcor.sva.voomaw/pval-histograms-PAGE1.pdf
  9064. lyxscale 33
  9065. width 30col%
  9066. groupId meth-pval-hist
  9067. \end_inset
  9068. \end_layout
  9069. \begin_layout Plain Layout
  9070. \series bold
  9071. \begin_inset Caption Standard
  9072. \begin_layout Plain Layout
  9073. AR vs.
  9074. TX, Analysis C
  9075. \end_layout
  9076. \end_inset
  9077. \end_layout
  9078. \end_inset
  9079. \begin_inset space \hfill{}
  9080. \end_inset
  9081. \begin_inset Float figure
  9082. wide false
  9083. sideways false
  9084. status collapsed
  9085. \begin_layout Plain Layout
  9086. \align center
  9087. \begin_inset Graphics
  9088. filename graphics/methylvoom/unadj.dupcor.sva.voomaw/pval-histograms-PAGE2.pdf
  9089. lyxscale 33
  9090. width 30col%
  9091. groupId meth-pval-hist
  9092. \end_inset
  9093. \end_layout
  9094. \begin_layout Plain Layout
  9095. \series bold
  9096. \begin_inset Caption Standard
  9097. \begin_layout Plain Layout
  9098. ADNR vs.
  9099. TX, Analysis C
  9100. \end_layout
  9101. \end_inset
  9102. \end_layout
  9103. \end_inset
  9104. \begin_inset space \hfill{}
  9105. \end_inset
  9106. \begin_inset Float figure
  9107. wide false
  9108. sideways false
  9109. status collapsed
  9110. \begin_layout Plain Layout
  9111. \align center
  9112. \begin_inset Graphics
  9113. filename graphics/methylvoom/unadj.dupcor.sva.voomaw/pval-histograms-PAGE3.pdf
  9114. lyxscale 33
  9115. width 30col%
  9116. groupId meth-pval-hist
  9117. \end_inset
  9118. \end_layout
  9119. \begin_layout Plain Layout
  9120. \series bold
  9121. \begin_inset Caption Standard
  9122. \begin_layout Plain Layout
  9123. CAN vs.
  9124. TX, Analysis C
  9125. \end_layout
  9126. \end_inset
  9127. \end_layout
  9128. \end_inset
  9129. \end_layout
  9130. \begin_layout Plain Layout
  9131. \begin_inset Caption Standard
  9132. \begin_layout Plain Layout
  9133. \series bold
  9134. \begin_inset CommandInset label
  9135. LatexCommand label
  9136. name "fig:meth-p-value-histograms"
  9137. \end_inset
  9138. Probe p-value histograms for each contrast in each analysis.
  9139. \series default
  9140. For each differential methylation test of interest, the distribution of
  9141. p-values across all probes is plotted as a histogram.
  9142. The red solid line indicates the density that would be expected under the
  9143. null hypothesis for all probes (a
  9144. \begin_inset Formula $\mathrm{Uniform}(0,1)$
  9145. \end_inset
  9146. distribution), while the blue dotted line indicates the fraction of p-values
  9147. that actually follow the null hypothesis (
  9148. \begin_inset Formula $\hat{\pi}_{0}$
  9149. \end_inset
  9150. ) estimated using the method of averaging local FDR values
  9151. \begin_inset CommandInset citation
  9152. LatexCommand cite
  9153. key "Phipson2013Thesis"
  9154. literal "false"
  9155. \end_inset
  9156. .
  9157. the blue line is only shown in each plot if the estimate of
  9158. \begin_inset Formula $\hat{\pi}_{0}$
  9159. \end_inset
  9160. for that p-value distribution is different from 1.
  9161. \end_layout
  9162. \end_inset
  9163. \end_layout
  9164. \end_inset
  9165. \end_layout
  9166. \begin_layout Standard
  9167. Table
  9168. \begin_inset CommandInset ref
  9169. LatexCommand ref
  9170. reference "tab:methyl-num-signif"
  9171. plural "false"
  9172. caps "false"
  9173. noprefix "false"
  9174. \end_inset
  9175. shows the number of significantly differentially methylated probes reported
  9176. by each analysis for each comparison of interest at an FDR of 10%.
  9177. As expected, the more elaborate analyses, B and C, report more significant
  9178. probes than the more basic analysis A, consistent with the conclusions
  9179. above that the data contain hidden systematic variations that must be modeled.
  9180. Table
  9181. \begin_inset CommandInset ref
  9182. LatexCommand ref
  9183. reference "tab:methyl-est-nonnull"
  9184. plural "false"
  9185. caps "false"
  9186. noprefix "false"
  9187. \end_inset
  9188. shows the estimated number differentially methylated probes for each test
  9189. from each analysis.
  9190. This was computed by estimating the proportion of null hypotheses that
  9191. were true using the method of
  9192. \begin_inset CommandInset citation
  9193. LatexCommand cite
  9194. key "Phipson2013Thesis"
  9195. literal "false"
  9196. \end_inset
  9197. and subtracting that fraction from the total number of probes, yielding
  9198. an estimate of the number of null hypotheses that are false based on the
  9199. distribution of p-values across the entire dataset.
  9200. Note that this does not identify which null hypotheses should be rejected
  9201. (i.e.
  9202. which probes are significant); it only estimates the true number of such
  9203. probes.
  9204. Once again, analyses B and C result it much larger estimates for the number
  9205. of differentially methylated probes.
  9206. In this case, analysis C, the only analysis that includes voom, estimates
  9207. the largest number of differentially methylated probes for all 3 contrasts.
  9208. If the assumptions of all the methods employed hold, then this represents
  9209. a gain in statistical power over the simpler analysis A.
  9210. Figure
  9211. \begin_inset CommandInset ref
  9212. LatexCommand ref
  9213. reference "fig:meth-p-value-histograms"
  9214. plural "false"
  9215. caps "false"
  9216. noprefix "false"
  9217. \end_inset
  9218. shows the p-value distributions for each test, from which the numbers in
  9219. Table
  9220. \begin_inset CommandInset ref
  9221. LatexCommand ref
  9222. reference "tab:methyl-est-nonnull"
  9223. plural "false"
  9224. caps "false"
  9225. noprefix "false"
  9226. \end_inset
  9227. were generated.
  9228. The distributions for analysis A all have a dip in density near zero, which
  9229. is a strong sign of a poor model fit.
  9230. The histograms for analyses B and C are more well-behaved, with a uniform
  9231. component stretching all the way from 0 to 1 representing the probes for
  9232. which the null hypotheses is true (no differential methylation), and a
  9233. zero-biased component representing the probes for which the null hypothesis
  9234. is false (differentially methylated).
  9235. These histograms do not indicate any major issues with the model fit.
  9236. \end_layout
  9237. \begin_layout Standard
  9238. \begin_inset Flex TODO Note (inline)
  9239. status open
  9240. \begin_layout Plain Layout
  9241. If time allows, maybe generate the PCA plots before/after SVA effect subtraction
  9242. ?
  9243. \end_layout
  9244. \end_inset
  9245. \end_layout
  9246. \begin_layout Section
  9247. Discussion
  9248. \end_layout
  9249. \begin_layout Subsection
  9250. fRMA achieves clinically applicable normalization without sacrificing classifica
  9251. tion performance
  9252. \end_layout
  9253. \begin_layout Standard
  9254. As shown in Figure
  9255. \begin_inset CommandInset ref
  9256. LatexCommand ref
  9257. reference "fig:Classifier-probabilities-RMA"
  9258. plural "false"
  9259. caps "false"
  9260. noprefix "false"
  9261. \end_inset
  9262. , improper normalization, particularly separate normalization of training
  9263. and test samples, leads to unwanted biases in classification.
  9264. In a controlled experimental context, it is always possible to correct
  9265. this issue by normalizing all experimental samples together.
  9266. However, because it is not feasible to normalize all samples together in
  9267. a clinical context, a single-channel normalization is required is required.
  9268. \end_layout
  9269. \begin_layout Standard
  9270. The major concern in using a single-channel normalization is that non-single-cha
  9271. nnel methods can share information between arrays to improve the normalization,
  9272. and single-channel methods risk sacrificing the gains in normalization
  9273. accuracy that come from this information sharing.
  9274. In the case of RMA, this information sharing is accomplished through quantile
  9275. normalization and median polish steps.
  9276. The need for information sharing in quantile normalization can easily be
  9277. removed by learning a fixed set of quantiles from external data and normalizing
  9278. each array to these fixed quantiles, instead of the quantiles of the data
  9279. itself.
  9280. As long as the fixed quantiles are reasonable, the result will be similar
  9281. to standard RMA.
  9282. However, there is no analogous way to eliminate cross-array information
  9283. sharing in the median polish step, so fRMA replaces this with a weighted
  9284. average of probes on each array, with the weights learned from external
  9285. data.
  9286. This step of fRMA has the greatest potential to diverge from RMA un undesirable
  9287. ways.
  9288. \end_layout
  9289. \begin_layout Standard
  9290. However, when run on real data, fRMA performed at least as well as RMA in
  9291. both the internal validation and external validation tests.
  9292. This shows that fRMA can be used to normalize individual clinical samples
  9293. in a class prediction context without sacrificing the classifier performance
  9294. that would be obtained by using the more well-established RMA for normalization.
  9295. The other single-channel normalization method considered, SCAN, showed
  9296. some loss of AUC in the external validation test.
  9297. Based on these results, fRMA is the preferred normalization for clinical
  9298. samples in a class prediction context.
  9299. \end_layout
  9300. \begin_layout Subsection
  9301. Robust fRMA vectors can be generated for new array platforms
  9302. \end_layout
  9303. \begin_layout Standard
  9304. \begin_inset Flex TODO Note (inline)
  9305. status open
  9306. \begin_layout Plain Layout
  9307. Look up the exact numbers, do a find & replace for
  9308. \begin_inset Quotes eld
  9309. \end_inset
  9310. 850
  9311. \begin_inset Quotes erd
  9312. \end_inset
  9313. \end_layout
  9314. \end_inset
  9315. \end_layout
  9316. \begin_layout Standard
  9317. The published fRMA normalization vectors for the hgu133plus2 platform were
  9318. generated from a set of about 850 samples chosen from a wide range of tissues,
  9319. which the authors determined was sufficient to generate a robust set of
  9320. normalization vectors that could be applied across all tissues
  9321. \begin_inset CommandInset citation
  9322. LatexCommand cite
  9323. key "McCall2010"
  9324. literal "false"
  9325. \end_inset
  9326. .
  9327. Since we only had hthgu133pluspm for 2 tissues of interest, our needs were
  9328. more modest.
  9329. Even using only 130 samples in 26 batches of 5 samples each for kidney
  9330. biopsies, we were able to train a robust set of fRMA normalization vectors
  9331. that were not meaningfully affected by the random selection of 5 samples
  9332. from each batch.
  9333. As expected, the training process was just as robust for the blood samples
  9334. with 230 samples in 46 batches of 5 samples each.
  9335. Because these vectors were each generated using training samples from a
  9336. single tissue, they are not suitable for general use, unlike the vectors
  9337. provided with fRMA itself.
  9338. They are purpose-built for normalizing a specific type of sample on a specific
  9339. platform.
  9340. This is a mostly acceptable limitation in the context of developing a machine
  9341. learning classifier for diagnosing a disease based on samples of a specific
  9342. tissue.
  9343. \end_layout
  9344. \begin_layout Standard
  9345. \begin_inset Flex TODO Note (inline)
  9346. status open
  9347. \begin_layout Plain Layout
  9348. Talk about how these vectors can be used for any data from these tissues
  9349. on this platform even though they were custom made for this data set.
  9350. \end_layout
  9351. \end_inset
  9352. \end_layout
  9353. \begin_layout Standard
  9354. \begin_inset Flex TODO Note (inline)
  9355. status open
  9356. \begin_layout Plain Layout
  9357. How to bring up that these custom vectors were used in another project by
  9358. someone else that was never published?
  9359. \end_layout
  9360. \end_inset
  9361. \end_layout
  9362. \begin_layout Subsection
  9363. Methylation array data can be successfully analyzed using existing techniques,
  9364. but machine learning poses additional challenges
  9365. \end_layout
  9366. \begin_layout Standard
  9367. Both analysis strategies B and C both yield a reasonable analysis, with
  9368. a mean-variance trend that matches the expected behavior for the non-linear
  9369. M-value transformation (Figure
  9370. \begin_inset CommandInset ref
  9371. LatexCommand ref
  9372. reference "fig:meanvar-sva-aw"
  9373. plural "false"
  9374. caps "false"
  9375. noprefix "false"
  9376. \end_inset
  9377. ) and well-behaved p-value distributions (Figure
  9378. \begin_inset CommandInset ref
  9379. LatexCommand ref
  9380. reference "fig:meth-p-value-histograms"
  9381. plural "false"
  9382. caps "false"
  9383. noprefix "false"
  9384. \end_inset
  9385. ).
  9386. These two analyses also yield similar numbers of significant probes (Table
  9387. \begin_inset CommandInset ref
  9388. LatexCommand ref
  9389. reference "tab:methyl-num-signif"
  9390. plural "false"
  9391. caps "false"
  9392. noprefix "false"
  9393. \end_inset
  9394. ) and similar estimates of the number of differentially methylated probes
  9395. (Table
  9396. \begin_inset CommandInset ref
  9397. LatexCommand ref
  9398. reference "tab:methyl-est-nonnull"
  9399. plural "false"
  9400. caps "false"
  9401. noprefix "false"
  9402. \end_inset
  9403. ).
  9404. The main difference between these two analyses is the method used to account
  9405. for the mean-variance trend.
  9406. In analysis B, the trend is estimated and applied at the probe level: each
  9407. probe's estimated variance is squeezed toward the trend using an empirical
  9408. Bayes procedure (Figure
  9409. \begin_inset CommandInset ref
  9410. LatexCommand ref
  9411. reference "fig:meanvar-sva-aw"
  9412. plural "false"
  9413. caps "false"
  9414. noprefix "false"
  9415. \end_inset
  9416. ).
  9417. In analysis C, the trend is still estimated at the probe level, but instead
  9418. of estimating a single variance value shared across all observations for
  9419. a given probe, the voom method computes an initial estimate of the variance
  9420. for each observation individually based on where its model-fitted M-value
  9421. falls on the trend line and then assigns inverse-variance weights to model
  9422. the difference in variance between observations.
  9423. An overall variance is still estimated for each probe using the same empirical
  9424. Bayes method, but now the residual trend is flat (Figure
  9425. \begin_inset CommandInset ref
  9426. LatexCommand ref
  9427. reference "fig:meanvar-sva-voomaw"
  9428. plural "false"
  9429. caps "false"
  9430. noprefix "false"
  9431. \end_inset
  9432. ), indicating that the mean-variance trend is adequately modeled by scaling
  9433. the estimated variance for each observation using the weights computed
  9434. by voom.
  9435. \end_layout
  9436. \begin_layout Standard
  9437. The difference between the standard empirical Bayes trended variance modeling
  9438. (analysis B) and voom (analysis C) is analogous to the difference between
  9439. a t-test with equal variance and a t-test with unequal variance, except
  9440. that the unequal group variances used in the latter test are estimated
  9441. based on the mean-variance trend from all the probes rather than the data
  9442. for the specific probe being tested, thus stabilizing the group variance
  9443. estimates by sharing information between probes.
  9444. Allowing voom to model the variance using observation weights in this manner
  9445. allows the linear model fit to concentrate statistical power where it will
  9446. do the most good.
  9447. For example, if a particular probe's M-values are always at the extreme
  9448. of the M-value range (e.g.
  9449. less than -4) for ADNR samples, but the M-values for that probe in TX and
  9450. CAN samples are within the flat region of the mean-variance trend (between
  9451. -3 and +3), voom is able to down-weight the contribution of the high-variance
  9452. M-values from the ADNR samples in order to gain more statistical power
  9453. while testing for differential methylation between TX and CAN.
  9454. In contrast, modeling the mean-variance trend only at the probe level would
  9455. combine the high-variance ADNR samples and lower-variance samples from
  9456. other conditions and estimate an intermediate variance for this probe.
  9457. In practice, analysis B shows that this approach is adequate, but the voom
  9458. approach in analysis C is at least as good on all model fit criteria and
  9459. yields a larger estimate for the number of differentially methylated genes,
  9460. \emph on
  9461. and
  9462. \emph default
  9463. it matches up better with the theoretical
  9464. \end_layout
  9465. \begin_layout Standard
  9466. The significant association of diabetes diagnosis with sample quality is
  9467. interesting.
  9468. The samples with Type 2 diabetes tended to have more variation, averaged
  9469. across all probes, than those with Type 1 diabetes.
  9470. This is consistent with the consensus that type 2 diabetes and the associated
  9471. metabolic syndrome represent a broad dysregulation of the body's endocrine
  9472. signaling related to metabolism [citation needed].
  9473. This dysregulation could easily manifest as a greater degree of variation
  9474. in the DNA methylation patterns of affected tissues.
  9475. In contrast, Type 1 diabetes has a more specific cause and effect, so a
  9476. less variable methylation signature is expected.
  9477. \end_layout
  9478. \begin_layout Standard
  9479. This preliminary analysis suggests that some degree of differential methylation
  9480. exists between TX and each of the three types of transplant disfunction
  9481. studied.
  9482. Hence, it may be feasible to train a classifier to diagnose transplant
  9483. disfunction from DNA methylation array data.
  9484. However, the major importance of both SVA and sample quality weighting
  9485. for proper modeling of this data poses significant challenges for any attempt
  9486. at a machine learning on data of similar quality.
  9487. While these are easily used in a modeling context with full sample information,
  9488. neither of these methods is directly applicable in a machine learning context,
  9489. where the diagnosis is not known ahead of time.
  9490. If a machine learning approach for methylation-based diagnosis is to be
  9491. pursued, it will either require machine-learning-friendly methods to address
  9492. the same systematic trends in the data that SVA and sample quality weighting
  9493. address, or it will require higher quality data with substantially less
  9494. systematic perturbation of the data.
  9495. \end_layout
  9496. \begin_layout Section
  9497. Future Directions
  9498. \end_layout
  9499. \begin_layout Standard
  9500. \begin_inset Flex TODO Note (inline)
  9501. status open
  9502. \begin_layout Plain Layout
  9503. Some work was already being done with the existing fRMA vectors.
  9504. Do I mention that here?
  9505. \end_layout
  9506. \end_inset
  9507. \end_layout
  9508. \begin_layout Subsection
  9509. Improving fRMA to allow training from batches of unequal size
  9510. \end_layout
  9511. \begin_layout Standard
  9512. Because the tools for building fRMA normalization vectors require equal-size
  9513. batches, many samples must be discarded from the training data.
  9514. This is undesirable for a few reasons.
  9515. First, more data is simply better, all other things being equal.
  9516. In this case,
  9517. \begin_inset Quotes eld
  9518. \end_inset
  9519. better
  9520. \begin_inset Quotes erd
  9521. \end_inset
  9522. means a more precise estimate of normalization parameters.
  9523. In addition, the samples to be discarded must be chosen arbitrarily, which
  9524. introduces an unnecessary element of randomness into the estimation process.
  9525. While the randomness can be made deterministic by setting a consistent
  9526. random seed, the need for equal size batches also introduces a need for
  9527. the analyst to decide on the appropriate trade-off between batch size and
  9528. the number of batches.
  9529. This introduces an unnecessary and undesirable
  9530. \begin_inset Quotes eld
  9531. \end_inset
  9532. researcher degree of freedom
  9533. \begin_inset Quotes erd
  9534. \end_inset
  9535. into the analysis, since the generated normalization vectors now depend
  9536. on the choice of batch size based on vague selection criteria and instinct,
  9537. which can unintentionally introduce bias if the researcher chooses a batch
  9538. size based on what seems to yield the most favorable downstream results
  9539. \begin_inset CommandInset citation
  9540. LatexCommand cite
  9541. key "Simmons2011"
  9542. literal "false"
  9543. \end_inset
  9544. .
  9545. \end_layout
  9546. \begin_layout Standard
  9547. Fortunately, the requirement for equal-size batches is not inherent to the
  9548. fRMA algorithm but rather a limitation of the implementation in the frmaTools
  9549. package.
  9550. In personal communication, the package's author, Matthew McCall, has indicated
  9551. that with some work, it should be possible to improve the implementation
  9552. to work with batches of unequal sizes.
  9553. The current implementation ignores the batch size when calculating with-batch
  9554. and between-batch residual variances, since the batch size constant cancels
  9555. out later in the calculations as long as all batches are of equal size.
  9556. Hence, the calculations of these parameters would need to be modified to
  9557. remove this optimization and properly calculate the variances using the
  9558. full formula.
  9559. Once this modification is made, a new strategy would need to be developed
  9560. for assessing the stability of parameter estimates, since the random subsamplin
  9561. g step is eliminated, meaning that different subsamplings can no longer
  9562. be compared as in Figures
  9563. \begin_inset CommandInset ref
  9564. LatexCommand ref
  9565. reference "fig:frma-violin"
  9566. plural "false"
  9567. caps "false"
  9568. noprefix "false"
  9569. \end_inset
  9570. and
  9571. \begin_inset CommandInset ref
  9572. LatexCommand ref
  9573. reference "fig:Representative-MA-plots"
  9574. plural "false"
  9575. caps "false"
  9576. noprefix "false"
  9577. \end_inset
  9578. .
  9579. Bootstrap resampling is likely a good candidate here: sample many training
  9580. sets of equal size from the existing training set with replacement, estimate
  9581. parameters from each resampled training set, and compare the estimated
  9582. parameters between bootstraps in order to quantify the variability in each
  9583. parameter's estimation.
  9584. \end_layout
  9585. \begin_layout Subsection
  9586. Developing methylation arrays as a diagnostic tool for kidney transplant
  9587. rejection
  9588. \end_layout
  9589. \begin_layout Standard
  9590. The current study has showed that DNA methylation, as assayed by Illumina
  9591. 450k methylation arrays, has some potential for diagnosing transplant dysfuncti
  9592. ons, including rejection.
  9593. However, very few probes could be confidently identified as differentially
  9594. methylated between healthy and dysfunctional transplants.
  9595. One likely explanation for this is the predominant influence of unobserved
  9596. confounding factors.
  9597. SVA can model and correct for such factors, but the correction can never
  9598. be perfect, so some degree of unwanted systematic variation will always
  9599. remain after SVA correction.
  9600. If the effect size of the confounding factors was similar to that of the
  9601. factor of interest (in this case, transplant status), this would be an
  9602. acceptable limitation, since removing most of the confounding factors'
  9603. effects would allow the main effect to stand out.
  9604. However, in this data set, the confounding factors have a much larger effect
  9605. size than transplant status, which means that the small degree of remaining
  9606. variation not removed by SVA can still swamp the effect of interest, making
  9607. it difficult to detect.
  9608. This is, of course, a major issue when the end goal is to develop a classifier
  9609. to diagnose transplant rejection from methylation data, since batch-correction
  9610. methods like SVA that work in a linear modeling context cannot be applied
  9611. in a machine learning context.
  9612. \end_layout
  9613. \begin_layout Standard
  9614. Currently, the source of these unwanted systematic variations in the data
  9615. is unknown.
  9616. The best solution would be to determine the cause of the variation and
  9617. eliminate it, thereby eliminating the need to model and remove that variation.
  9618. However, if this proves impractical, another option is to use SVA to identify
  9619. probes that are highly associated with the surrogate variables that describe
  9620. the unwanted variation in the data.
  9621. These probes could be discarded prior to classifier training, in order
  9622. to maximize the chance that the training algorithm will be able to identify
  9623. highly predictive probes from those remaining.
  9624. Lastly, it is possible that some of this unwanted variation is a result
  9625. of the array-based assay being used and would be eliminated by switching
  9626. to assaying DNA methylation using bisulphite sequencing.
  9627. However, this carries the risk that the sequencing assay will have its
  9628. own set of biases that must be corrected for in a different way.
  9629. \end_layout
  9630. \begin_layout Chapter
  9631. Globin-blocking for more effective blood RNA-seq analysis in primate animal
  9632. model
  9633. \end_layout
  9634. \begin_layout Standard
  9635. \begin_inset Flex TODO Note (inline)
  9636. status open
  9637. \begin_layout Plain Layout
  9638. Choose between above and the paper title: Optimizing yield of deep RNA sequencin
  9639. g for gene expression profiling by globin reduction of peripheral blood
  9640. samples from cynomolgus monkeys (Macaca fascicularis).
  9641. \end_layout
  9642. \end_inset
  9643. \end_layout
  9644. \begin_layout Standard
  9645. \begin_inset Flex TODO Note (inline)
  9646. status open
  9647. \begin_layout Plain Layout
  9648. Chapter author list:
  9649. \begin_inset CommandInset href
  9650. LatexCommand href
  9651. target "https://tex.stackexchange.com/questions/156862/displaying-author-for-each-chapter-in-book"
  9652. \end_inset
  9653. Every chapter gets an author list, which may or may not be part of a citation
  9654. to a published/preprinted paper.
  9655. \end_layout
  9656. \end_inset
  9657. \end_layout
  9658. \begin_layout Standard
  9659. \begin_inset Flex TODO Note (inline)
  9660. status open
  9661. \begin_layout Plain Layout
  9662. Preprint then cite the paper
  9663. \end_layout
  9664. \end_inset
  9665. \end_layout
  9666. \begin_layout Section*
  9667. Abstract
  9668. \end_layout
  9669. \begin_layout Paragraph
  9670. Background
  9671. \end_layout
  9672. \begin_layout Standard
  9673. Primate blood contains high concentrations of globin messenger RNA.
  9674. Globin reduction is a standard technique used to improve the expression
  9675. results obtained by DNA microarrays on RNA from blood samples.
  9676. However, with whole transcriptome RNA-sequencing (RNA-seq) quickly replacing
  9677. microarrays for many applications, the impact of globin reduction for RNA-seq
  9678. has not been previously studied.
  9679. Moreover, no off-the-shelf kits are available for globin reduction in nonhuman
  9680. primates.
  9681. \end_layout
  9682. \begin_layout Paragraph
  9683. Results
  9684. \end_layout
  9685. \begin_layout Standard
  9686. Here we report a protocol for RNA-seq in primate blood samples that uses
  9687. complimentary oligonucleotides to block reverse transcription of the alpha
  9688. and beta globin genes.
  9689. In test samples from cynomolgus monkeys (Macaca fascicularis), this globin
  9690. blocking protocol approximately doubles the yield of informative (non-globin)
  9691. reads by greatly reducing the fraction of globin reads, while also improving
  9692. the consistency in sequencing depth between samples.
  9693. The increased yield enables detection of about 2000 more genes, significantly
  9694. increases the correlation in measured gene expression levels between samples,
  9695. and increases the sensitivity of differential gene expression tests.
  9696. \end_layout
  9697. \begin_layout Paragraph
  9698. Conclusions
  9699. \end_layout
  9700. \begin_layout Standard
  9701. These results show that globin blocking significantly improves the cost-effectiv
  9702. eness of mRNA sequencing in primate blood samples by doubling the yield
  9703. of useful reads, allowing detection of more genes, and improving the precision
  9704. of gene expression measurements.
  9705. Based on these results, a globin reducing or blocking protocol is recommended
  9706. for all RNA-seq studies of primate blood samples.
  9707. \end_layout
  9708. \begin_layout Section
  9709. Approach
  9710. \end_layout
  9711. \begin_layout Standard
  9712. \begin_inset Note Note
  9713. status open
  9714. \begin_layout Plain Layout
  9715. Consider putting some of this in the Intro chapter
  9716. \end_layout
  9717. \begin_layout Itemize
  9718. Cynomolgus monkeys as a model organism
  9719. \end_layout
  9720. \begin_deeper
  9721. \begin_layout Itemize
  9722. Highly related to humans
  9723. \end_layout
  9724. \begin_layout Itemize
  9725. Small size and short life cycle - good research animal
  9726. \end_layout
  9727. \begin_layout Itemize
  9728. Genomics resources still in development
  9729. \end_layout
  9730. \end_deeper
  9731. \begin_layout Itemize
  9732. Inadequacy of existing blood RNA-seq protocols
  9733. \end_layout
  9734. \begin_deeper
  9735. \begin_layout Itemize
  9736. Existing protocols use a separate globin pulldown step, slowing down processing
  9737. \end_layout
  9738. \end_deeper
  9739. \end_inset
  9740. \end_layout
  9741. \begin_layout Standard
  9742. Increasingly, researchers are turning to high-throughput mRNA sequencing
  9743. technologies (RNA-seq) in preference to expression microarrays for analysis
  9744. of gene expression
  9745. \begin_inset CommandInset citation
  9746. LatexCommand cite
  9747. key "Mutz2012"
  9748. literal "false"
  9749. \end_inset
  9750. .
  9751. The advantages are even greater for study of model organisms with no well-estab
  9752. lished array platforms available, such as the cynomolgus monkey (Macaca
  9753. fascicularis).
  9754. High fractions of globin mRNA are naturally present in mammalian peripheral
  9755. blood samples (up to 70% of total mRNA) and these are known to interfere
  9756. with the results of array-based expression profiling
  9757. \begin_inset CommandInset citation
  9758. LatexCommand cite
  9759. key "Winn2010"
  9760. literal "false"
  9761. \end_inset
  9762. .
  9763. The importance of globin reduction for RNA-seq of blood has only been evaluated
  9764. for a deepSAGE protocol on human samples
  9765. \begin_inset CommandInset citation
  9766. LatexCommand cite
  9767. key "Mastrokolias2012"
  9768. literal "false"
  9769. \end_inset
  9770. .
  9771. In the present report, we evaluated globin reduction using custom blocking
  9772. oligonucleotides for deep RNA-seq of peripheral blood samples from a nonhuman
  9773. primate, cynomolgus monkey, using the Illumina technology platform.
  9774. We demonstrate that globin reduction significantly improves the cost-effectiven
  9775. ess of RNA-seq in blood samples.
  9776. Thus, our protocol offers a significant advantage to any investigator planning
  9777. to use RNA-seq for gene expression profiling of nonhuman primate blood
  9778. samples.
  9779. Our method can be generally applied to any species by designing complementary
  9780. oligonucleotide blocking probes to the globin gene sequences of that species.
  9781. Indeed, any highly expressed but biologically uninformative transcripts
  9782. can also be blocked to further increase sequencing efficiency and value
  9783. \begin_inset CommandInset citation
  9784. LatexCommand cite
  9785. key "Arnaud2016"
  9786. literal "false"
  9787. \end_inset
  9788. .
  9789. \end_layout
  9790. \begin_layout Section
  9791. Methods
  9792. \end_layout
  9793. \begin_layout Subsection
  9794. Sample collection
  9795. \end_layout
  9796. \begin_layout Standard
  9797. All research reported here was done under IACUC-approved protocols at the
  9798. University of Miami and complied with all applicable federal and state
  9799. regulations and ethical principles for nonhuman primate research.
  9800. Blood draws occurred between 16 April 2012 and 18 June 2015.
  9801. The experimental system involved intrahepatic pancreatic islet transplantation
  9802. into Cynomolgus monkeys with induced diabetes mellitus with or without
  9803. concomitant infusion of mesenchymal stem cells.
  9804. Blood was collected at serial time points before and after transplantation
  9805. into PAXgene Blood RNA tubes (PreAnalytiX/Qiagen, Valencia, CA) at the
  9806. precise volume:volume ratio of 2.5 ml whole blood into 6.9 ml of PAX gene
  9807. additive.
  9808. \end_layout
  9809. \begin_layout Subsection
  9810. Globin Blocking
  9811. \end_layout
  9812. \begin_layout Standard
  9813. Four oligonucleotides were designed to hybridize to the 3’ end of the transcript
  9814. s for Cynomolgus HBA1, HBA2 and HBB, with two hybridization sites for HBB
  9815. and 2 sites for HBA (the chosen sites were identical in both HBA genes).
  9816. All oligos were purchased from Sigma and were entirely composed of 2’O-Me
  9817. bases with a C3 spacer positioned at the 3’ ends to prevent any polymerase
  9818. mediated primer extension.
  9819. \end_layout
  9820. \begin_layout Quote
  9821. HBA1/2 site 1: GCCCACUCAGACUUUAUUCAAAG-C3spacer
  9822. \end_layout
  9823. \begin_layout Quote
  9824. HBA1/2 site 2: GGUGCAAGGAGGGGAGGAG-C3spacer
  9825. \end_layout
  9826. \begin_layout Quote
  9827. HBB site 1: AAUGAAAAUAAAUGUUUUUUAUUAG-C3spacer
  9828. \end_layout
  9829. \begin_layout Quote
  9830. HBB site 2: CUCAAGGCCCUUCAUAAUAUCCC-C3spacer
  9831. \end_layout
  9832. \begin_layout Subsection
  9833. RNA-seq Library Preparation
  9834. \end_layout
  9835. \begin_layout Standard
  9836. \begin_inset Flex TODO Note (inline)
  9837. status open
  9838. \begin_layout Plain Layout
  9839. Add protected spaces where appropriate to prevent unwanted line breaks.
  9840. \end_layout
  9841. \end_inset
  9842. \end_layout
  9843. \begin_layout Standard
  9844. Sequencing libraries were prepared with 200
  9845. \begin_inset space ~
  9846. \end_inset
  9847. ng total RNA from each sample.
  9848. Polyadenylated mRNA was selected from 200 ng aliquots of cynomolgus blood-deriv
  9849. ed total RNA using Ambion Dynabeads Oligo(dT)25 beads (Invitrogen) following
  9850. manufacturer’s recommended protocol.
  9851. PolyA selected RNA was then combined with 8 pmol of HBA1/2 (site 1), 8
  9852. pmol of HBA1/2 (site 2), 12 pmol of HBB (site 1) and 12 pmol of HBB (site
  9853. 2) oligonucleotides.
  9854. In addition, 20 pmol of RT primer containing a portion of the Illumina
  9855. adapter sequence (B-oligo-dTV: GAGTTCCTTGGCACCCGAGAATTCCATTTTTTTTTTTTTTTTTTTV)
  9856. and 4 µL of 5X First Strand buffer (250 mM Tris-HCl pH 8.3, 375 mM KCl,
  9857. 15mM MgCl2) were added in a total volume of 15 µL.
  9858. The RNA was fragmented by heating this cocktail for 3 minutes at 95°C and
  9859. then placed on ice.
  9860. This was followed by the addition of 2 µL 0.1 M DTT, 1 µL RNaseOUT, 1 µL
  9861. 10mM dNTPs 10% biotin-16 aminoallyl-2’- dUTP and 10% biotin-16 aminoallyl-2’-
  9862. dCTP (TriLink Biotech, San Diego, CA), 1 µL Superscript II (200U/ µL, Thermo-Fi
  9863. sher).
  9864. A second “unblocked” library was prepared in the same way for each sample
  9865. but replacing the blocking oligos with an equivalent volume of water.
  9866. The reaction was carried out at 25°C for 15 minutes and 42°C for 40 minutes,
  9867. followed by incubation at 75°C for 10 minutes to inactivate the reverse
  9868. transcriptase.
  9869. \end_layout
  9870. \begin_layout Standard
  9871. The cDNA/RNA hybrid molecules were purified using 1.8X Ampure XP beads (Agencourt
  9872. ) following supplier’s recommended protocol.
  9873. The cDNA/RNA hybrid was eluted in 25 µL of 10 mM Tris-HCl pH 8.0, and then
  9874. bound to 25 µL of M280 Magnetic Streptavidin beads washed per recommended
  9875. protocol (Thermo-Fisher).
  9876. After 30 minutes of binding, beads were washed one time in 100 µL 0.1N NaOH
  9877. to denature and remove the bound RNA, followed by two 100 µL washes with
  9878. 1X TE buffer.
  9879. \end_layout
  9880. \begin_layout Standard
  9881. Subsequent attachment of the 5-prime Illumina A adapter was performed by
  9882. on-bead random primer extension of the following sequence (A-N8 primer:
  9883. TTCAGAGTTCTACAGTCCGACGATCNNNNNNNN).
  9884. Briefly, beads were resuspended in a 20 µL reaction containing 5 µM A-N8
  9885. primer, 40mM Tris-HCl pH 7.5, 20mM MgCl2, 50mM NaCl, 0.325U/µL Sequenase
  9886. 2.0 (Affymetrix, Santa Clara, CA), 0.0025U/µL inorganic pyrophosphatase (Affymetr
  9887. ix) and 300 µM each dNTP.
  9888. Reaction was incubated at 22°C for 30 minutes, then beads were washed 2
  9889. times with 1X TE buffer (200µL).
  9890. \end_layout
  9891. \begin_layout Standard
  9892. The magnetic streptavidin beads were resuspended in 34 µL nuclease-free
  9893. water and added directly to a PCR tube.
  9894. The two Illumina protocol-specified PCR primers were added at 0.53 µM (Illumina
  9895. TruSeq Universal Primer 1 and Illumina TruSeq barcoded PCR primer 2), along
  9896. with 40 µL 2X KAPA HiFi Hotstart ReadyMix (KAPA, Willmington MA) and thermocycl
  9897. ed as follows: starting with 98°C (2 min-hold); 15 cycles of 98°C, 20sec;
  9898. 60°C, 30sec; 72°C, 30sec; and finished with a 72°C (2 min-hold).
  9899. \end_layout
  9900. \begin_layout Standard
  9901. PCR products were purified with 1X Ampure Beads following manufacturer’s
  9902. recommended protocol.
  9903. Libraries were then analyzed using the Agilent TapeStation and quantitation
  9904. of desired size range was performed by “smear analysis”.
  9905. Samples were pooled in equimolar batches of 16 samples.
  9906. Pooled libraries were size selected on 2% agarose gels (E-Gel EX Agarose
  9907. Gels; Thermo-Fisher).
  9908. Products were cut between 250 and 350 bp (corresponding to insert sizes
  9909. of 130 to 230 bps).
  9910. Finished library pools were then sequenced on the Illumina NextSeq500 instrumen
  9911. t with 75 base read lengths.
  9912. \end_layout
  9913. \begin_layout Subsection
  9914. Read alignment and counting
  9915. \end_layout
  9916. \begin_layout Standard
  9917. Reads were aligned to the cynomolgus genome using STAR
  9918. \begin_inset CommandInset citation
  9919. LatexCommand cite
  9920. key "Dobin2013,Wilson2013"
  9921. literal "false"
  9922. \end_inset
  9923. .
  9924. Counts of uniquely mapped reads were obtained for every gene in each sample
  9925. with the
  9926. \begin_inset Flex Code
  9927. status open
  9928. \begin_layout Plain Layout
  9929. featureCounts
  9930. \end_layout
  9931. \end_inset
  9932. function from the
  9933. \begin_inset Flex Code
  9934. status open
  9935. \begin_layout Plain Layout
  9936. Rsubread
  9937. \end_layout
  9938. \end_inset
  9939. package, using each of the three possibilities for the
  9940. \begin_inset Flex Code
  9941. status open
  9942. \begin_layout Plain Layout
  9943. strandSpecific
  9944. \end_layout
  9945. \end_inset
  9946. option: sense, antisense, and unstranded
  9947. \begin_inset CommandInset citation
  9948. LatexCommand cite
  9949. key "Liao2014"
  9950. literal "false"
  9951. \end_inset
  9952. .
  9953. A few artifacts in the cynomolgus genome annotation complicated read counting.
  9954. First, no ortholog is annotated for alpha globin in the cynomolgus genome,
  9955. presumably because the human genome has two alpha globin genes with nearly
  9956. identical sequences, making the orthology relationship ambiguous.
  9957. However, two loci in the cynomolgus genome are as “hemoglobin subunit alpha-lik
  9958. e” (LOC102136192 and LOC102136846).
  9959. LOC102136192 is annotated as a pseudogene while LOC102136846 is annotated
  9960. as protein-coding.
  9961. Our globin reduction protocol was designed to include blocking of these
  9962. two genes.
  9963. Indeed, these two genes have almost the same read counts in each library
  9964. as the properly-annotated HBB gene and much larger counts than any other
  9965. gene in the unblocked libraries, giving confidence that reads derived from
  9966. the real alpha globin are mapping to both genes.
  9967. Thus, reads from both of these loci were counted as alpha globin reads
  9968. in all further analyses.
  9969. The second artifact is a small, uncharacterized non-coding RNA gene (LOC1021365
  9970. 91), which overlaps the HBA-like gene (LOC102136192) on the opposite strand.
  9971. If counting is not performed in stranded mode (or if a non-strand-specific
  9972. sequencing protocol is used), many reads mapping to the globin gene will
  9973. be discarded as ambiguous due to their overlap with this ncRNA gene, resulting
  9974. in significant undercounting of globin reads.
  9975. Therefore, stranded sense counts were used for all further analysis in
  9976. the present study to insure that we accurately accounted for globin transcript
  9977. reduction.
  9978. However, we note that stranded reads are not necessary for RNA-seq using
  9979. our protocol in standard practice.
  9980. \end_layout
  9981. \begin_layout Subsection
  9982. Normalization and Exploratory Data Analysis
  9983. \end_layout
  9984. \begin_layout Standard
  9985. Libraries were normalized by computing scaling factors using the
  9986. \begin_inset Flex Code
  9987. status open
  9988. \begin_layout Plain Layout
  9989. edgeR
  9990. \end_layout
  9991. \end_inset
  9992. package’s Trimmed Mean of M-values method
  9993. \begin_inset CommandInset citation
  9994. LatexCommand cite
  9995. key "Robinson2010"
  9996. literal "false"
  9997. \end_inset
  9998. .
  9999. Log2 counts per million values (logCPM) were calculated using the cpm function
  10000. in
  10001. \begin_inset Flex Code
  10002. status open
  10003. \begin_layout Plain Layout
  10004. edgeR
  10005. \end_layout
  10006. \end_inset
  10007. for individual samples and aveLogCPM function for averages across groups
  10008. of samples, using those functions’ default prior count values to avoid
  10009. taking the logarithm of 0.
  10010. Genes were considered “present” if their average normalized logCPM values
  10011. across all libraries were at least -1.
  10012. Normalizing for gene length was unnecessary because the sequencing protocol
  10013. is 3’-biased and hence the expected read count for each gene is related
  10014. to the transcript’s copy number but not its length.
  10015. \end_layout
  10016. \begin_layout Standard
  10017. In order to assess the effect of blocking on reproducibility, Pearson and
  10018. Spearman correlation coefficients were computed between the logCPM values
  10019. for every pair of libraries within the globin-blocked (GB) and unblocked
  10020. (non-GB) groups, and
  10021. \begin_inset Flex Code
  10022. status open
  10023. \begin_layout Plain Layout
  10024. edgeR
  10025. \end_layout
  10026. \end_inset
  10027. 's
  10028. \begin_inset Flex Code
  10029. status open
  10030. \begin_layout Plain Layout
  10031. estimateDisp
  10032. \end_layout
  10033. \end_inset
  10034. function was used to compute negative binomial dispersions separately for
  10035. the two groups
  10036. \begin_inset CommandInset citation
  10037. LatexCommand cite
  10038. key "Chen2014"
  10039. literal "false"
  10040. \end_inset
  10041. .
  10042. \end_layout
  10043. \begin_layout Subsection
  10044. Differential Expression Analysis
  10045. \end_layout
  10046. \begin_layout Standard
  10047. All tests for differential gene expression were performed using
  10048. \begin_inset Flex Code
  10049. status open
  10050. \begin_layout Plain Layout
  10051. edgeR
  10052. \end_layout
  10053. \end_inset
  10054. , by first fitting a negative binomial generalized linear model to the counts
  10055. and normalization factors and then performing a quasi-likelihood F-test
  10056. with robust estimation of outlier gene dispersions
  10057. \begin_inset CommandInset citation
  10058. LatexCommand cite
  10059. key "Lund2012,Phipson2016"
  10060. literal "false"
  10061. \end_inset
  10062. .
  10063. To investigate the effects of globin blocking on each gene, an additive
  10064. model was fit to the full data with coefficients for globin blocking and
  10065. SampleID.
  10066. To test the effect of globin blocking on detection of differentially expressed
  10067. genes, the GB samples and non-GB samples were each analyzed independently
  10068. as follows: for each animal with both a pre-transplant and a post-transplant
  10069. time point in the data set, the pre-transplant sample and the earliest
  10070. post-transplant sample were selected, and all others were excluded, yielding
  10071. a pre-/post-transplant pair of samples for each animal (N=7 animals with
  10072. paired samples).
  10073. These samples were analyzed for pre-transplant vs.
  10074. post-transplant differential gene expression while controlling for inter-animal
  10075. variation using an additive model with coefficients for transplant and
  10076. animal ID.
  10077. In all analyses, p-values were adjusted using the Benjamini-Hochberg procedure
  10078. for FDR control
  10079. \begin_inset CommandInset citation
  10080. LatexCommand cite
  10081. key "Benjamini1995"
  10082. literal "false"
  10083. \end_inset
  10084. .
  10085. \end_layout
  10086. \begin_layout Standard
  10087. \begin_inset Note Note
  10088. status open
  10089. \begin_layout Itemize
  10090. New blood RNA-seq protocol to block reverse transcription of globin genes
  10091. \end_layout
  10092. \begin_layout Itemize
  10093. Blood RNA-seq time course after transplants with/without MSC infusion
  10094. \end_layout
  10095. \end_inset
  10096. \end_layout
  10097. \begin_layout Section
  10098. Results
  10099. \end_layout
  10100. \begin_layout Subsection
  10101. Globin blocking yields a larger and more consistent fraction of useful reads
  10102. \end_layout
  10103. \begin_layout Standard
  10104. \begin_inset ERT
  10105. status open
  10106. \begin_layout Plain Layout
  10107. \backslash
  10108. afterpage{
  10109. \end_layout
  10110. \begin_layout Plain Layout
  10111. \backslash
  10112. begin{landscape}
  10113. \end_layout
  10114. \end_inset
  10115. \end_layout
  10116. \begin_layout Standard
  10117. \begin_inset Float table
  10118. placement p
  10119. wide false
  10120. sideways false
  10121. status open
  10122. \begin_layout Plain Layout
  10123. \align center
  10124. \begin_inset Tabular
  10125. <lyxtabular version="3" rows="4" columns="7">
  10126. <features tabularvalignment="middle">
  10127. <column alignment="center" valignment="top">
  10128. <column alignment="center" valignment="top">
  10129. <column alignment="center" valignment="top">
  10130. <column alignment="center" valignment="top">
  10131. <column alignment="center" valignment="top">
  10132. <column alignment="center" valignment="top">
  10133. <column alignment="center" valignment="top">
  10134. <row>
  10135. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  10136. \begin_inset Text
  10137. \begin_layout Plain Layout
  10138. \end_layout
  10139. \end_inset
  10140. </cell>
  10141. <cell multicolumn="1" alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  10142. \begin_inset Text
  10143. \begin_layout Plain Layout
  10144. \family roman
  10145. \series medium
  10146. \shape up
  10147. \size normal
  10148. \emph off
  10149. \bar no
  10150. \strikeout off
  10151. \xout off
  10152. \uuline off
  10153. \uwave off
  10154. \noun off
  10155. \color none
  10156. Percent of Total Reads
  10157. \end_layout
  10158. \end_inset
  10159. </cell>
  10160. <cell multicolumn="2" alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  10161. \begin_inset Text
  10162. \begin_layout Plain Layout
  10163. \end_layout
  10164. \end_inset
  10165. </cell>
  10166. <cell multicolumn="2" alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  10167. \begin_inset Text
  10168. \begin_layout Plain Layout
  10169. \end_layout
  10170. \end_inset
  10171. </cell>
  10172. <cell multicolumn="2" alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  10173. \begin_inset Text
  10174. \begin_layout Plain Layout
  10175. \end_layout
  10176. \end_inset
  10177. </cell>
  10178. <cell multicolumn="1" alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  10179. \begin_inset Text
  10180. \begin_layout Plain Layout
  10181. \family roman
  10182. \series medium
  10183. \shape up
  10184. \size normal
  10185. \emph off
  10186. \bar no
  10187. \strikeout off
  10188. \xout off
  10189. \uuline off
  10190. \uwave off
  10191. \noun off
  10192. \color none
  10193. Percent of Genic Reads
  10194. \end_layout
  10195. \end_inset
  10196. </cell>
  10197. <cell multicolumn="2" alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  10198. \begin_inset Text
  10199. \begin_layout Plain Layout
  10200. \end_layout
  10201. \end_inset
  10202. </cell>
  10203. </row>
  10204. <row>
  10205. <cell alignment="center" valignment="top" bottomline="true" leftline="true" usebox="none">
  10206. \begin_inset Text
  10207. \begin_layout Plain Layout
  10208. GB
  10209. \end_layout
  10210. \end_inset
  10211. </cell>
  10212. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  10213. \begin_inset Text
  10214. \begin_layout Plain Layout
  10215. \family roman
  10216. \series medium
  10217. \shape up
  10218. \size normal
  10219. \emph off
  10220. \bar no
  10221. \strikeout off
  10222. \xout off
  10223. \uuline off
  10224. \uwave off
  10225. \noun off
  10226. \color none
  10227. Non-globin Reads
  10228. \end_layout
  10229. \end_inset
  10230. </cell>
  10231. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  10232. \begin_inset Text
  10233. \begin_layout Plain Layout
  10234. \family roman
  10235. \series medium
  10236. \shape up
  10237. \size normal
  10238. \emph off
  10239. \bar no
  10240. \strikeout off
  10241. \xout off
  10242. \uuline off
  10243. \uwave off
  10244. \noun off
  10245. \color none
  10246. Globin Reads
  10247. \end_layout
  10248. \end_inset
  10249. </cell>
  10250. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  10251. \begin_inset Text
  10252. \begin_layout Plain Layout
  10253. \family roman
  10254. \series medium
  10255. \shape up
  10256. \size normal
  10257. \emph off
  10258. \bar no
  10259. \strikeout off
  10260. \xout off
  10261. \uuline off
  10262. \uwave off
  10263. \noun off
  10264. \color none
  10265. All Genic Reads
  10266. \end_layout
  10267. \end_inset
  10268. </cell>
  10269. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  10270. \begin_inset Text
  10271. \begin_layout Plain Layout
  10272. \family roman
  10273. \series medium
  10274. \shape up
  10275. \size normal
  10276. \emph off
  10277. \bar no
  10278. \strikeout off
  10279. \xout off
  10280. \uuline off
  10281. \uwave off
  10282. \noun off
  10283. \color none
  10284. All Aligned Reads
  10285. \end_layout
  10286. \end_inset
  10287. </cell>
  10288. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  10289. \begin_inset Text
  10290. \begin_layout Plain Layout
  10291. \family roman
  10292. \series medium
  10293. \shape up
  10294. \size normal
  10295. \emph off
  10296. \bar no
  10297. \strikeout off
  10298. \xout off
  10299. \uuline off
  10300. \uwave off
  10301. \noun off
  10302. \color none
  10303. Non-globin Reads
  10304. \end_layout
  10305. \end_inset
  10306. </cell>
  10307. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  10308. \begin_inset Text
  10309. \begin_layout Plain Layout
  10310. \family roman
  10311. \series medium
  10312. \shape up
  10313. \size normal
  10314. \emph off
  10315. \bar no
  10316. \strikeout off
  10317. \xout off
  10318. \uuline off
  10319. \uwave off
  10320. \noun off
  10321. \color none
  10322. Globin Reads
  10323. \end_layout
  10324. \end_inset
  10325. </cell>
  10326. </row>
  10327. <row>
  10328. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  10329. \begin_inset Text
  10330. \begin_layout Plain Layout
  10331. \family roman
  10332. \series medium
  10333. \shape up
  10334. \size normal
  10335. \emph off
  10336. \bar no
  10337. \strikeout off
  10338. \xout off
  10339. \uuline off
  10340. \uwave off
  10341. \noun off
  10342. \color none
  10343. Yes
  10344. \end_layout
  10345. \end_inset
  10346. </cell>
  10347. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  10348. \begin_inset Text
  10349. \begin_layout Plain Layout
  10350. \family roman
  10351. \series medium
  10352. \shape up
  10353. \size normal
  10354. \emph off
  10355. \bar no
  10356. \strikeout off
  10357. \xout off
  10358. \uuline off
  10359. \uwave off
  10360. \noun off
  10361. \color none
  10362. 50.4% ± 6.82
  10363. \end_layout
  10364. \end_inset
  10365. </cell>
  10366. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  10367. \begin_inset Text
  10368. \begin_layout Plain Layout
  10369. \family roman
  10370. \series medium
  10371. \shape up
  10372. \size normal
  10373. \emph off
  10374. \bar no
  10375. \strikeout off
  10376. \xout off
  10377. \uuline off
  10378. \uwave off
  10379. \noun off
  10380. \color none
  10381. 3.48% ± 2.94
  10382. \end_layout
  10383. \end_inset
  10384. </cell>
  10385. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  10386. \begin_inset Text
  10387. \begin_layout Plain Layout
  10388. \family roman
  10389. \series medium
  10390. \shape up
  10391. \size normal
  10392. \emph off
  10393. \bar no
  10394. \strikeout off
  10395. \xout off
  10396. \uuline off
  10397. \uwave off
  10398. \noun off
  10399. \color none
  10400. 53.9% ± 6.81
  10401. \end_layout
  10402. \end_inset
  10403. </cell>
  10404. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  10405. \begin_inset Text
  10406. \begin_layout Plain Layout
  10407. \family roman
  10408. \series medium
  10409. \shape up
  10410. \size normal
  10411. \emph off
  10412. \bar no
  10413. \strikeout off
  10414. \xout off
  10415. \uuline off
  10416. \uwave off
  10417. \noun off
  10418. \color none
  10419. 89.7% ± 2.40
  10420. \end_layout
  10421. \end_inset
  10422. </cell>
  10423. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  10424. \begin_inset Text
  10425. \begin_layout Plain Layout
  10426. \family roman
  10427. \series medium
  10428. \shape up
  10429. \size normal
  10430. \emph off
  10431. \bar no
  10432. \strikeout off
  10433. \xout off
  10434. \uuline off
  10435. \uwave off
  10436. \noun off
  10437. \color none
  10438. 93.5% ± 5.25
  10439. \end_layout
  10440. \end_inset
  10441. </cell>
  10442. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  10443. \begin_inset Text
  10444. \begin_layout Plain Layout
  10445. \family roman
  10446. \series medium
  10447. \shape up
  10448. \size normal
  10449. \emph off
  10450. \bar no
  10451. \strikeout off
  10452. \xout off
  10453. \uuline off
  10454. \uwave off
  10455. \noun off
  10456. \color none
  10457. 6.49% ± 5.25
  10458. \end_layout
  10459. \end_inset
  10460. </cell>
  10461. </row>
  10462. <row>
  10463. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  10464. \begin_inset Text
  10465. \begin_layout Plain Layout
  10466. \family roman
  10467. \series medium
  10468. \shape up
  10469. \size normal
  10470. \emph off
  10471. \bar no
  10472. \strikeout off
  10473. \xout off
  10474. \uuline off
  10475. \uwave off
  10476. \noun off
  10477. \color none
  10478. No
  10479. \end_layout
  10480. \end_inset
  10481. </cell>
  10482. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  10483. \begin_inset Text
  10484. \begin_layout Plain Layout
  10485. \family roman
  10486. \series medium
  10487. \shape up
  10488. \size normal
  10489. \emph off
  10490. \bar no
  10491. \strikeout off
  10492. \xout off
  10493. \uuline off
  10494. \uwave off
  10495. \noun off
  10496. \color none
  10497. 26.3% ± 8.95
  10498. \end_layout
  10499. \end_inset
  10500. </cell>
  10501. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  10502. \begin_inset Text
  10503. \begin_layout Plain Layout
  10504. \family roman
  10505. \series medium
  10506. \shape up
  10507. \size normal
  10508. \emph off
  10509. \bar no
  10510. \strikeout off
  10511. \xout off
  10512. \uuline off
  10513. \uwave off
  10514. \noun off
  10515. \color none
  10516. 44.6% ± 16.6
  10517. \end_layout
  10518. \end_inset
  10519. </cell>
  10520. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  10521. \begin_inset Text
  10522. \begin_layout Plain Layout
  10523. \family roman
  10524. \series medium
  10525. \shape up
  10526. \size normal
  10527. \emph off
  10528. \bar no
  10529. \strikeout off
  10530. \xout off
  10531. \uuline off
  10532. \uwave off
  10533. \noun off
  10534. \color none
  10535. 70.1% ± 9.38
  10536. \end_layout
  10537. \end_inset
  10538. </cell>
  10539. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  10540. \begin_inset Text
  10541. \begin_layout Plain Layout
  10542. \family roman
  10543. \series medium
  10544. \shape up
  10545. \size normal
  10546. \emph off
  10547. \bar no
  10548. \strikeout off
  10549. \xout off
  10550. \uuline off
  10551. \uwave off
  10552. \noun off
  10553. \color none
  10554. 90.7% ± 5.16
  10555. \end_layout
  10556. \end_inset
  10557. </cell>
  10558. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  10559. \begin_inset Text
  10560. \begin_layout Plain Layout
  10561. \family roman
  10562. \series medium
  10563. \shape up
  10564. \size normal
  10565. \emph off
  10566. \bar no
  10567. \strikeout off
  10568. \xout off
  10569. \uuline off
  10570. \uwave off
  10571. \noun off
  10572. \color none
  10573. 38.8% ± 17.1
  10574. \end_layout
  10575. \end_inset
  10576. </cell>
  10577. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  10578. \begin_inset Text
  10579. \begin_layout Plain Layout
  10580. \family roman
  10581. \series medium
  10582. \shape up
  10583. \size normal
  10584. \emph off
  10585. \bar no
  10586. \strikeout off
  10587. \xout off
  10588. \uuline off
  10589. \uwave off
  10590. \noun off
  10591. \color none
  10592. 61.2% ± 17.1
  10593. \end_layout
  10594. \end_inset
  10595. </cell>
  10596. </row>
  10597. </lyxtabular>
  10598. \end_inset
  10599. \end_layout
  10600. \begin_layout Plain Layout
  10601. \begin_inset Caption Standard
  10602. \begin_layout Plain Layout
  10603. \series bold
  10604. \begin_inset Argument 1
  10605. status collapsed
  10606. \begin_layout Plain Layout
  10607. Fractions of reads mapping to genomic features in GB and non-GB samples.
  10608. \end_layout
  10609. \end_inset
  10610. \begin_inset CommandInset label
  10611. LatexCommand label
  10612. name "tab:Fractions-of-reads"
  10613. \end_inset
  10614. Fractions of reads mapping to genomic features in GB and non-GB samples.
  10615. \series default
  10616. All values are given as mean ± standard deviation.
  10617. \end_layout
  10618. \end_inset
  10619. \end_layout
  10620. \end_inset
  10621. \end_layout
  10622. \begin_layout Standard
  10623. \begin_inset ERT
  10624. status open
  10625. \begin_layout Plain Layout
  10626. \backslash
  10627. end{landscape}
  10628. \end_layout
  10629. \begin_layout Plain Layout
  10630. }
  10631. \end_layout
  10632. \end_inset
  10633. \end_layout
  10634. \begin_layout Standard
  10635. The objective of the present study was to validate a new protocol for deep
  10636. RNA-seq of whole blood drawn into PaxGene tubes from cynomolgus monkeys
  10637. undergoing islet transplantation, with particular focus on minimizing the
  10638. loss of useful sequencing space to uninformative globin reads.
  10639. The details of the analysis with respect to transplant outcomes and the
  10640. impact of mesenchymal stem cell treatment will be reported in a separate
  10641. manuscript (in preparation).
  10642. To focus on the efficacy of our globin blocking protocol, 37 blood samples,
  10643. 16 from pre-transplant and 21 from post-transplant time points, were each
  10644. prepped once with and once without globin blocking oligos, and were then
  10645. sequenced on an Illumina NextSeq500 instrument.
  10646. The number of reads aligning to each gene in the cynomolgus genome was
  10647. counted.
  10648. Table 1 summarizes the distribution of read fractions among the GB and
  10649. non-GB libraries.
  10650. In the libraries with no globin blocking, globin reads made up an average
  10651. of 44.6% of total input reads, while reads assigned to all other genes made
  10652. up an average of 26.3%.
  10653. The remaining reads either aligned to intergenic regions (that include
  10654. long non-coding RNAs) or did not align with any annotated transcripts in
  10655. the current build of the cynomolgus genome.
  10656. In the GB libraries, globin reads made up only 3.48% and reads assigned
  10657. to all other genes increased to 50.4%.
  10658. Thus, globin blocking resulted in a 92.2% reduction in globin reads and
  10659. a 91.6% increase in yield of useful non-globin reads.
  10660. \end_layout
  10661. \begin_layout Standard
  10662. This reduction is not quite as efficient as the previous analysis showed
  10663. for human samples by DeepSAGE (<0.4% globin reads after globin reduction)
  10664. \begin_inset CommandInset citation
  10665. LatexCommand cite
  10666. key "Mastrokolias2012"
  10667. literal "false"
  10668. \end_inset
  10669. .
  10670. Nonetheless, this degree of globin reduction is sufficient to nearly double
  10671. the yield of useful reads.
  10672. Thus, globin blocking cuts the required sequencing effort (and costs) to
  10673. achieve a target coverage depth by almost 50%.
  10674. Consistent with this near doubling of yield, the average difference in
  10675. un-normalized logCPM across all genes between the GB libraries and non-GB
  10676. libraries is approximately 1 (mean = 1.01, median = 1.08), an overall 2-fold
  10677. increase.
  10678. Un-normalized values are used here because the TMM normalization correctly
  10679. identifies this 2-fold difference as biologically irrelevant and removes
  10680. it.
  10681. \end_layout
  10682. \begin_layout Standard
  10683. \begin_inset Float figure
  10684. wide false
  10685. sideways false
  10686. status collapsed
  10687. \begin_layout Plain Layout
  10688. \align center
  10689. \begin_inset Graphics
  10690. filename graphics/Globin Paper/figure1 - globin-fractions.pdf
  10691. lyxscale 50
  10692. width 75col%
  10693. \end_inset
  10694. \end_layout
  10695. \begin_layout Plain Layout
  10696. \begin_inset Caption Standard
  10697. \begin_layout Plain Layout
  10698. \series bold
  10699. \begin_inset Argument 1
  10700. status collapsed
  10701. \begin_layout Plain Layout
  10702. Fraction of genic reads in each sample aligned to non-globin genes, with
  10703. and without globin blocking (GB).
  10704. \end_layout
  10705. \end_inset
  10706. \begin_inset CommandInset label
  10707. LatexCommand label
  10708. name "fig:Fraction-of-genic-reads"
  10709. \end_inset
  10710. Fraction of genic reads in each sample aligned to non-globin genes, with
  10711. and without globin blocking (GB).
  10712. \series default
  10713. All reads in each sequencing library were aligned to the cyno genome, and
  10714. the number of reads uniquely aligning to each gene was counted.
  10715. For each sample, counts were summed separately for all globin genes and
  10716. for the remainder of the genes (non-globin genes), and the fraction of
  10717. genic reads aligned to non-globin genes was computed.
  10718. Each point represents an individual sample.
  10719. Gray + signs indicate the means for globin-blocked libraries and unblocked
  10720. libraries.
  10721. The overall distribution for each group is represented as a notched box
  10722. plots.
  10723. Points are randomly spread vertically to avoid excessive overlapping.
  10724. \end_layout
  10725. \end_inset
  10726. \end_layout
  10727. \end_inset
  10728. \end_layout
  10729. \begin_layout Standard
  10730. Another important aspect is that the standard deviations in Table
  10731. \begin_inset CommandInset ref
  10732. LatexCommand ref
  10733. reference "tab:Fractions-of-reads"
  10734. plural "false"
  10735. caps "false"
  10736. noprefix "false"
  10737. \end_inset
  10738. are uniformly smaller in the GB samples than the non-GB ones, indicating
  10739. much greater consistency of yield.
  10740. This is best seen in the percentage of non-globin reads as a fraction of
  10741. total reads aligned to annotated genes (genic reads).
  10742. For the non-GB samples, this measure ranges from 10.9% to 80.9%, while for
  10743. the GB samples it ranges from 81.9% to 99.9% (Figure
  10744. \begin_inset CommandInset ref
  10745. LatexCommand ref
  10746. reference "fig:Fraction-of-genic-reads"
  10747. plural "false"
  10748. caps "false"
  10749. noprefix "false"
  10750. \end_inset
  10751. ).
  10752. This means that for applications where it is critical that each sample
  10753. achieve a specified minimum coverage in order to provide useful information,
  10754. it would be necessary to budget up to 10 times the sequencing depth per
  10755. sample without globin blocking, even though the average yield improvement
  10756. for globin blocking is only 2-fold, because every sample has a chance of
  10757. being 90% globin and 10% useful reads.
  10758. Hence, the more consistent behavior of GB samples makes planning an experiment
  10759. easier and more efficient because it eliminates the need to over-sequence
  10760. every sample in order to guard against the worst case of a high-globin
  10761. fraction.
  10762. \end_layout
  10763. \begin_layout Subsection
  10764. Globin blocking lowers the noise floor and allows detection of about 2000
  10765. more low-expression genes
  10766. \end_layout
  10767. \begin_layout Standard
  10768. \begin_inset Flex TODO Note (inline)
  10769. status open
  10770. \begin_layout Plain Layout
  10771. Remove redundant titles from figures
  10772. \end_layout
  10773. \end_inset
  10774. \end_layout
  10775. \begin_layout Standard
  10776. \begin_inset Float figure
  10777. wide false
  10778. sideways false
  10779. status collapsed
  10780. \begin_layout Plain Layout
  10781. \align center
  10782. \begin_inset Graphics
  10783. filename graphics/Globin Paper/figure2 - aveLogCPM-colored.pdf
  10784. lyxscale 50
  10785. height 60theight%
  10786. \end_inset
  10787. \end_layout
  10788. \begin_layout Plain Layout
  10789. \begin_inset Caption Standard
  10790. \begin_layout Plain Layout
  10791. \series bold
  10792. \begin_inset Argument 1
  10793. status collapsed
  10794. \begin_layout Plain Layout
  10795. Distributions of average group gene abundances when normalized separately
  10796. or together.
  10797. \end_layout
  10798. \end_inset
  10799. \begin_inset CommandInset label
  10800. LatexCommand label
  10801. name "fig:logcpm-dists"
  10802. \end_inset
  10803. Distributions of average group gene abundances when normalized separately
  10804. or together.
  10805. \series default
  10806. All reads in each sequencing library were aligned to the cyno genome, and
  10807. the number of reads uniquely aligning to each gene was counted.
  10808. Genes with zero counts in all libraries were discarded.
  10809. Libraries were normalized using the TMM method.
  10810. Libraries were split into globin-blocked (GB) and non-GB groups and the
  10811. average abundance for each gene in both groups, measured in log2 counts
  10812. per million reads counted, was computed using the aveLogCPM function.
  10813. The distribution of average gene logCPM values was plotted for both groups
  10814. using a kernel density plot to approximate a continuous distribution.
  10815. The logCPM GB distributions are marked in red, non-GB in blue.
  10816. The black vertical line denotes the chosen detection threshold of -1.
  10817. Top panel: Libraries were split into GB and non-GB groups first and normalized
  10818. separately.
  10819. Bottom panel: Libraries were all normalized together first and then split
  10820. into groups.
  10821. \end_layout
  10822. \end_inset
  10823. \end_layout
  10824. \begin_layout Plain Layout
  10825. \end_layout
  10826. \end_inset
  10827. \end_layout
  10828. \begin_layout Standard
  10829. Since globin blocking yields more usable sequencing depth, it should also
  10830. allow detection of more genes at any given threshold.
  10831. When we looked at the distribution of average normalized logCPM values
  10832. across all libraries for genes with at least one read assigned to them,
  10833. we observed the expected bimodal distribution, with a high-abundance "signal"
  10834. peak representing detected genes and a low-abundance "noise" peak representing
  10835. genes whose read count did not rise above the noise floor (Figure
  10836. \begin_inset CommandInset ref
  10837. LatexCommand ref
  10838. reference "fig:logcpm-dists"
  10839. plural "false"
  10840. caps "false"
  10841. noprefix "false"
  10842. \end_inset
  10843. ).
  10844. Consistent with the 2-fold increase in raw counts assigned to non-globin
  10845. genes, the signal peak for GB samples is shifted to the right relative
  10846. to the non-GB signal peak.
  10847. When all the samples are normalized together, this difference is normalized
  10848. out, lining up the signal peaks, and this reveals that, as expected, the
  10849. noise floor for the GB samples is about 2-fold lower.
  10850. This greater separation between signal and noise peaks in the GB samples
  10851. means that low-expression genes should be more easily detected and more
  10852. precisely quantified than in the non-GB samples.
  10853. \end_layout
  10854. \begin_layout Standard
  10855. \begin_inset Float figure
  10856. wide false
  10857. sideways false
  10858. status collapsed
  10859. \begin_layout Plain Layout
  10860. \align center
  10861. \begin_inset Graphics
  10862. filename graphics/Globin Paper/figure3 - detection.pdf
  10863. lyxscale 50
  10864. width 70col%
  10865. \end_inset
  10866. \end_layout
  10867. \begin_layout Plain Layout
  10868. \begin_inset Caption Standard
  10869. \begin_layout Plain Layout
  10870. \series bold
  10871. \begin_inset Argument 1
  10872. status collapsed
  10873. \begin_layout Plain Layout
  10874. Gene detections as a function of abundance thresholds in globin-blocked
  10875. (GB) and non-GB samples.
  10876. \end_layout
  10877. \end_inset
  10878. \begin_inset CommandInset label
  10879. LatexCommand label
  10880. name "fig:Gene-detections"
  10881. \end_inset
  10882. Gene detections as a function of abundance thresholds in globin-blocked
  10883. (GB) and non-GB samples.
  10884. \series default
  10885. Average abundance (logCPM,
  10886. \begin_inset Formula $\log_{2}$
  10887. \end_inset
  10888. counts per million reads counted) was computed by separate group normalization
  10889. as described in Figure
  10890. \begin_inset CommandInset ref
  10891. LatexCommand ref
  10892. reference "fig:logcpm-dists"
  10893. plural "false"
  10894. caps "false"
  10895. noprefix "false"
  10896. \end_inset
  10897. for both the GB and non-GB groups, as well as for all samples considered
  10898. as one large group.
  10899. For each every integer threshold from -2 to 3, the number of genes detected
  10900. at or above that logCPM threshold was plotted for each group.
  10901. \end_layout
  10902. \end_inset
  10903. \end_layout
  10904. \begin_layout Plain Layout
  10905. \end_layout
  10906. \end_inset
  10907. \end_layout
  10908. \begin_layout Standard
  10909. Based on these distributions, we selected a detection threshold of -1, which
  10910. is approximately the leftmost edge of the trough between the signal and
  10911. noise peaks.
  10912. This represents the most liberal possible detection threshold that doesn't
  10913. call substantial numbers of noise genes as detected.
  10914. Among the full dataset, 13429 genes were detected at this threshold, and
  10915. 22276 were not.
  10916. When considering the GB libraries and non-GB libraries separately and re-comput
  10917. ing normalization factors independently within each group, 14535 genes were
  10918. detected in the GB libraries while only 12460 were detected in the non-GB
  10919. libraries.
  10920. Thus, GB allowed the detection of 2000 extra genes that were buried under
  10921. the noise floor without GB.
  10922. This pattern of at least 2000 additional genes detected with GB was also
  10923. consistent across a wide range of possible detection thresholds, from -2
  10924. to 3 (see Figure
  10925. \begin_inset CommandInset ref
  10926. LatexCommand ref
  10927. reference "fig:Gene-detections"
  10928. plural "false"
  10929. caps "false"
  10930. noprefix "false"
  10931. \end_inset
  10932. ).
  10933. \end_layout
  10934. \begin_layout Subsection
  10935. Globin blocking does not add significant additional noise or decrease sample
  10936. quality
  10937. \end_layout
  10938. \begin_layout Standard
  10939. One potential worry is that the globin blocking protocol could perturb the
  10940. levels of non-globin genes.
  10941. There are two kinds of possible perturbations: systematic and random.
  10942. The former is not a major concern for detection of differential expression,
  10943. since a 2-fold change in every sample has no effect on the relative fold
  10944. change between samples.
  10945. In contrast, random perturbations would increase the noise and obscure
  10946. the signal in the dataset, reducing the capacity to detect differential
  10947. expression.
  10948. \end_layout
  10949. \begin_layout Standard
  10950. \begin_inset Float figure
  10951. wide false
  10952. sideways false
  10953. status collapsed
  10954. \begin_layout Plain Layout
  10955. \align center
  10956. \begin_inset Graphics
  10957. filename graphics/Globin Paper/figure4 - maplot-colored.pdf
  10958. lyxscale 50
  10959. width 60col%
  10960. groupId colwidth
  10961. \end_inset
  10962. \end_layout
  10963. \begin_layout Plain Layout
  10964. \begin_inset Caption Standard
  10965. \begin_layout Plain Layout
  10966. \begin_inset Argument 1
  10967. status collapsed
  10968. \begin_layout Plain Layout
  10969. MA plot showing effects of globin blocking on each gene's abundance.
  10970. \end_layout
  10971. \end_inset
  10972. \begin_inset CommandInset label
  10973. LatexCommand label
  10974. name "fig:MA-plot"
  10975. \end_inset
  10976. \series bold
  10977. MA plot showing effects of globin blocking on each gene's abundance.
  10978. \series default
  10979. All libraries were normalized together as described in Figure
  10980. \begin_inset CommandInset ref
  10981. LatexCommand ref
  10982. reference "fig:logcpm-dists"
  10983. plural "false"
  10984. caps "false"
  10985. noprefix "false"
  10986. \end_inset
  10987. , and genes with an average logCPM below -1 were filtered out.
  10988. Each remaining gene was tested for differential abundance with respect
  10989. to globin blocking (GB) using
  10990. \begin_inset Flex Code
  10991. status open
  10992. \begin_layout Plain Layout
  10993. edgeR
  10994. \end_layout
  10995. \end_inset
  10996. ’s quasi-likelihood F-test, fitting a negative binomial generalized linear
  10997. model to table of read counts in each library.
  10998. For each gene,
  10999. \begin_inset Flex Code
  11000. status open
  11001. \begin_layout Plain Layout
  11002. edgeR
  11003. \end_layout
  11004. \end_inset
  11005. reported average abundance (logCPM),
  11006. \begin_inset Formula $\log_{2}$
  11007. \end_inset
  11008. fold change (logFC), p-value, and Benjamini-Hochberg adjusted false discovery
  11009. rate (FDR).
  11010. Each gene's logFC was plotted against its logCPM, colored by FDR.
  11011. Red points are significant at ≤10% FDR, and blue are not significant at
  11012. that threshold.
  11013. The alpha and beta globin genes targeted for blocking are marked with large
  11014. triangles, while all other genes are represented as small points.
  11015. \end_layout
  11016. \end_inset
  11017. \end_layout
  11018. \begin_layout Plain Layout
  11019. \end_layout
  11020. \end_inset
  11021. \end_layout
  11022. \begin_layout Standard
  11023. \begin_inset Flex TODO Note (inline)
  11024. status open
  11025. \begin_layout Plain Layout
  11026. Standardize on
  11027. \begin_inset Quotes eld
  11028. \end_inset
  11029. log2
  11030. \begin_inset Quotes erd
  11031. \end_inset
  11032. notation
  11033. \end_layout
  11034. \end_inset
  11035. \end_layout
  11036. \begin_layout Standard
  11037. The data do indeed show small systematic perturbations in gene levels (Figure
  11038. \begin_inset CommandInset ref
  11039. LatexCommand ref
  11040. reference "fig:MA-plot"
  11041. plural "false"
  11042. caps "false"
  11043. noprefix "false"
  11044. \end_inset
  11045. ).
  11046. Other than the 3 designated alpha and beta globin genes, two other genes
  11047. stand out as having especially large negative log fold changes: HBD and
  11048. LOC1021365.
  11049. HBD, delta globin, is most likely targeted by the blocking oligos due to
  11050. high sequence homology with the other globin genes.
  11051. LOC1021365 is the aforementioned ncRNA that is reverse-complementary to
  11052. one of the alpha-like genes and that would be expected to be removed during
  11053. the globin blocking step.
  11054. All other genes appear in a cluster centered vertically at 0, and the vast
  11055. majority of genes in this cluster show an absolute log2(FC) of 0.5 or less.
  11056. Nevertheless, many of these small perturbations are still statistically
  11057. significant, indicating that the globin blocking oligos likely cause very
  11058. small but non-zero systematic perturbations in measured gene expression
  11059. levels.
  11060. \end_layout
  11061. \begin_layout Standard
  11062. \begin_inset Float figure
  11063. wide false
  11064. sideways false
  11065. status collapsed
  11066. \begin_layout Plain Layout
  11067. \align center
  11068. \begin_inset Graphics
  11069. filename graphics/Globin Paper/figure5 - corrplot.pdf
  11070. lyxscale 50
  11071. width 70col%
  11072. \end_inset
  11073. \end_layout
  11074. \begin_layout Plain Layout
  11075. \begin_inset Caption Standard
  11076. \begin_layout Plain Layout
  11077. \series bold
  11078. \begin_inset Argument 1
  11079. status collapsed
  11080. \begin_layout Plain Layout
  11081. Comparison of inter-sample gene abundance correlations with and without
  11082. globin blocking.
  11083. \end_layout
  11084. \end_inset
  11085. \begin_inset CommandInset label
  11086. LatexCommand label
  11087. name "fig:gene-abundance-correlations"
  11088. \end_inset
  11089. Comparison of inter-sample gene abundance correlations with and without
  11090. globin blocking (GB).
  11091. \series default
  11092. All libraries were normalized together as described in Figure 2, and genes
  11093. with an average abundance (logCPM, log2 counts per million reads counted)
  11094. less than -1 were filtered out.
  11095. Each gene’s logCPM was computed in each library using the
  11096. \begin_inset Flex Code
  11097. status open
  11098. \begin_layout Plain Layout
  11099. edgeR
  11100. \end_layout
  11101. \end_inset
  11102. cpm function.
  11103. For each pair of biological samples, the Pearson correlation between those
  11104. samples' GB libraries was plotted against the correlation between the same
  11105. samples’ non-GB libraries.
  11106. Each point represents an unique pair of samples.
  11107. The solid gray line shows a quantile-quantile plot of distribution of GB
  11108. correlations vs.
  11109. that of non-GB correlations.
  11110. The thin dashed line is the identity line, provided for reference.
  11111. \end_layout
  11112. \end_inset
  11113. \end_layout
  11114. \begin_layout Plain Layout
  11115. \end_layout
  11116. \end_inset
  11117. \end_layout
  11118. \begin_layout Standard
  11119. To evaluate the possibility of globin blocking causing random perturbations
  11120. and reducing sample quality, we computed the Pearson correlation between
  11121. logCPM values for every pair of samples with and without GB and plotted
  11122. them against each other (Figure
  11123. \begin_inset CommandInset ref
  11124. LatexCommand ref
  11125. reference "fig:gene-abundance-correlations"
  11126. plural "false"
  11127. caps "false"
  11128. noprefix "false"
  11129. \end_inset
  11130. ).
  11131. The plot indicated that the GB libraries have higher sample-to-sample correlati
  11132. ons than the non-GB libraries.
  11133. Parametric and nonparametric tests for differences between the correlations
  11134. with and without GB both confirmed that this difference was highly significant
  11135. (2-sided paired t-test: t = 37.2, df = 665, P ≪ 2.2e-16; 2-sided Wilcoxon
  11136. sign-rank test: V = 2195, P ≪ 2.2e-16).
  11137. Performing the same tests on the Spearman correlations gave the same conclusion
  11138. (t-test: t = 26.8, df = 665, P ≪ 2.2e-16; sign-rank test: V = 8781, P ≪ 2.2e-16).
  11139. The
  11140. \begin_inset Flex Code
  11141. status open
  11142. \begin_layout Plain Layout
  11143. edgeR
  11144. \end_layout
  11145. \end_inset
  11146. package was used to compute the overall biological coefficient of variation
  11147. (BCV) for GB and non-GB libraries, and found that globin blocking resulted
  11148. in a negligible increase in the BCV (0.417 with GB vs.
  11149. 0.400 without).
  11150. The near equality of the BCVs for both sets indicates that the higher correlati
  11151. ons in the GB libraries are most likely a result of the increased yield
  11152. of useful reads, which reduces the contribution of Poisson counting uncertainty
  11153. to the overall variance of the logCPM values
  11154. \begin_inset CommandInset citation
  11155. LatexCommand cite
  11156. key "McCarthy2012"
  11157. literal "false"
  11158. \end_inset
  11159. .
  11160. This improves the precision of expression measurements and more than offsets
  11161. the negligible increase in BCV.
  11162. \end_layout
  11163. \begin_layout Subsection
  11164. More differentially expressed genes are detected with globin blocking
  11165. \end_layout
  11166. \begin_layout Standard
  11167. \begin_inset Float table
  11168. wide false
  11169. sideways false
  11170. status collapsed
  11171. \begin_layout Plain Layout
  11172. \align center
  11173. \begin_inset Tabular
  11174. <lyxtabular version="3" rows="5" columns="5">
  11175. <features tabularvalignment="middle">
  11176. <column alignment="center" valignment="top">
  11177. <column alignment="center" valignment="top">
  11178. <column alignment="center" valignment="top">
  11179. <column alignment="center" valignment="top">
  11180. <column alignment="center" valignment="top">
  11181. <row>
  11182. <cell alignment="center" valignment="top" usebox="none">
  11183. \begin_inset Text
  11184. \begin_layout Plain Layout
  11185. \end_layout
  11186. \end_inset
  11187. </cell>
  11188. <cell alignment="center" valignment="top" usebox="none">
  11189. \begin_inset Text
  11190. \begin_layout Plain Layout
  11191. \end_layout
  11192. \end_inset
  11193. </cell>
  11194. <cell multicolumn="1" alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  11195. \begin_inset Text
  11196. \begin_layout Plain Layout
  11197. \series bold
  11198. No Globin Blocking
  11199. \end_layout
  11200. \end_inset
  11201. </cell>
  11202. <cell multicolumn="2" alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  11203. \begin_inset Text
  11204. \begin_layout Plain Layout
  11205. \end_layout
  11206. \end_inset
  11207. </cell>
  11208. <cell multicolumn="2" alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  11209. \begin_inset Text
  11210. \begin_layout Plain Layout
  11211. \end_layout
  11212. \end_inset
  11213. </cell>
  11214. </row>
  11215. <row>
  11216. <cell alignment="center" valignment="top" usebox="none">
  11217. \begin_inset Text
  11218. \begin_layout Plain Layout
  11219. \end_layout
  11220. \end_inset
  11221. </cell>
  11222. <cell alignment="center" valignment="top" usebox="none">
  11223. \begin_inset Text
  11224. \begin_layout Plain Layout
  11225. \end_layout
  11226. \end_inset
  11227. </cell>
  11228. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  11229. \begin_inset Text
  11230. \begin_layout Plain Layout
  11231. \series bold
  11232. Up
  11233. \end_layout
  11234. \end_inset
  11235. </cell>
  11236. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  11237. \begin_inset Text
  11238. \begin_layout Plain Layout
  11239. \series bold
  11240. NS
  11241. \end_layout
  11242. \end_inset
  11243. </cell>
  11244. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  11245. \begin_inset Text
  11246. \begin_layout Plain Layout
  11247. \series bold
  11248. Down
  11249. \end_layout
  11250. \end_inset
  11251. </cell>
  11252. </row>
  11253. <row>
  11254. <cell multirow="3" alignment="center" valignment="middle" topline="true" bottomline="true" leftline="true" usebox="none">
  11255. \begin_inset Text
  11256. \begin_layout Plain Layout
  11257. \series bold
  11258. Globin-Blocking
  11259. \end_layout
  11260. \end_inset
  11261. </cell>
  11262. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  11263. \begin_inset Text
  11264. \begin_layout Plain Layout
  11265. \series bold
  11266. Up
  11267. \end_layout
  11268. \end_inset
  11269. </cell>
  11270. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  11271. \begin_inset Text
  11272. \begin_layout Plain Layout
  11273. \family roman
  11274. \series medium
  11275. \shape up
  11276. \size normal
  11277. \emph off
  11278. \bar no
  11279. \strikeout off
  11280. \xout off
  11281. \uuline off
  11282. \uwave off
  11283. \noun off
  11284. \color none
  11285. 231
  11286. \end_layout
  11287. \end_inset
  11288. </cell>
  11289. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  11290. \begin_inset Text
  11291. \begin_layout Plain Layout
  11292. \family roman
  11293. \series medium
  11294. \shape up
  11295. \size normal
  11296. \emph off
  11297. \bar no
  11298. \strikeout off
  11299. \xout off
  11300. \uuline off
  11301. \uwave off
  11302. \noun off
  11303. \color none
  11304. 515
  11305. \end_layout
  11306. \end_inset
  11307. </cell>
  11308. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  11309. \begin_inset Text
  11310. \begin_layout Plain Layout
  11311. \family roman
  11312. \series medium
  11313. \shape up
  11314. \size normal
  11315. \emph off
  11316. \bar no
  11317. \strikeout off
  11318. \xout off
  11319. \uuline off
  11320. \uwave off
  11321. \noun off
  11322. \color none
  11323. 2
  11324. \end_layout
  11325. \end_inset
  11326. </cell>
  11327. </row>
  11328. <row>
  11329. <cell multirow="4" alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  11330. \begin_inset Text
  11331. \begin_layout Plain Layout
  11332. \end_layout
  11333. \end_inset
  11334. </cell>
  11335. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  11336. \begin_inset Text
  11337. \begin_layout Plain Layout
  11338. \series bold
  11339. NS
  11340. \end_layout
  11341. \end_inset
  11342. </cell>
  11343. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  11344. \begin_inset Text
  11345. \begin_layout Plain Layout
  11346. \family roman
  11347. \series medium
  11348. \shape up
  11349. \size normal
  11350. \emph off
  11351. \bar no
  11352. \strikeout off
  11353. \xout off
  11354. \uuline off
  11355. \uwave off
  11356. \noun off
  11357. \color none
  11358. 160
  11359. \end_layout
  11360. \end_inset
  11361. </cell>
  11362. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  11363. \begin_inset Text
  11364. \begin_layout Plain Layout
  11365. \family roman
  11366. \series medium
  11367. \shape up
  11368. \size normal
  11369. \emph off
  11370. \bar no
  11371. \strikeout off
  11372. \xout off
  11373. \uuline off
  11374. \uwave off
  11375. \noun off
  11376. \color none
  11377. 11235
  11378. \end_layout
  11379. \end_inset
  11380. </cell>
  11381. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  11382. \begin_inset Text
  11383. \begin_layout Plain Layout
  11384. \family roman
  11385. \series medium
  11386. \shape up
  11387. \size normal
  11388. \emph off
  11389. \bar no
  11390. \strikeout off
  11391. \xout off
  11392. \uuline off
  11393. \uwave off
  11394. \noun off
  11395. \color none
  11396. 136
  11397. \end_layout
  11398. \end_inset
  11399. </cell>
  11400. </row>
  11401. <row>
  11402. <cell multirow="4" alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  11403. \begin_inset Text
  11404. \begin_layout Plain Layout
  11405. \end_layout
  11406. \end_inset
  11407. </cell>
  11408. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  11409. \begin_inset Text
  11410. \begin_layout Plain Layout
  11411. \series bold
  11412. Down
  11413. \end_layout
  11414. \end_inset
  11415. </cell>
  11416. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  11417. \begin_inset Text
  11418. \begin_layout Plain Layout
  11419. \family roman
  11420. \series medium
  11421. \shape up
  11422. \size normal
  11423. \emph off
  11424. \bar no
  11425. \strikeout off
  11426. \xout off
  11427. \uuline off
  11428. \uwave off
  11429. \noun off
  11430. \color none
  11431. 0
  11432. \end_layout
  11433. \end_inset
  11434. </cell>
  11435. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  11436. \begin_inset Text
  11437. \begin_layout Plain Layout
  11438. \family roman
  11439. \series medium
  11440. \shape up
  11441. \size normal
  11442. \emph off
  11443. \bar no
  11444. \strikeout off
  11445. \xout off
  11446. \uuline off
  11447. \uwave off
  11448. \noun off
  11449. \color none
  11450. 548
  11451. \end_layout
  11452. \end_inset
  11453. </cell>
  11454. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  11455. \begin_inset Text
  11456. \begin_layout Plain Layout
  11457. \family roman
  11458. \series medium
  11459. \shape up
  11460. \size normal
  11461. \emph off
  11462. \bar no
  11463. \strikeout off
  11464. \xout off
  11465. \uuline off
  11466. \uwave off
  11467. \noun off
  11468. \color none
  11469. 127
  11470. \end_layout
  11471. \end_inset
  11472. </cell>
  11473. </row>
  11474. </lyxtabular>
  11475. \end_inset
  11476. \end_layout
  11477. \begin_layout Plain Layout
  11478. \begin_inset Caption Standard
  11479. \begin_layout Plain Layout
  11480. \series bold
  11481. \begin_inset Argument 1
  11482. status open
  11483. \begin_layout Plain Layout
  11484. Comparison of significantly differentially expressed genes with and without
  11485. globin blocking.
  11486. \end_layout
  11487. \end_inset
  11488. \begin_inset CommandInset label
  11489. LatexCommand label
  11490. name "tab:Comparison-of-significant"
  11491. \end_inset
  11492. Comparison of significantly differentially expressed genes with and without
  11493. globin blocking.
  11494. \series default
  11495. Up, Down: Genes significantly up/down-regulated in post-transplant samples
  11496. relative to pre-transplant samples, with a false discovery rate of 10%
  11497. or less.
  11498. NS: Non-significant genes (false discovery rate greater than 10%).
  11499. \end_layout
  11500. \end_inset
  11501. \end_layout
  11502. \begin_layout Plain Layout
  11503. \end_layout
  11504. \end_inset
  11505. \end_layout
  11506. \begin_layout Standard
  11507. To compare performance on differential gene expression tests, we took subsets
  11508. of both the GB and non-GB libraries with exactly one pre-transplant and
  11509. one post-transplant sample for each animal that had paired samples available
  11510. for analysis (N=7 animals, N=14 samples in each subset).
  11511. The same test for pre- vs.
  11512. post-transplant differential gene expression was performed on the same
  11513. 7 pairs of samples from GB libraries and non-GB libraries, in each case
  11514. using an FDR of 10% as the threshold of significance.
  11515. Out of 12954 genes that passed the detection threshold in both subsets,
  11516. 358 were called significantly differentially expressed in the same direction
  11517. in both sets; 1063 were differentially expressed in the GB set only; 296
  11518. were differentially expressed in the non-GB set only; 2 genes were called
  11519. significantly up in the GB set but significantly down in the non-GB set;
  11520. and the remaining 11235 were not called differentially expressed in either
  11521. set.
  11522. These data are summarized in Table
  11523. \begin_inset CommandInset ref
  11524. LatexCommand ref
  11525. reference "tab:Comparison-of-significant"
  11526. plural "false"
  11527. caps "false"
  11528. noprefix "false"
  11529. \end_inset
  11530. .
  11531. The differences in BCV calculated by EdgeR for these subsets of samples
  11532. were negligible (BCV = 0.302 for GB and 0.297 for non-GB).
  11533. \end_layout
  11534. \begin_layout Standard
  11535. The key point is that the GB data results in substantially more differentially
  11536. expressed calls than the non-GB data.
  11537. Since there is no gold standard for this dataset, it is impossible to be
  11538. certain whether this is due to under-calling of differential expression
  11539. in the non-GB samples or over-calling in the GB samples.
  11540. However, given that both datasets are derived from the same biological
  11541. samples and have nearly equal BCVs, it is more likely that the larger number
  11542. of DE calls in the GB samples are genuine detections that were enabled
  11543. by the higher sequencing depth and measurement precision of the GB samples.
  11544. Note that the same set of genes was considered in both subsets, so the
  11545. larger number of differentially expressed gene calls in the GB data set
  11546. reflects a greater sensitivity to detect significant differential gene
  11547. expression and not simply the larger total number of detected genes in
  11548. GB samples described earlier.
  11549. \end_layout
  11550. \begin_layout Section
  11551. Discussion
  11552. \end_layout
  11553. \begin_layout Standard
  11554. The original experience with whole blood gene expression profiling on DNA
  11555. microarrays demonstrated that the high concentration of globin transcripts
  11556. reduced the sensitivity to detect genes with relatively low expression
  11557. levels, in effect, significantly reducing the sensitivity.
  11558. To address this limitation, commercial protocols for globin reduction were
  11559. developed based on strategies to block globin transcript amplification
  11560. during labeling or physically removing globin transcripts by affinity bead
  11561. methods
  11562. \begin_inset CommandInset citation
  11563. LatexCommand cite
  11564. key "Winn2010"
  11565. literal "false"
  11566. \end_inset
  11567. .
  11568. More recently, using the latest generation of labeling protocols and arrays,
  11569. it was determined that globin reduction was no longer necessary to obtain
  11570. sufficient sensitivity to detect differential transcript expression
  11571. \begin_inset CommandInset citation
  11572. LatexCommand cite
  11573. key "NuGEN2010"
  11574. literal "false"
  11575. \end_inset
  11576. .
  11577. However, we are not aware of any publications using these currently available
  11578. protocols the with latest generation of microarrays that actually compare
  11579. the detection sensitivity with and without globin reduction.
  11580. However, in practice this has now been adopted generally primarily driven
  11581. by concerns for cost control.
  11582. The main objective of our work was to directly test the impact of globin
  11583. gene transcripts and a new globin blocking protocol for application to
  11584. the newest generation of differential gene expression profiling determined
  11585. using next generation sequencing.
  11586. \end_layout
  11587. \begin_layout Standard
  11588. The challenge of doing global gene expression profiling in cynomolgus monkeys
  11589. is that the current available arrays were never designed to comprehensively
  11590. cover this genome and have not been updated since the first assemblies
  11591. of the cynomolgus genome were published.
  11592. Therefore, we determined that the best strategy for peripheral blood profiling
  11593. was to do deep RNA-seq and inform the workflow using the latest available
  11594. genome assembly and annotation
  11595. \begin_inset CommandInset citation
  11596. LatexCommand cite
  11597. key "Wilson2013"
  11598. literal "false"
  11599. \end_inset
  11600. .
  11601. However, it was not immediately clear whether globin reduction was necessary
  11602. for RNA-seq or how much improvement in efficiency or sensitivity to detect
  11603. differential gene expression would be achieved for the added cost and work.
  11604. \end_layout
  11605. \begin_layout Standard
  11606. We only found one report that demonstrated that globin reduction significantly
  11607. improved the effective read yields for sequencing of human peripheral blood
  11608. cell RNA using a DeepSAGE protocol
  11609. \begin_inset CommandInset citation
  11610. LatexCommand cite
  11611. key "Mastrokolias2012"
  11612. literal "false"
  11613. \end_inset
  11614. .
  11615. The approach to DeepSAGE involves two different restriction enzymes that
  11616. purify and then tag small fragments of transcripts at specific locations
  11617. and thus, significantly reduces the complexity of the transcriptome.
  11618. Therefore, we could not determine how DeepSAGE results would translate
  11619. to the common strategy in the field for assaying the entire transcript
  11620. population by whole-transcriptome 3’-end RNA-seq.
  11621. Furthermore, if globin reduction is necessary, we also needed a globin
  11622. reduction method specific to cynomolgus globin sequences that would work
  11623. an organism for which no kit is available off the shelf.
  11624. \end_layout
  11625. \begin_layout Standard
  11626. As mentioned above, the addition of globin blocking oligos has a very small
  11627. impact on measured expression levels of gene expression.
  11628. However, this is a non-issue for the purposes of differential expression
  11629. testing, since a systematic change in a gene in all samples does not affect
  11630. relative expression levels between samples.
  11631. However, we must acknowledge that simple comparisons of gene expression
  11632. data obtained by GB and non-GB protocols are not possible without additional
  11633. normalization.
  11634. \end_layout
  11635. \begin_layout Standard
  11636. More importantly, globin blocking not only nearly doubles the yield of usable
  11637. reads, it also increases inter-sample correlation and sensitivity to detect
  11638. differential gene expression relative to the same set of samples profiled
  11639. without blocking.
  11640. In addition, globin blocking does not add a significant amount of random
  11641. noise to the data.
  11642. Globin blocking thus represents a cost-effective way to squeeze more data
  11643. and statistical power out of the same blood samples and the same amount
  11644. of sequencing.
  11645. In conclusion, globin reduction greatly increases the yield of useful RNA-seq
  11646. reads mapping to the rest of the genome, with minimal perturbations in
  11647. the relative levels of non-globin genes.
  11648. Based on these results, globin transcript reduction using sequence-specific,
  11649. complementary blocking oligonucleotides is recommended for all deep RNA-seq
  11650. of cynomolgus and other nonhuman primate blood samples.
  11651. \end_layout
  11652. \begin_layout Section
  11653. Future Directions
  11654. \end_layout
  11655. \begin_layout Standard
  11656. One drawback of the globin blocking method presented in this analysis is
  11657. a poor yield of genic reads, only around 50%.
  11658. In a separate experiment, the reagent mixture was modified so as to address
  11659. this drawback, resulting in a method that produces an even better reduction
  11660. in globin reads without reducing the overall fraction of genic reads.
  11661. However, the data showing this improvement consists of only a few test
  11662. samples, so the larger data set analyzed above was chosen in order to demonstra
  11663. te the effectiveness of the method in reducing globin reads while preserving
  11664. the biological signal.
  11665. \end_layout
  11666. \begin_layout Standard
  11667. The motivation for developing a fast practical way to enrich for non-globin
  11668. reads in cyno blood samples was to enable a large-scale RNA-seq experiment
  11669. investigating the effects of mesenchymal stem cell infusion on blood gene
  11670. expression in cynomologus transplant recipients in a time course after
  11671. transplantation.
  11672. With the globin blocking method in place, the way is now clear for this
  11673. experiment to proceed.
  11674. \end_layout
  11675. \begin_layout Chapter
  11676. Future Directions
  11677. \end_layout
  11678. \begin_layout Standard
  11679. \begin_inset Flex TODO Note (inline)
  11680. status open
  11681. \begin_layout Plain Layout
  11682. If there are any chapter-independent future directions, put them here.
  11683. Otherwise, delete this section.
  11684. Check in the directions if this is OK.
  11685. \end_layout
  11686. \end_inset
  11687. \end_layout
  11688. \begin_layout Chapter
  11689. Closing remarks
  11690. \end_layout
  11691. \begin_layout Standard
  11692. \begin_inset ERT
  11693. status collapsed
  11694. \begin_layout Plain Layout
  11695. % Use "References" as the title of the Bibliography
  11696. \end_layout
  11697. \begin_layout Plain Layout
  11698. \backslash
  11699. renewcommand{
  11700. \backslash
  11701. bibname}{References}
  11702. \end_layout
  11703. \end_inset
  11704. \end_layout
  11705. \begin_layout Standard
  11706. \begin_inset CommandInset bibtex
  11707. LatexCommand bibtex
  11708. btprint "btPrintCited"
  11709. bibfiles "code-refs,refs-PROCESSED"
  11710. options "bibtotoc,unsrt"
  11711. \end_inset
  11712. \end_layout
  11713. \begin_layout Standard
  11714. \begin_inset Flex TODO Note (inline)
  11715. status open
  11716. \begin_layout Plain Layout
  11717. Check bib entry formatting & sort order
  11718. \end_layout
  11719. \end_inset
  11720. \end_layout
  11721. \begin_layout Standard
  11722. \begin_inset Flex TODO Note (inline)
  11723. status open
  11724. \begin_layout Plain Layout
  11725. Check in-text citation format.
  11726. Probably don't just want [1], [2], etc.
  11727. \end_layout
  11728. \end_inset
  11729. \end_layout
  11730. \end_body
  11731. \end_document