thesis.lyx 226 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687688689690691692693694695696697698699700701702703704705706707708709710711712713714715716717718719720721722723724725726727728729730731732733734735736737738739740741742743744745746747748749750751752753754755756757758759760761762763764765766767768769770771772773774775776777778779780781782783784785786787788789790791792793794795796797798799800801802803804805806807808809810811812813814815816817818819820821822823824825826827828829830831832833834835836837838839840841842843844845846847848849850851852853854855856857858859860861862863864865866867868869870871872873874875876877878879880881882883884885886887888889890891892893894895896897898899900901902903904905906907908909910911912913914915916917918919920921922923924925926927928929930931932933934935936937938939940941942943944945946947948949950951952953954955956957958959960961962963964965966967968969970971972973974975976977978979980981982983984985986987988989990991992993994995996997998999100010011002100310041005100610071008100910101011101210131014101510161017101810191020102110221023102410251026102710281029103010311032103310341035103610371038103910401041104210431044104510461047104810491050105110521053105410551056105710581059106010611062106310641065106610671068106910701071107210731074107510761077107810791080108110821083108410851086108710881089109010911092109310941095109610971098109911001101110211031104110511061107110811091110111111121113111411151116111711181119112011211122112311241125112611271128112911301131113211331134113511361137113811391140114111421143114411451146114711481149115011511152115311541155115611571158115911601161116211631164116511661167116811691170117111721173117411751176117711781179118011811182118311841185118611871188118911901191119211931194119511961197119811991200120112021203120412051206120712081209121012111212121312141215121612171218121912201221122212231224122512261227122812291230123112321233123412351236123712381239124012411242124312441245124612471248124912501251125212531254125512561257125812591260126112621263126412651266126712681269127012711272127312741275127612771278127912801281128212831284128512861287128812891290129112921293129412951296129712981299130013011302130313041305130613071308130913101311131213131314131513161317131813191320132113221323132413251326132713281329133013311332133313341335133613371338133913401341134213431344134513461347134813491350135113521353135413551356135713581359136013611362136313641365136613671368136913701371137213731374137513761377137813791380138113821383138413851386138713881389139013911392139313941395139613971398139914001401140214031404140514061407140814091410141114121413141414151416141714181419142014211422142314241425142614271428142914301431143214331434143514361437143814391440144114421443144414451446144714481449145014511452145314541455145614571458145914601461146214631464146514661467146814691470147114721473147414751476147714781479148014811482148314841485148614871488148914901491149214931494149514961497149814991500150115021503150415051506150715081509151015111512151315141515151615171518151915201521152215231524152515261527152815291530153115321533153415351536153715381539154015411542154315441545154615471548154915501551155215531554155515561557155815591560156115621563156415651566156715681569157015711572157315741575157615771578157915801581158215831584158515861587158815891590159115921593159415951596159715981599160016011602160316041605160616071608160916101611161216131614161516161617161816191620162116221623162416251626162716281629163016311632163316341635163616371638163916401641164216431644164516461647164816491650165116521653165416551656165716581659166016611662166316641665166616671668166916701671167216731674167516761677167816791680168116821683168416851686168716881689169016911692169316941695169616971698169917001701170217031704170517061707170817091710171117121713171417151716171717181719172017211722172317241725172617271728172917301731173217331734173517361737173817391740174117421743174417451746174717481749175017511752175317541755175617571758175917601761176217631764176517661767176817691770177117721773177417751776177717781779178017811782178317841785178617871788178917901791179217931794179517961797179817991800180118021803180418051806180718081809181018111812181318141815181618171818181918201821182218231824182518261827182818291830183118321833183418351836183718381839184018411842184318441845184618471848184918501851185218531854185518561857185818591860186118621863186418651866186718681869187018711872187318741875187618771878187918801881188218831884188518861887188818891890189118921893189418951896189718981899190019011902190319041905190619071908190919101911191219131914191519161917191819191920192119221923192419251926192719281929193019311932193319341935193619371938193919401941194219431944194519461947194819491950195119521953195419551956195719581959196019611962196319641965196619671968196919701971197219731974197519761977197819791980198119821983198419851986198719881989199019911992199319941995199619971998199920002001200220032004200520062007200820092010201120122013201420152016201720182019202020212022202320242025202620272028202920302031203220332034203520362037203820392040204120422043204420452046204720482049205020512052205320542055205620572058205920602061206220632064206520662067206820692070207120722073207420752076207720782079208020812082208320842085208620872088208920902091209220932094209520962097209820992100210121022103210421052106210721082109211021112112211321142115211621172118211921202121212221232124212521262127212821292130213121322133213421352136213721382139214021412142214321442145214621472148214921502151215221532154215521562157215821592160216121622163216421652166216721682169217021712172217321742175217621772178217921802181218221832184218521862187218821892190219121922193219421952196219721982199220022012202220322042205220622072208220922102211221222132214221522162217221822192220222122222223222422252226222722282229223022312232223322342235223622372238223922402241224222432244224522462247224822492250225122522253225422552256225722582259226022612262226322642265226622672268226922702271227222732274227522762277227822792280228122822283228422852286228722882289229022912292229322942295229622972298229923002301230223032304230523062307230823092310231123122313231423152316231723182319232023212322232323242325232623272328232923302331233223332334233523362337233823392340234123422343234423452346234723482349235023512352235323542355235623572358235923602361236223632364236523662367236823692370237123722373237423752376237723782379238023812382238323842385238623872388238923902391239223932394239523962397239823992400240124022403240424052406240724082409241024112412241324142415241624172418241924202421242224232424242524262427242824292430243124322433243424352436243724382439244024412442244324442445244624472448244924502451245224532454245524562457245824592460246124622463246424652466246724682469247024712472247324742475247624772478247924802481248224832484248524862487248824892490249124922493249424952496249724982499250025012502250325042505250625072508250925102511251225132514251525162517251825192520252125222523252425252526252725282529253025312532253325342535253625372538253925402541254225432544254525462547254825492550255125522553255425552556255725582559256025612562256325642565256625672568256925702571257225732574257525762577257825792580258125822583258425852586258725882589259025912592259325942595259625972598259926002601260226032604260526062607260826092610261126122613261426152616261726182619262026212622262326242625262626272628262926302631263226332634263526362637263826392640264126422643264426452646264726482649265026512652265326542655265626572658265926602661266226632664266526662667266826692670267126722673267426752676267726782679268026812682268326842685268626872688268926902691269226932694269526962697269826992700270127022703270427052706270727082709271027112712271327142715271627172718271927202721272227232724272527262727272827292730273127322733273427352736273727382739274027412742274327442745274627472748274927502751275227532754275527562757275827592760276127622763276427652766276727682769277027712772277327742775277627772778277927802781278227832784278527862787278827892790279127922793279427952796279727982799280028012802280328042805280628072808280928102811281228132814281528162817281828192820282128222823282428252826282728282829283028312832283328342835283628372838283928402841284228432844284528462847284828492850285128522853285428552856285728582859286028612862286328642865286628672868286928702871287228732874287528762877287828792880288128822883288428852886288728882889289028912892289328942895289628972898289929002901290229032904290529062907290829092910291129122913291429152916291729182919292029212922292329242925292629272928292929302931293229332934293529362937293829392940294129422943294429452946294729482949295029512952295329542955295629572958295929602961296229632964296529662967296829692970297129722973297429752976297729782979298029812982298329842985298629872988298929902991299229932994299529962997299829993000300130023003300430053006300730083009301030113012301330143015301630173018301930203021302230233024302530263027302830293030303130323033303430353036303730383039304030413042304330443045304630473048304930503051305230533054305530563057305830593060306130623063306430653066306730683069307030713072307330743075307630773078307930803081308230833084308530863087308830893090309130923093309430953096309730983099310031013102310331043105310631073108310931103111311231133114311531163117311831193120312131223123312431253126312731283129313031313132313331343135313631373138313931403141314231433144314531463147314831493150315131523153315431553156315731583159316031613162316331643165316631673168316931703171317231733174317531763177317831793180318131823183318431853186318731883189319031913192319331943195319631973198319932003201320232033204320532063207320832093210321132123213321432153216321732183219322032213222322332243225322632273228322932303231323232333234323532363237323832393240324132423243324432453246324732483249325032513252325332543255325632573258325932603261326232633264326532663267326832693270327132723273327432753276327732783279328032813282328332843285328632873288328932903291329232933294329532963297329832993300330133023303330433053306330733083309331033113312331333143315331633173318331933203321332233233324332533263327332833293330333133323333333433353336333733383339334033413342334333443345334633473348334933503351335233533354335533563357335833593360336133623363336433653366336733683369337033713372337333743375337633773378337933803381338233833384338533863387338833893390339133923393339433953396339733983399340034013402340334043405340634073408340934103411341234133414341534163417341834193420342134223423342434253426342734283429343034313432343334343435343634373438343934403441344234433444344534463447344834493450345134523453345434553456345734583459346034613462346334643465346634673468346934703471347234733474347534763477347834793480348134823483348434853486348734883489349034913492349334943495349634973498349935003501350235033504350535063507350835093510351135123513351435153516351735183519352035213522352335243525352635273528352935303531353235333534353535363537353835393540354135423543354435453546354735483549355035513552355335543555355635573558355935603561356235633564356535663567356835693570357135723573357435753576357735783579358035813582358335843585358635873588358935903591359235933594359535963597359835993600360136023603360436053606360736083609361036113612361336143615361636173618361936203621362236233624362536263627362836293630363136323633363436353636363736383639364036413642364336443645364636473648364936503651365236533654365536563657365836593660366136623663366436653666366736683669367036713672367336743675367636773678367936803681368236833684368536863687368836893690369136923693369436953696369736983699370037013702370337043705370637073708370937103711371237133714371537163717371837193720372137223723372437253726372737283729373037313732373337343735373637373738373937403741374237433744374537463747374837493750375137523753375437553756375737583759376037613762376337643765376637673768376937703771377237733774377537763777377837793780378137823783378437853786378737883789379037913792379337943795379637973798379938003801380238033804380538063807380838093810381138123813381438153816381738183819382038213822382338243825382638273828382938303831383238333834383538363837383838393840384138423843384438453846384738483849385038513852385338543855385638573858385938603861386238633864386538663867386838693870387138723873387438753876387738783879388038813882388338843885388638873888388938903891389238933894389538963897389838993900390139023903390439053906390739083909391039113912391339143915391639173918391939203921392239233924392539263927392839293930393139323933393439353936393739383939394039413942394339443945394639473948394939503951395239533954395539563957395839593960396139623963396439653966396739683969397039713972397339743975397639773978397939803981398239833984398539863987398839893990399139923993399439953996399739983999400040014002400340044005400640074008400940104011401240134014401540164017401840194020402140224023402440254026402740284029403040314032403340344035403640374038403940404041404240434044404540464047404840494050405140524053405440554056405740584059406040614062406340644065406640674068406940704071407240734074407540764077407840794080408140824083408440854086408740884089409040914092409340944095409640974098409941004101410241034104410541064107410841094110411141124113411441154116411741184119412041214122412341244125412641274128412941304131413241334134413541364137413841394140414141424143414441454146414741484149415041514152415341544155415641574158415941604161416241634164416541664167416841694170417141724173417441754176417741784179418041814182418341844185418641874188418941904191419241934194419541964197419841994200420142024203420442054206420742084209421042114212421342144215421642174218421942204221422242234224422542264227422842294230423142324233423442354236423742384239424042414242424342444245424642474248424942504251425242534254425542564257425842594260426142624263426442654266426742684269427042714272427342744275427642774278427942804281428242834284428542864287428842894290429142924293429442954296429742984299430043014302430343044305430643074308430943104311431243134314431543164317431843194320432143224323432443254326432743284329433043314332433343344335433643374338433943404341434243434344434543464347434843494350435143524353435443554356435743584359436043614362436343644365436643674368436943704371437243734374437543764377437843794380438143824383438443854386438743884389439043914392439343944395439643974398439944004401440244034404440544064407440844094410441144124413441444154416441744184419442044214422442344244425442644274428442944304431443244334434443544364437443844394440444144424443444444454446444744484449445044514452445344544455445644574458445944604461446244634464446544664467446844694470447144724473447444754476447744784479448044814482448344844485448644874488448944904491449244934494449544964497449844994500450145024503450445054506450745084509451045114512451345144515451645174518451945204521452245234524452545264527452845294530453145324533453445354536453745384539454045414542454345444545454645474548454945504551455245534554455545564557455845594560456145624563456445654566456745684569457045714572457345744575457645774578457945804581458245834584458545864587458845894590459145924593459445954596459745984599460046014602460346044605460646074608460946104611461246134614461546164617461846194620462146224623462446254626462746284629463046314632463346344635463646374638463946404641464246434644464546464647464846494650465146524653465446554656465746584659466046614662466346644665466646674668466946704671467246734674467546764677467846794680468146824683468446854686468746884689469046914692469346944695469646974698469947004701470247034704470547064707470847094710471147124713471447154716471747184719472047214722472347244725472647274728472947304731473247334734473547364737473847394740474147424743474447454746474747484749475047514752475347544755475647574758475947604761476247634764476547664767476847694770477147724773477447754776477747784779478047814782478347844785478647874788478947904791479247934794479547964797479847994800480148024803480448054806480748084809481048114812481348144815481648174818481948204821482248234824482548264827482848294830483148324833483448354836483748384839484048414842484348444845484648474848484948504851485248534854485548564857485848594860486148624863486448654866486748684869487048714872487348744875487648774878487948804881488248834884488548864887488848894890489148924893489448954896489748984899490049014902490349044905490649074908490949104911491249134914491549164917491849194920492149224923492449254926492749284929493049314932493349344935493649374938493949404941494249434944494549464947494849494950495149524953495449554956495749584959496049614962496349644965496649674968496949704971497249734974497549764977497849794980498149824983498449854986498749884989499049914992499349944995499649974998499950005001500250035004500550065007500850095010501150125013501450155016501750185019502050215022502350245025502650275028502950305031503250335034503550365037503850395040504150425043504450455046504750485049505050515052505350545055505650575058505950605061506250635064506550665067506850695070507150725073507450755076507750785079508050815082508350845085508650875088508950905091509250935094509550965097509850995100510151025103510451055106510751085109511051115112511351145115511651175118511951205121512251235124512551265127512851295130513151325133513451355136513751385139514051415142514351445145514651475148514951505151515251535154515551565157515851595160516151625163516451655166516751685169517051715172517351745175517651775178517951805181518251835184518551865187518851895190519151925193519451955196519751985199520052015202520352045205520652075208520952105211521252135214521552165217521852195220522152225223522452255226522752285229523052315232523352345235523652375238523952405241524252435244524552465247524852495250525152525253525452555256525752585259526052615262526352645265526652675268526952705271527252735274527552765277527852795280528152825283528452855286528752885289529052915292529352945295529652975298529953005301530253035304530553065307530853095310531153125313531453155316531753185319532053215322532353245325532653275328532953305331533253335334533553365337533853395340534153425343534453455346534753485349535053515352535353545355535653575358535953605361536253635364536553665367536853695370537153725373537453755376537753785379538053815382538353845385538653875388538953905391539253935394539553965397539853995400540154025403540454055406540754085409541054115412541354145415541654175418541954205421542254235424542554265427542854295430543154325433543454355436543754385439544054415442544354445445544654475448544954505451545254535454545554565457545854595460546154625463546454655466546754685469547054715472547354745475547654775478547954805481548254835484548554865487548854895490549154925493549454955496549754985499550055015502550355045505550655075508550955105511551255135514551555165517551855195520552155225523552455255526552755285529553055315532553355345535553655375538553955405541554255435544554555465547554855495550555155525553555455555556555755585559556055615562556355645565556655675568556955705571557255735574557555765577557855795580558155825583558455855586558755885589559055915592559355945595559655975598559956005601560256035604560556065607560856095610561156125613561456155616561756185619562056215622562356245625562656275628562956305631563256335634563556365637563856395640564156425643564456455646564756485649565056515652565356545655565656575658565956605661566256635664566556665667566856695670567156725673567456755676567756785679568056815682568356845685568656875688568956905691569256935694569556965697569856995700570157025703570457055706570757085709571057115712571357145715571657175718571957205721572257235724572557265727572857295730573157325733573457355736573757385739574057415742574357445745574657475748574957505751575257535754575557565757575857595760576157625763576457655766576757685769577057715772577357745775577657775778577957805781578257835784578557865787578857895790579157925793579457955796579757985799580058015802580358045805580658075808580958105811581258135814581558165817581858195820582158225823582458255826582758285829583058315832583358345835583658375838583958405841584258435844584558465847584858495850585158525853585458555856585758585859586058615862586358645865586658675868586958705871587258735874587558765877587858795880588158825883588458855886588758885889589058915892589358945895589658975898589959005901590259035904590559065907590859095910591159125913591459155916591759185919592059215922592359245925592659275928592959305931593259335934593559365937593859395940594159425943594459455946594759485949595059515952595359545955595659575958595959605961596259635964596559665967596859695970597159725973597459755976597759785979598059815982598359845985598659875988598959905991599259935994599559965997599859996000600160026003600460056006600760086009601060116012601360146015601660176018601960206021602260236024602560266027602860296030603160326033603460356036603760386039604060416042604360446045604660476048604960506051605260536054605560566057605860596060606160626063606460656066606760686069607060716072607360746075607660776078607960806081608260836084608560866087608860896090609160926093609460956096609760986099610061016102610361046105610661076108610961106111611261136114611561166117611861196120612161226123612461256126612761286129613061316132613361346135613661376138613961406141614261436144614561466147614861496150615161526153615461556156615761586159616061616162616361646165616661676168616961706171617261736174617561766177617861796180618161826183618461856186618761886189619061916192619361946195619661976198619962006201620262036204620562066207620862096210621162126213621462156216621762186219622062216222622362246225622662276228622962306231623262336234623562366237623862396240624162426243624462456246624762486249625062516252625362546255625662576258625962606261626262636264626562666267626862696270627162726273627462756276627762786279628062816282628362846285628662876288628962906291629262936294629562966297629862996300630163026303630463056306630763086309631063116312631363146315631663176318631963206321632263236324632563266327632863296330633163326333633463356336633763386339634063416342634363446345634663476348634963506351635263536354635563566357635863596360636163626363636463656366636763686369637063716372637363746375637663776378637963806381638263836384638563866387638863896390639163926393639463956396639763986399640064016402640364046405640664076408640964106411641264136414641564166417641864196420642164226423642464256426642764286429643064316432643364346435643664376438643964406441644264436444644564466447644864496450645164526453645464556456645764586459646064616462646364646465646664676468646964706471647264736474647564766477647864796480648164826483648464856486648764886489649064916492649364946495649664976498649965006501650265036504650565066507650865096510651165126513651465156516651765186519652065216522652365246525652665276528652965306531653265336534653565366537653865396540654165426543654465456546654765486549655065516552655365546555655665576558655965606561656265636564656565666567656865696570657165726573657465756576657765786579658065816582658365846585658665876588658965906591659265936594659565966597659865996600660166026603660466056606660766086609661066116612661366146615661666176618661966206621662266236624662566266627662866296630663166326633663466356636663766386639664066416642664366446645664666476648664966506651665266536654665566566657665866596660666166626663666466656666666766686669667066716672667366746675667666776678667966806681668266836684668566866687668866896690669166926693669466956696669766986699670067016702670367046705670667076708670967106711671267136714671567166717671867196720672167226723672467256726672767286729673067316732673367346735673667376738673967406741674267436744674567466747674867496750675167526753675467556756675767586759676067616762676367646765676667676768676967706771677267736774677567766777677867796780678167826783678467856786678767886789679067916792679367946795679667976798679968006801680268036804680568066807680868096810681168126813681468156816681768186819682068216822682368246825682668276828682968306831683268336834683568366837683868396840684168426843684468456846684768486849685068516852685368546855685668576858685968606861686268636864686568666867686868696870687168726873687468756876687768786879688068816882688368846885688668876888688968906891689268936894689568966897689868996900690169026903690469056906690769086909691069116912691369146915691669176918691969206921692269236924692569266927692869296930693169326933693469356936693769386939694069416942694369446945694669476948694969506951695269536954695569566957695869596960696169626963696469656966696769686969697069716972697369746975697669776978697969806981698269836984698569866987698869896990699169926993699469956996699769986999700070017002700370047005700670077008700970107011701270137014701570167017701870197020702170227023702470257026702770287029703070317032703370347035703670377038703970407041704270437044704570467047704870497050705170527053705470557056705770587059706070617062706370647065706670677068706970707071707270737074707570767077707870797080708170827083708470857086708770887089709070917092709370947095709670977098709971007101710271037104710571067107710871097110711171127113711471157116711771187119712071217122712371247125712671277128712971307131713271337134713571367137713871397140714171427143714471457146714771487149715071517152715371547155715671577158715971607161716271637164716571667167716871697170717171727173717471757176717771787179718071817182718371847185718671877188718971907191719271937194719571967197719871997200720172027203720472057206720772087209721072117212721372147215721672177218721972207221722272237224722572267227722872297230723172327233723472357236723772387239724072417242724372447245724672477248724972507251725272537254725572567257725872597260726172627263726472657266726772687269727072717272727372747275727672777278727972807281728272837284728572867287728872897290729172927293729472957296729772987299730073017302730373047305730673077308730973107311731273137314731573167317731873197320732173227323732473257326732773287329733073317332733373347335733673377338733973407341734273437344734573467347734873497350735173527353735473557356735773587359736073617362736373647365736673677368736973707371737273737374737573767377737873797380738173827383738473857386738773887389739073917392739373947395739673977398739974007401740274037404740574067407740874097410741174127413741474157416741774187419742074217422742374247425742674277428742974307431743274337434743574367437743874397440744174427443744474457446744774487449745074517452745374547455745674577458745974607461746274637464746574667467746874697470747174727473747474757476747774787479748074817482748374847485748674877488748974907491749274937494749574967497749874997500750175027503750475057506750775087509751075117512751375147515751675177518751975207521752275237524752575267527752875297530753175327533753475357536753775387539754075417542754375447545754675477548754975507551755275537554755575567557755875597560756175627563756475657566756775687569757075717572757375747575757675777578757975807581758275837584758575867587758875897590759175927593759475957596759775987599760076017602760376047605760676077608760976107611761276137614761576167617761876197620762176227623762476257626762776287629763076317632763376347635763676377638763976407641764276437644764576467647764876497650765176527653765476557656765776587659766076617662766376647665766676677668766976707671767276737674767576767677767876797680768176827683768476857686768776887689769076917692769376947695769676977698769977007701770277037704770577067707770877097710771177127713771477157716771777187719772077217722772377247725772677277728772977307731773277337734773577367737773877397740774177427743774477457746774777487749775077517752775377547755775677577758775977607761776277637764776577667767776877697770777177727773777477757776777777787779778077817782778377847785778677877788778977907791779277937794779577967797779877997800780178027803780478057806780778087809781078117812781378147815781678177818781978207821782278237824782578267827782878297830783178327833783478357836783778387839784078417842784378447845784678477848784978507851785278537854785578567857785878597860786178627863786478657866786778687869787078717872787378747875787678777878787978807881788278837884788578867887788878897890789178927893789478957896789778987899790079017902790379047905790679077908790979107911791279137914791579167917791879197920792179227923792479257926792779287929793079317932793379347935793679377938793979407941794279437944794579467947794879497950795179527953795479557956795779587959796079617962796379647965796679677968796979707971797279737974797579767977797879797980798179827983798479857986798779887989799079917992799379947995799679977998799980008001800280038004800580068007800880098010801180128013801480158016801780188019802080218022802380248025802680278028802980308031803280338034803580368037803880398040804180428043804480458046804780488049805080518052805380548055805680578058805980608061806280638064806580668067806880698070807180728073807480758076807780788079808080818082808380848085808680878088808980908091809280938094809580968097809880998100810181028103810481058106810781088109811081118112811381148115811681178118811981208121812281238124812581268127812881298130813181328133813481358136813781388139814081418142814381448145814681478148814981508151815281538154815581568157815881598160816181628163816481658166816781688169817081718172817381748175817681778178817981808181818281838184818581868187818881898190819181928193819481958196819781988199820082018202820382048205820682078208820982108211821282138214821582168217821882198220822182228223822482258226822782288229823082318232823382348235823682378238823982408241824282438244824582468247824882498250825182528253825482558256825782588259826082618262826382648265826682678268826982708271827282738274827582768277827882798280828182828283828482858286828782888289829082918292829382948295829682978298829983008301830283038304830583068307830883098310831183128313831483158316831783188319832083218322832383248325832683278328832983308331833283338334833583368337833883398340834183428343834483458346834783488349835083518352835383548355835683578358835983608361836283638364836583668367836883698370837183728373837483758376837783788379838083818382838383848385838683878388838983908391839283938394839583968397839883998400840184028403840484058406840784088409841084118412841384148415841684178418841984208421842284238424842584268427842884298430843184328433843484358436843784388439844084418442844384448445844684478448844984508451845284538454845584568457845884598460846184628463846484658466846784688469847084718472847384748475847684778478847984808481848284838484848584868487848884898490849184928493849484958496849784988499850085018502850385048505850685078508850985108511851285138514851585168517851885198520852185228523852485258526852785288529853085318532853385348535853685378538853985408541854285438544854585468547854885498550855185528553855485558556855785588559856085618562856385648565856685678568856985708571857285738574857585768577857885798580858185828583858485858586858785888589859085918592859385948595859685978598859986008601860286038604860586068607860886098610861186128613861486158616861786188619862086218622862386248625862686278628862986308631863286338634863586368637863886398640864186428643864486458646864786488649865086518652865386548655865686578658865986608661866286638664866586668667866886698670867186728673867486758676867786788679868086818682868386848685868686878688868986908691869286938694869586968697869886998700870187028703870487058706870787088709871087118712871387148715871687178718871987208721872287238724872587268727872887298730873187328733873487358736873787388739874087418742874387448745874687478748874987508751875287538754875587568757875887598760876187628763876487658766876787688769877087718772877387748775877687778778877987808781878287838784878587868787878887898790879187928793879487958796879787988799880088018802880388048805880688078808880988108811881288138814881588168817881888198820882188228823882488258826882788288829883088318832883388348835883688378838883988408841884288438844884588468847884888498850885188528853885488558856885788588859886088618862886388648865886688678868886988708871887288738874887588768877887888798880888188828883888488858886888788888889889088918892889388948895889688978898889989008901890289038904890589068907890889098910891189128913891489158916891789188919892089218922892389248925892689278928892989308931893289338934893589368937893889398940894189428943894489458946894789488949895089518952895389548955895689578958895989608961896289638964896589668967896889698970897189728973897489758976897789788979898089818982898389848985898689878988898989908991899289938994899589968997899889999000900190029003900490059006900790089009901090119012901390149015901690179018901990209021902290239024902590269027902890299030903190329033903490359036903790389039904090419042904390449045904690479048904990509051905290539054905590569057905890599060906190629063906490659066906790689069907090719072907390749075907690779078907990809081908290839084908590869087908890899090909190929093909490959096909790989099910091019102910391049105910691079108910991109111911291139114911591169117911891199120912191229123912491259126912791289129913091319132913391349135913691379138913991409141914291439144914591469147914891499150915191529153915491559156915791589159916091619162916391649165916691679168916991709171917291739174917591769177917891799180918191829183918491859186918791889189919091919192919391949195919691979198919992009201920292039204920592069207920892099210921192129213921492159216921792189219922092219222922392249225922692279228922992309231923292339234923592369237923892399240924192429243924492459246924792489249925092519252925392549255925692579258925992609261926292639264926592669267926892699270927192729273927492759276927792789279928092819282928392849285928692879288928992909291929292939294929592969297929892999300930193029303930493059306930793089309931093119312931393149315931693179318931993209321932293239324932593269327932893299330933193329333933493359336933793389339934093419342934393449345934693479348934993509351935293539354935593569357935893599360936193629363936493659366936793689369937093719372937393749375937693779378937993809381938293839384938593869387938893899390939193929393939493959396939793989399940094019402940394049405940694079408940994109411941294139414941594169417941894199420942194229423942494259426942794289429943094319432943394349435943694379438943994409441944294439444944594469447944894499450945194529453945494559456945794589459946094619462946394649465946694679468946994709471947294739474947594769477947894799480948194829483948494859486948794889489949094919492949394949495949694979498949995009501950295039504950595069507950895099510951195129513951495159516951795189519952095219522952395249525952695279528952995309531953295339534953595369537953895399540954195429543954495459546954795489549955095519552955395549555955695579558955995609561956295639564956595669567956895699570957195729573957495759576957795789579958095819582958395849585958695879588958995909591959295939594959595969597959895999600960196029603960496059606960796089609961096119612961396149615961696179618961996209621962296239624962596269627962896299630963196329633963496359636963796389639964096419642964396449645964696479648964996509651965296539654965596569657965896599660966196629663966496659666966796689669967096719672967396749675967696779678967996809681968296839684968596869687968896899690969196929693969496959696969796989699970097019702970397049705970697079708970997109711971297139714971597169717971897199720972197229723972497259726972797289729973097319732973397349735973697379738973997409741974297439744974597469747974897499750975197529753975497559756975797589759976097619762976397649765976697679768976997709771977297739774977597769777977897799780978197829783978497859786978797889789979097919792979397949795979697979798979998009801980298039804980598069807980898099810981198129813981498159816981798189819982098219822982398249825982698279828982998309831983298339834983598369837983898399840984198429843984498459846984798489849985098519852985398549855985698579858985998609861986298639864986598669867986898699870987198729873987498759876987798789879988098819882988398849885988698879888988998909891989298939894989598969897989898999900990199029903990499059906990799089909991099119912991399149915991699179918991999209921992299239924992599269927992899299930993199329933993499359936993799389939994099419942994399449945994699479948994999509951995299539954995599569957995899599960996199629963996499659966996799689969997099719972997399749975997699779978997999809981998299839984998599869987998899899990999199929993999499959996999799989999100001000110002100031000410005100061000710008100091001010011100121001310014100151001610017100181001910020100211002210023100241002510026100271002810029100301003110032100331003410035100361003710038100391004010041100421004310044100451004610047100481004910050100511005210053100541005510056100571005810059100601006110062100631006410065100661006710068100691007010071100721007310074100751007610077100781007910080100811008210083100841008510086100871008810089100901009110092100931009410095100961009710098100991010010101101021010310104101051010610107101081010910110101111011210113101141011510116101171011810119101201012110122101231012410125101261012710128101291013010131101321013310134101351013610137101381013910140101411014210143101441014510146101471014810149101501015110152101531015410155101561015710158101591016010161101621016310164101651016610167101681016910170101711017210173101741017510176101771017810179101801018110182101831018410185101861018710188101891019010191101921019310194101951019610197101981019910200102011020210203102041020510206102071020810209102101021110212102131021410215102161021710218102191022010221102221022310224102251022610227102281022910230102311023210233102341023510236102371023810239102401024110242102431024410245102461024710248102491025010251102521025310254102551025610257102581025910260102611026210263102641026510266102671026810269102701027110272102731027410275102761027710278102791028010281102821028310284102851028610287102881028910290102911029210293102941029510296102971029810299103001030110302103031030410305103061030710308103091031010311103121031310314103151031610317103181031910320103211032210323103241032510326103271032810329103301033110332103331033410335103361033710338103391034010341103421034310344103451034610347103481034910350103511035210353103541035510356103571035810359103601036110362103631036410365103661036710368103691037010371103721037310374103751037610377103781037910380103811038210383103841038510386103871038810389103901039110392103931039410395103961039710398103991040010401104021040310404104051040610407104081040910410104111041210413104141041510416104171041810419104201042110422104231042410425104261042710428104291043010431104321043310434104351043610437104381043910440104411044210443104441044510446104471044810449104501045110452104531045410455104561045710458104591046010461104621046310464104651046610467104681046910470104711047210473104741047510476104771047810479104801048110482104831048410485104861048710488104891049010491104921049310494104951049610497104981049910500105011050210503105041050510506105071050810509105101051110512105131051410515105161051710518105191052010521105221052310524105251052610527105281052910530105311053210533105341053510536105371053810539105401054110542105431054410545105461054710548105491055010551105521055310554105551055610557105581055910560105611056210563105641056510566105671056810569105701057110572105731057410575105761057710578105791058010581105821058310584105851058610587105881058910590105911059210593105941059510596105971059810599106001060110602106031060410605106061060710608106091061010611106121061310614106151061610617106181061910620106211062210623106241062510626106271062810629106301063110632106331063410635106361063710638106391064010641106421064310644106451064610647106481064910650106511065210653106541065510656106571065810659106601066110662106631066410665106661066710668106691067010671106721067310674106751067610677106781067910680106811068210683106841068510686106871068810689106901069110692106931069410695106961069710698106991070010701107021070310704107051070610707107081070910710107111071210713107141071510716107171071810719107201072110722107231072410725107261072710728107291073010731107321073310734107351073610737107381073910740107411074210743107441074510746107471074810749107501075110752107531075410755107561075710758107591076010761107621076310764107651076610767107681076910770107711077210773107741077510776107771077810779107801078110782107831078410785107861078710788107891079010791107921079310794107951079610797107981079910800108011080210803108041080510806108071080810809108101081110812108131081410815108161081710818108191082010821108221082310824108251082610827108281082910830108311083210833108341083510836
  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. \SetWatermarkLightness{0.97}
  14. \SetWatermarkScale{1}
  15. % Set up required header format
  16. \usepackage{fancyhdr}
  17. \pagestyle{fancy}
  18. \renewcommand{\headrulewidth}{0pt}
  19. \rhead{}
  20. \lhead{}
  21. \rfoot{}
  22. \lfoot{}
  23. \cfoot{\thepage} % Page number bottom center
  24. % Allow FloatBarrier command
  25. \usepackage{placeins}
  26. % This one breaks subfigs so it's disabled
  27. % https://tex.stackexchange.com/questions/65680/automatically-bold-first-sentence-of-a-floats-caption
  28. \end_preamble
  29. \use_default_options true
  30. \begin_modules
  31. todonotes
  32. \end_modules
  33. \maintain_unincluded_children false
  34. \language english
  35. \language_package default
  36. \inputencoding utf8
  37. \fontencoding default
  38. \font_roman "default" "default"
  39. \font_sans "default" "default"
  40. \font_typewriter "default" "default"
  41. \font_math "auto" "auto"
  42. \font_default_family default
  43. \use_non_tex_fonts false
  44. \font_sc false
  45. \font_osf false
  46. \font_sf_scale 100 100
  47. \font_tt_scale 100 100
  48. \use_microtype false
  49. \use_dash_ligatures true
  50. \graphics default
  51. \default_output_format pdf4
  52. \output_sync 0
  53. \bibtex_command default
  54. \index_command default
  55. \paperfontsize 12
  56. \spacing double
  57. \use_hyperref true
  58. \pdf_bookmarks true
  59. \pdf_bookmarksnumbered false
  60. \pdf_bookmarksopen false
  61. \pdf_bookmarksopenlevel 1
  62. \pdf_breaklinks false
  63. \pdf_pdfborder false
  64. \pdf_colorlinks false
  65. \pdf_backref false
  66. \pdf_pdfusetitle true
  67. \papersize letterpaper
  68. \use_geometry true
  69. \use_package amsmath 1
  70. \use_package amssymb 1
  71. \use_package cancel 1
  72. \use_package esint 1
  73. \use_package mathdots 1
  74. \use_package mathtools 1
  75. \use_package mhchem 1
  76. \use_package stackrel 1
  77. \use_package stmaryrd 1
  78. \use_package undertilde 1
  79. \cite_engine basic
  80. \cite_engine_type default
  81. \biblio_style plain
  82. \use_bibtopic false
  83. \use_indices false
  84. \paperorientation portrait
  85. \suppress_date false
  86. \justification true
  87. \use_refstyle 1
  88. \use_minted 0
  89. \index Index
  90. \shortcut idx
  91. \color #008000
  92. \end_index
  93. \leftmargin 1.5in
  94. \topmargin 1in
  95. \rightmargin 1in
  96. \bottommargin 1in
  97. \secnumdepth 3
  98. \tocdepth 3
  99. \paragraph_separation indent
  100. \paragraph_indentation default
  101. \is_math_indent 0
  102. \math_numbering_side default
  103. \quotes_style english
  104. \dynamic_quotes 0
  105. \papercolumns 1
  106. \papersides 1
  107. \paperpagestyle default
  108. \tracking_changes false
  109. \output_changes false
  110. \html_math_output 0
  111. \html_css_as_file 0
  112. \html_be_strict false
  113. \end_header
  114. \begin_body
  115. \begin_layout Title
  116. Bioinformatic analysis of complex, high-throughput genomic and epigenomic
  117. data in the context of immunology and transplant rejection
  118. \end_layout
  119. \begin_layout Author
  120. A thesis presented
  121. \begin_inset Newline newline
  122. \end_inset
  123. by
  124. \begin_inset Newline newline
  125. \end_inset
  126. Ryan C.
  127. Thompson
  128. \begin_inset Newline newline
  129. \end_inset
  130. to
  131. \begin_inset Newline newline
  132. \end_inset
  133. The Scripps Research Institute Graduate Program
  134. \begin_inset Newline newline
  135. \end_inset
  136. in partial fulfillment of the requirements for the degree of
  137. \begin_inset Newline newline
  138. \end_inset
  139. Doctor of Philosophy in the subject of Biology
  140. \begin_inset Newline newline
  141. \end_inset
  142. for
  143. \begin_inset Newline newline
  144. \end_inset
  145. The Scripps Research Institute
  146. \begin_inset Newline newline
  147. \end_inset
  148. La Jolla, California
  149. \end_layout
  150. \begin_layout Date
  151. October 2019
  152. \end_layout
  153. \begin_layout Standard
  154. [Copyright notice]
  155. \end_layout
  156. \begin_layout Standard
  157. [Thesis acceptance form]
  158. \end_layout
  159. \begin_layout Standard
  160. [Dedication]
  161. \end_layout
  162. \begin_layout Standard
  163. [Acknowledgements]
  164. \end_layout
  165. \begin_layout Standard
  166. \begin_inset CommandInset toc
  167. LatexCommand tableofcontents
  168. \end_inset
  169. \end_layout
  170. \begin_layout Standard
  171. \begin_inset FloatList table
  172. \end_inset
  173. \end_layout
  174. \begin_layout Standard
  175. \begin_inset FloatList figure
  176. \end_inset
  177. \end_layout
  178. \begin_layout Standard
  179. [List of Abbreviations]
  180. \end_layout
  181. \begin_layout Standard
  182. \begin_inset Flex TODO Note (inline)
  183. status open
  184. \begin_layout Plain Layout
  185. Look into auto-generated nomenclature list: https://wiki.lyx.org/Tips/Nomenclature
  186. \end_layout
  187. \end_inset
  188. \end_layout
  189. \begin_layout List of TODOs
  190. \end_layout
  191. \begin_layout Standard
  192. [Abstract]
  193. \end_layout
  194. \begin_layout Chapter*
  195. Abstract
  196. \end_layout
  197. \begin_layout Chapter
  198. Introduction
  199. \end_layout
  200. \begin_layout Section
  201. Background & Significance
  202. \end_layout
  203. \begin_layout Subsection
  204. Biological motivation
  205. \end_layout
  206. \begin_layout Itemize
  207. Rejection is the major long-term threat to organ and tissue grafts
  208. \end_layout
  209. \begin_deeper
  210. \begin_layout Itemize
  211. Common mechanisms of rejection
  212. \end_layout
  213. \begin_layout Itemize
  214. Effective immune suppression requires monitoring for rejection and tuning
  215. \end_layout
  216. \begin_layout Itemize
  217. Current tests for rejection (tissue biopsy) are invasive and biased
  218. \end_layout
  219. \begin_layout Itemize
  220. A blood test based on microarrays would be less biased and invasive
  221. \end_layout
  222. \end_deeper
  223. \begin_layout Itemize
  224. Memory cells are resistant to immune suppression
  225. \end_layout
  226. \begin_deeper
  227. \begin_layout Itemize
  228. Mechanisms of resistance in memory cells are poorly understood
  229. \end_layout
  230. \begin_layout Itemize
  231. A better understanding of immune memory formation is needed
  232. \end_layout
  233. \end_deeper
  234. \begin_layout Itemize
  235. Mesenchymal stem cell infusion is a promising new treatment to prevent/delay
  236. rejection
  237. \end_layout
  238. \begin_deeper
  239. \begin_layout Itemize
  240. Demonstrated in mice, but not yet in primates
  241. \end_layout
  242. \begin_layout Itemize
  243. Mechanism currently unknown, but MSC are known to be immune modulatory
  244. \end_layout
  245. \end_deeper
  246. \begin_layout Subsection
  247. Overview of bioinformatic analysis methods
  248. \end_layout
  249. \begin_layout Standard
  250. An overview of all the methods used, including what problem they solve,
  251. what assumptions they make, and a basic description of how they work.
  252. \end_layout
  253. \begin_layout Itemize
  254. ChIP-seq Peak calling
  255. \end_layout
  256. \begin_deeper
  257. \begin_layout Itemize
  258. Cross-correlation analysis to determine fragment size
  259. \end_layout
  260. \begin_layout Itemize
  261. Broad vs narrow peaks
  262. \end_layout
  263. \begin_layout Itemize
  264. SICER for broad peaks
  265. \end_layout
  266. \begin_layout Itemize
  267. IDR for biologically reproducible peaks
  268. \end_layout
  269. \begin_layout Itemize
  270. csaw peak filtering guidelines for unbiased downstream analysis
  271. \end_layout
  272. \end_deeper
  273. \begin_layout Itemize
  274. Normalization is non-trivial and application-dependant
  275. \end_layout
  276. \begin_deeper
  277. \begin_layout Itemize
  278. Expression arrays: RMA & fRMA; why fRMA is needed
  279. \end_layout
  280. \begin_layout Itemize
  281. Methylation arrays: M-value transformation approximates normal data but
  282. induces heteroskedasticity
  283. \end_layout
  284. \begin_layout Itemize
  285. RNA-seq: normalize based on assumption that the average gene is not changing
  286. \end_layout
  287. \begin_layout Itemize
  288. ChIP-seq: complex with many considerations, dependent on experimental methods,
  289. biological system, and analysis goals
  290. \end_layout
  291. \end_deeper
  292. \begin_layout Itemize
  293. Limma: The standard linear modeling framework for genomics
  294. \end_layout
  295. \begin_deeper
  296. \begin_layout Itemize
  297. empirical Bayes variance modeling: limma's core feature
  298. \end_layout
  299. \begin_layout Itemize
  300. edgeR & DESeq2: Extend with negative bonomial GLM for RNA-seq and other
  301. count data
  302. \end_layout
  303. \begin_layout Itemize
  304. voom: Extend with precision weights to model mean-variance trend
  305. \end_layout
  306. \begin_layout Itemize
  307. arrayWeights and duplicateCorrelation to handle complex variance structures
  308. \end_layout
  309. \end_deeper
  310. \begin_layout Itemize
  311. sva and ComBat for batch correction
  312. \end_layout
  313. \begin_layout Itemize
  314. Factor analysis: PCA, MDS, MOFA
  315. \end_layout
  316. \begin_deeper
  317. \begin_layout Itemize
  318. Batch-corrected PCA is informative, but careful application is required
  319. to avoid bias
  320. \end_layout
  321. \end_deeper
  322. \begin_layout Itemize
  323. Gene set analysis: camera and SPIA
  324. \end_layout
  325. \begin_layout Section
  326. Innovation
  327. \end_layout
  328. \begin_layout Itemize
  329. MSC infusion to improve transplant outcomes (prevent/delay rejection)
  330. \end_layout
  331. \begin_deeper
  332. \begin_layout Itemize
  333. Characterize MSC response to interferon gamma
  334. \end_layout
  335. \begin_layout Itemize
  336. IFN-g is thought to stimulate their function
  337. \end_layout
  338. \begin_layout Itemize
  339. Test IFN-g treated MSC infusion as a therapy to delay graft rejection in
  340. cynomolgus monkeys
  341. \end_layout
  342. \begin_layout Itemize
  343. Monitor animals post-transplant using blood RNA-seq at serial time points
  344. \end_layout
  345. \end_deeper
  346. \begin_layout Itemize
  347. Investigate dynamics of histone marks in CD4 T-cell activation and memory
  348. \end_layout
  349. \begin_deeper
  350. \begin_layout Itemize
  351. Previous studies have looked at single snapshots of histone marks
  352. \end_layout
  353. \begin_layout Itemize
  354. Instead, look at changes in histone marks across activation and memory
  355. \end_layout
  356. \end_deeper
  357. \begin_layout Itemize
  358. High-throughput sequencing and microarray technologies
  359. \end_layout
  360. \begin_deeper
  361. \begin_layout Itemize
  362. Powerful methods for assaying gene expression and epigenetics across entire
  363. genomes
  364. \end_layout
  365. \begin_layout Itemize
  366. Proper analysis requires finding and exploiting systematic genome-wide trends
  367. \end_layout
  368. \end_deeper
  369. \begin_layout Chapter
  370. Reproducible genome-wide epigenetic analysis of H3K4 and H3K27 methylation
  371. in naive and memory CD4 T-cell activation
  372. \end_layout
  373. \begin_layout Standard
  374. \begin_inset Flex TODO Note (inline)
  375. status open
  376. \begin_layout Plain Layout
  377. Chapter author list: Me, Sarah, Dan
  378. \end_layout
  379. \end_inset
  380. \end_layout
  381. \begin_layout Standard
  382. \begin_inset Flex TODO Note (inline)
  383. status open
  384. \begin_layout Plain Layout
  385. Need better section titles throughout the chapter
  386. \end_layout
  387. \end_inset
  388. \end_layout
  389. \begin_layout Section
  390. Approach
  391. \end_layout
  392. \begin_layout Itemize
  393. CD4 T-cells are central to all adaptive immune responses and memory
  394. \end_layout
  395. \begin_layout Itemize
  396. H3K4 and H3K27 methylation are major epigenetic regulators of gene expression
  397. \end_layout
  398. \begin_layout Itemize
  399. Canonically, H3K4 is activating and H3K27 is inhibitory, but the reality
  400. is complex
  401. \end_layout
  402. \begin_layout Itemize
  403. Looking at these marks during CD4 activation and memory should reveal new
  404. mechanistic details
  405. \end_layout
  406. \begin_layout Itemize
  407. Test
  408. \begin_inset Quotes eld
  409. \end_inset
  410. poised promoter
  411. \begin_inset Quotes erd
  412. \end_inset
  413. hypothesis in which H3K4 and H3K27 are both methylated
  414. \end_layout
  415. \begin_layout Itemize
  416. Expand scope of analysis beyond simple promoter counts
  417. \end_layout
  418. \begin_deeper
  419. \begin_layout Itemize
  420. Analyze peaks genome-wide, including in intergenic regions
  421. \end_layout
  422. \begin_layout Itemize
  423. Analysis of coverage distribution shape within promoters, e.g.
  424. upstream vs downstream coverage
  425. \end_layout
  426. \end_deeper
  427. \begin_layout Section
  428. Methods
  429. \end_layout
  430. \begin_layout Standard
  431. \begin_inset Flex TODO Note (inline)
  432. status open
  433. \begin_layout Plain Layout
  434. Move figures that are only justifying methods into this section
  435. \end_layout
  436. \end_inset
  437. \end_layout
  438. \begin_layout Standard
  439. A reproducible workflow
  440. \begin_inset CommandInset citation
  441. LatexCommand cite
  442. key "gh-cd4-csaw"
  443. literal "false"
  444. \end_inset
  445. was written to analyze the raw ChIP-seq and RNA-seq data from previous
  446. studies
  447. \begin_inset CommandInset citation
  448. LatexCommand cite
  449. key "LaMere2016,LaMere2017"
  450. literal "true"
  451. \end_inset
  452. .
  453. Briefly, this data consists of RNA-seq and ChIP-seq from CD4 T-cells cultured
  454. from 4 donors.
  455. From each donor, naive and memory CD4 T-cells were isolated separately.
  456. Then cultures of both cells were activated [how?], and samples were taken
  457. at 4 time points: Day 0 (pre-activation), Day 1 (early activation), Day
  458. 5 (peak activation), and Day 14 (post-activation).
  459. For each combination of cell type and time point, RNA was isolated, and
  460. ChIP-seq was performed for each of 3 histone marks: H3K4me2, H3K4me3, and
  461. H3K27me3.
  462. The ChIP-seq input was also sequenced for each sample.
  463. The result was 32 samples for each assay.
  464. \end_layout
  465. \begin_layout Standard
  466. Sequence reads were retrieved from the Sequence Read Archive (SRA)
  467. \begin_inset CommandInset citation
  468. LatexCommand cite
  469. key "Leinonen2011"
  470. literal "false"
  471. \end_inset
  472. .
  473. ChIP-seq (and input) reads were aligned to CRCh38 genome assembly using
  474. Bowtie 2
  475. \begin_inset CommandInset citation
  476. LatexCommand cite
  477. key "Langmead2012,Schneider2017,gh-hg38-ref"
  478. literal "false"
  479. \end_inset
  480. .
  481. Artifact regions were annotated using a custom implementation of the GreyListCh
  482. IP algorithm, and these
  483. \begin_inset Quotes eld
  484. \end_inset
  485. greylists
  486. \begin_inset Quotes erd
  487. \end_inset
  488. were merged with the ENCODE blacklist
  489. \begin_inset CommandInset citation
  490. LatexCommand cite
  491. key "greylistchip,Amemiya2019,Dunham2012"
  492. literal "false"
  493. \end_inset
  494. .
  495. Any read or peak overlapping one of these regions was regarded as artifactual
  496. and excluded from downstream analyses.
  497. \end_layout
  498. \begin_layout Standard
  499. Peaks are called using epic, an implementation of the SICER algorithm
  500. \begin_inset CommandInset citation
  501. LatexCommand cite
  502. key "Zang2009,gh-epic"
  503. literal "false"
  504. \end_inset
  505. .
  506. Peaks are also called separately using MACS, but MACS was determined to
  507. be a poor fit for the data, and these peak calls are not used further
  508. \begin_inset CommandInset citation
  509. LatexCommand cite
  510. key "Zhang2008"
  511. literal "false"
  512. \end_inset
  513. .
  514. \end_layout
  515. \begin_layout Itemize
  516. Re-analyze previously published CD4 ChIP-seq & RNA-seq data
  517. \end_layout
  518. \begin_deeper
  519. \begin_layout Itemize
  520. Completely reimplement analysis from scratch as a reproducible workflow
  521. \end_layout
  522. \begin_layout Itemize
  523. Use newly published methods & algorithms not available during the original
  524. analysis: SICER, csaw, MOFA
  525. \begin_inset CommandInset citation
  526. LatexCommand cite
  527. key "Argelaguet2018"
  528. literal "false"
  529. \end_inset
  530. , ComBat, sva, GREAT, and more
  531. \end_layout
  532. \end_deeper
  533. \begin_layout Itemize
  534. SICER, IDR, csaw, & GREAT to call ChIP-seq peaks genome-wide, perform differenti
  535. al abundance analysis, and relate those peaks to gene expression
  536. \end_layout
  537. \begin_layout Itemize
  538. Promoter counts in sliding windows around each gene's highest-expressed
  539. TSS to investigate coverage distribution within promoters
  540. \end_layout
  541. \begin_layout Subsection
  542. RNA-seq align+quant method comparison
  543. \end_layout
  544. \begin_layout Standard
  545. \begin_inset Flex TODO Note (inline)
  546. status open
  547. \begin_layout Plain Layout
  548. Maybe fix up the excessive axis ranges for these plots?
  549. \end_layout
  550. \end_inset
  551. \end_layout
  552. \begin_layout Standard
  553. \begin_inset Float figure
  554. wide false
  555. sideways true
  556. status collapsed
  557. \begin_layout Plain Layout
  558. \align center
  559. \begin_inset Float figure
  560. wide false
  561. sideways false
  562. status collapsed
  563. \begin_layout Plain Layout
  564. \align center
  565. \begin_inset Graphics
  566. filename graphics/CD4-csaw/rnaseq-compare/ensmebl-vs-entrez-star-CROP.png
  567. lyxscale 25
  568. width 30col%
  569. groupId rna-comp-subfig
  570. \end_inset
  571. \end_layout
  572. \begin_layout Plain Layout
  573. \begin_inset Caption Standard
  574. \begin_layout Plain Layout
  575. Comparison of STAR quantification between Ensembl and Entrez gene identifiers
  576. \end_layout
  577. \end_inset
  578. \end_layout
  579. \end_inset
  580. \begin_inset space \hfill{}
  581. \end_inset
  582. \begin_inset Float figure
  583. wide false
  584. sideways false
  585. status collapsed
  586. \begin_layout Plain Layout
  587. \align center
  588. \begin_inset Graphics
  589. filename graphics/CD4-csaw/rnaseq-compare/ensmebl-vs-entrez-shoal-CROP.png
  590. lyxscale 25
  591. width 30col%
  592. groupId rna-comp-subfig
  593. \end_inset
  594. \end_layout
  595. \begin_layout Plain Layout
  596. \begin_inset Caption Standard
  597. \begin_layout Plain Layout
  598. Comparison of Salmon+Shoal quantification between Ensembl and Entrez gene
  599. identifiers
  600. \end_layout
  601. \end_inset
  602. \end_layout
  603. \end_inset
  604. \begin_inset space \hfill{}
  605. \end_inset
  606. \begin_inset Float figure
  607. wide false
  608. sideways false
  609. status collapsed
  610. \begin_layout Plain Layout
  611. \align center
  612. \begin_inset Graphics
  613. filename graphics/CD4-csaw/rnaseq-compare/star-vs-hisat2-CROP.png
  614. lyxscale 25
  615. width 30col%
  616. groupId rna-comp-subfig
  617. \end_inset
  618. \end_layout
  619. \begin_layout Plain Layout
  620. \begin_inset Caption Standard
  621. \begin_layout Plain Layout
  622. Comparison of quantification between STAR and HISAT2 for identical annotation
  623. \end_layout
  624. \end_inset
  625. \end_layout
  626. \end_inset
  627. \end_layout
  628. \begin_layout Plain Layout
  629. \align center
  630. \begin_inset Float figure
  631. wide false
  632. sideways false
  633. status collapsed
  634. \begin_layout Plain Layout
  635. \align center
  636. \begin_inset Graphics
  637. filename graphics/CD4-csaw/rnaseq-compare/star-vs-salmon-CROP.png
  638. lyxscale 25
  639. width 30col%
  640. groupId rna-comp-subfig
  641. \end_inset
  642. \end_layout
  643. \begin_layout Plain Layout
  644. \begin_inset Caption Standard
  645. \begin_layout Plain Layout
  646. Comparison of quantification between STAR and Salmon for identical annotation
  647. \end_layout
  648. \end_inset
  649. \end_layout
  650. \end_inset
  651. \begin_inset space \hfill{}
  652. \end_inset
  653. \begin_inset Float figure
  654. wide false
  655. sideways false
  656. status collapsed
  657. \begin_layout Plain Layout
  658. \align center
  659. \begin_inset Graphics
  660. filename graphics/CD4-csaw/rnaseq-compare/salmon-vs-kallisto-CROP.png
  661. lyxscale 25
  662. width 30col%
  663. groupId rna-comp-subfig
  664. \end_inset
  665. \end_layout
  666. \begin_layout Plain Layout
  667. \begin_inset Caption Standard
  668. \begin_layout Plain Layout
  669. Comparison of quantification between Salmon and Kallisto for identical annotatio
  670. n
  671. \end_layout
  672. \end_inset
  673. \end_layout
  674. \end_inset
  675. \begin_inset space \hfill{}
  676. \end_inset
  677. \begin_inset Float figure
  678. wide false
  679. sideways false
  680. status collapsed
  681. \begin_layout Plain Layout
  682. \align center
  683. \begin_inset Graphics
  684. filename graphics/CD4-csaw/rnaseq-compare/salmon-vs-shoal-CROP.png
  685. lyxscale 25
  686. width 30col%
  687. groupId rna-comp-subfig
  688. \end_inset
  689. \end_layout
  690. \begin_layout Plain Layout
  691. \begin_inset Caption Standard
  692. \begin_layout Plain Layout
  693. Comparison of quantification between Salmon with and without Shoal for identical
  694. annotation
  695. \end_layout
  696. \end_inset
  697. \end_layout
  698. \end_inset
  699. \end_layout
  700. \begin_layout Plain Layout
  701. \begin_inset Caption Standard
  702. \begin_layout Plain Layout
  703. \begin_inset CommandInset label
  704. LatexCommand label
  705. name "fig:RNA-norm-comp"
  706. \end_inset
  707. RNA-seq comparisons
  708. \end_layout
  709. \end_inset
  710. \end_layout
  711. \end_inset
  712. \end_layout
  713. \begin_layout Itemize
  714. Ultimately selected shoal as quantification, Ensembl as annotation.
  715. Why? Running downstream analyses with all quant methods and both annotations
  716. showed very little practical difference, so choice was not terribly important.
  717. Prefer shoal due to theoretical advantages.
  718. To note in discussion: reproducible workflow made it easy to do this, enabling
  719. an informed decision.
  720. \end_layout
  721. \begin_layout Subsection
  722. RNA-seq has a large confounding batch effect
  723. \end_layout
  724. \begin_layout Standard
  725. \begin_inset Float figure
  726. wide false
  727. sideways false
  728. status collapsed
  729. \begin_layout Plain Layout
  730. \begin_inset Flex TODO Note (inline)
  731. status open
  732. \begin_layout Plain Layout
  733. Just take the top row
  734. \end_layout
  735. \end_inset
  736. \end_layout
  737. \begin_layout Plain Layout
  738. \align center
  739. \begin_inset Graphics
  740. filename graphics/CD4-csaw/RNA-seq/weights-vs-covars-CROP.png
  741. lyxscale 25
  742. width 100col%
  743. groupId colwidth-raster
  744. \end_inset
  745. \end_layout
  746. \begin_layout Plain Layout
  747. \begin_inset Caption Standard
  748. \begin_layout Plain Layout
  749. \series bold
  750. \begin_inset CommandInset label
  751. LatexCommand label
  752. name "fig:RNA-seq-weights-vs-covars"
  753. \end_inset
  754. RNA-seq sample weights, grouped by experimental and technical covariates.
  755. \end_layout
  756. \end_inset
  757. \end_layout
  758. \end_inset
  759. \end_layout
  760. \begin_layout Itemize
  761. Batch 1 is garbage quality.
  762. Analyses involving batch 1 samples are expected to yield poor statistical
  763. power.
  764. \end_layout
  765. \begin_layout Standard
  766. \begin_inset Float figure
  767. wide false
  768. sideways false
  769. status collapsed
  770. \begin_layout Plain Layout
  771. \align center
  772. \begin_inset Float figure
  773. wide false
  774. sideways false
  775. status collapsed
  776. \begin_layout Plain Layout
  777. \align center
  778. \begin_inset Graphics
  779. filename graphics/CD4-csaw/RNA-seq/PCA-no-batchsub-CROP.png
  780. lyxscale 25
  781. width 75col%
  782. groupId rna-pca-subfig
  783. \end_inset
  784. \end_layout
  785. \begin_layout Plain Layout
  786. \begin_inset Caption Standard
  787. \begin_layout Plain Layout
  788. \series bold
  789. \begin_inset CommandInset label
  790. LatexCommand label
  791. name "fig:RNA-PCA-no-batchsub"
  792. \end_inset
  793. Before batch correction
  794. \end_layout
  795. \end_inset
  796. \end_layout
  797. \end_inset
  798. \end_layout
  799. \begin_layout Plain Layout
  800. \align center
  801. \begin_inset Float figure
  802. wide false
  803. sideways false
  804. status collapsed
  805. \begin_layout Plain Layout
  806. \align center
  807. \begin_inset Graphics
  808. filename graphics/CD4-csaw/RNA-seq/PCA-combat-batchsub-CROP.png
  809. lyxscale 25
  810. width 75col%
  811. groupId rna-pca-subfig
  812. \end_inset
  813. \end_layout
  814. \begin_layout Plain Layout
  815. \begin_inset Caption Standard
  816. \begin_layout Plain Layout
  817. \series bold
  818. \begin_inset CommandInset label
  819. LatexCommand label
  820. name "fig:RNA-PCA-ComBat-batchsub"
  821. \end_inset
  822. After batch correction with ComBat
  823. \end_layout
  824. \end_inset
  825. \end_layout
  826. \end_inset
  827. \end_layout
  828. \begin_layout Plain Layout
  829. \begin_inset Caption Standard
  830. \begin_layout Plain Layout
  831. \series bold
  832. \begin_inset CommandInset label
  833. LatexCommand label
  834. name "fig:RNA-PCA"
  835. \end_inset
  836. PCoA plots of RNA-seq data showing effect of batch correction.
  837. \end_layout
  838. \end_inset
  839. \end_layout
  840. \end_inset
  841. \end_layout
  842. \begin_layout Itemize
  843. RNA-seq batch effect can be partially corrected, but still induces uncorrectable
  844. biases in downstream analysis
  845. \end_layout
  846. \begin_layout Subsection
  847. ChIP-seq blacklisting is important
  848. \end_layout
  849. \begin_layout Standard
  850. \begin_inset Float figure
  851. wide false
  852. sideways false
  853. status collapsed
  854. \begin_layout Plain Layout
  855. \align center
  856. \begin_inset Float figure
  857. wide false
  858. sideways false
  859. status collapsed
  860. \begin_layout Plain Layout
  861. \align center
  862. \begin_inset Graphics
  863. filename graphics/CD4-csaw/csaw/CCF-plots-PAGE2-CROP.pdf
  864. lyxscale 50
  865. width 100col%
  866. groupId ccf-subfig
  867. \end_inset
  868. \end_layout
  869. \begin_layout Plain Layout
  870. \begin_inset Caption Standard
  871. \begin_layout Plain Layout
  872. \series bold
  873. \begin_inset CommandInset label
  874. LatexCommand label
  875. name "fig:CCF-with-blacklist"
  876. \end_inset
  877. Cross-correlation plots with blacklisted reads removed
  878. \end_layout
  879. \end_inset
  880. \end_layout
  881. \end_inset
  882. \end_layout
  883. \begin_layout Plain Layout
  884. \align center
  885. \begin_inset Float figure
  886. wide false
  887. sideways false
  888. status collapsed
  889. \begin_layout Plain Layout
  890. \align center
  891. \begin_inset Graphics
  892. filename graphics/CD4-csaw/csaw/CCF-plots-noBL-PAGE2-CROP.pdf
  893. lyxscale 50
  894. width 100col%
  895. groupId ccf-subfig
  896. \end_inset
  897. \end_layout
  898. \begin_layout Plain Layout
  899. \begin_inset Caption Standard
  900. \begin_layout Plain Layout
  901. \series bold
  902. \begin_inset CommandInset label
  903. LatexCommand label
  904. name "fig:CCF-without-blacklist"
  905. \end_inset
  906. Cross-correlation plots without removing blacklisted reads
  907. \end_layout
  908. \end_inset
  909. \end_layout
  910. \end_inset
  911. \end_layout
  912. \begin_layout Plain Layout
  913. \begin_inset Caption Standard
  914. \begin_layout Plain Layout
  915. \series bold
  916. \begin_inset CommandInset label
  917. LatexCommand label
  918. name "fig:CCF-master"
  919. \end_inset
  920. Strand cross-correlation plots for ChIP-seq data.
  921. \end_layout
  922. \end_inset
  923. \end_layout
  924. \end_inset
  925. \end_layout
  926. \begin_layout Subsection
  927. ChIP-seq peak calling
  928. \end_layout
  929. \begin_layout Standard
  930. \begin_inset Float figure
  931. wide false
  932. sideways false
  933. status collapsed
  934. \begin_layout Plain Layout
  935. \align center
  936. \begin_inset Float figure
  937. wide false
  938. sideways false
  939. status collapsed
  940. \begin_layout Plain Layout
  941. \align center
  942. \begin_inset Graphics
  943. filename graphics/CD4-csaw/IDR/D4659vsD5053_epic-PAGE1-CROP.pdf
  944. lyxscale 50
  945. width 40theight%
  946. groupId idr-rc-subfig
  947. \end_inset
  948. \end_layout
  949. \begin_layout Plain Layout
  950. \begin_inset Caption Standard
  951. \begin_layout Plain Layout
  952. Peak ranks from SICER peak caller
  953. \end_layout
  954. \end_inset
  955. \end_layout
  956. \begin_layout Plain Layout
  957. \end_layout
  958. \end_inset
  959. \end_layout
  960. \begin_layout Plain Layout
  961. \align center
  962. \begin_inset Float figure
  963. wide false
  964. sideways false
  965. status collapsed
  966. \begin_layout Plain Layout
  967. \align center
  968. \begin_inset Graphics
  969. filename graphics/CD4-csaw/IDR/D4659vsD5053_macs-PAGE1-CROP.pdf
  970. lyxscale 50
  971. width 40theight%
  972. groupId idr-rc-subfig
  973. \end_inset
  974. \end_layout
  975. \begin_layout Plain Layout
  976. \begin_inset Caption Standard
  977. \begin_layout Plain Layout
  978. Peak ranks from MACS peak caller
  979. \end_layout
  980. \end_inset
  981. \end_layout
  982. \end_inset
  983. \end_layout
  984. \begin_layout Plain Layout
  985. \begin_inset Caption Standard
  986. \begin_layout Plain Layout
  987. \series bold
  988. \begin_inset CommandInset label
  989. LatexCommand label
  990. name "fig:IDR-rank-consist"
  991. \end_inset
  992. Irreproducible Discovery Rate rank consistency plots for H3K27me3.
  993. \series default
  994. Peaks are ranked by the scores assigned by the peak caller in each donor,
  995. and then the ranks are plotted against each other.
  996. Higher ranks are more significant.
  997. Peaks meeting various thresholds of reproducibility, measured by the irreproduc
  998. ible discovery rate (IDR), are shaded accordingly.
  999. \end_layout
  1000. \end_inset
  1001. \end_layout
  1002. \begin_layout Plain Layout
  1003. \end_layout
  1004. \end_inset
  1005. \end_layout
  1006. \begin_layout Standard
  1007. Figure
  1008. \begin_inset CommandInset ref
  1009. LatexCommand ref
  1010. reference "fig:IDR-rank-consist"
  1011. plural "false"
  1012. caps "false"
  1013. noprefix "false"
  1014. \end_inset
  1015. shows the IDR rank-consistency plots for peaks called in an arbitrarily-chosen
  1016. pair of donors.
  1017. when the peaks for each donor are ranked according to their scores, SICER
  1018. produces much more reproducible results between donors.
  1019. This is consistent with SICER's stated goal of identifying broad peaks,
  1020. in contrast to MACS, which is designed for identifying sharp peaks.
  1021. Based on this observation, the SICER peak calls were used for all downstream
  1022. analyses that involved ChIP-seq peaks.
  1023. \end_layout
  1024. \begin_layout Subsection
  1025. ChIP-seq normalization
  1026. \end_layout
  1027. \begin_layout Standard
  1028. \begin_inset Float figure
  1029. wide false
  1030. sideways false
  1031. status open
  1032. \begin_layout Plain Layout
  1033. \align center
  1034. \begin_inset Graphics
  1035. filename graphics/CD4-csaw/ChIP-seq/H3K4me2-sample-MAplot-bins-CROP.png
  1036. lyxscale 25
  1037. width 100col%
  1038. groupId colwidth-raster
  1039. \end_inset
  1040. \end_layout
  1041. \begin_layout Plain Layout
  1042. \begin_inset Caption Standard
  1043. \begin_layout Plain Layout
  1044. \series bold
  1045. \begin_inset CommandInset label
  1046. LatexCommand label
  1047. name "fig:MA-plot-bigbins"
  1048. \end_inset
  1049. MA plot of H3K4me2 read counts in 10kb bins for two arbitrary samples.
  1050. \end_layout
  1051. \end_inset
  1052. \end_layout
  1053. \end_inset
  1054. \end_layout
  1055. \begin_layout Subsection
  1056. ChIP-seq must be corrected for hidden confounding factors
  1057. \end_layout
  1058. \begin_layout Standard
  1059. \begin_inset Float figure
  1060. wide false
  1061. sideways true
  1062. status open
  1063. \begin_layout Plain Layout
  1064. \begin_inset Float figure
  1065. wide false
  1066. sideways false
  1067. status collapsed
  1068. \begin_layout Plain Layout
  1069. \align center
  1070. \begin_inset Graphics
  1071. filename graphics/CD4-csaw/ChIP-seq/H3K4me2-PCA-raw-CROP.png
  1072. lyxscale 25
  1073. width 30col%
  1074. groupId pcoa-subfig
  1075. \end_inset
  1076. \end_layout
  1077. \begin_layout Plain Layout
  1078. \begin_inset Caption Standard
  1079. \begin_layout Plain Layout
  1080. \series bold
  1081. \begin_inset CommandInset label
  1082. LatexCommand label
  1083. name "fig:PCoA-H3K4me2-bad"
  1084. \end_inset
  1085. H3K4me2, no correction
  1086. \end_layout
  1087. \end_inset
  1088. \end_layout
  1089. \end_inset
  1090. \begin_inset space \hfill{}
  1091. \end_inset
  1092. \begin_inset Float figure
  1093. wide false
  1094. sideways false
  1095. status collapsed
  1096. \begin_layout Plain Layout
  1097. \align center
  1098. \begin_inset Graphics
  1099. filename graphics/CD4-csaw/ChIP-seq/H3K4me3-PCA-raw-CROP.png
  1100. lyxscale 25
  1101. width 30col%
  1102. groupId pcoa-subfig
  1103. \end_inset
  1104. \end_layout
  1105. \begin_layout Plain Layout
  1106. \begin_inset Caption Standard
  1107. \begin_layout Plain Layout
  1108. \series bold
  1109. \begin_inset CommandInset label
  1110. LatexCommand label
  1111. name "fig:PCoA-H3K4me3-bad"
  1112. \end_inset
  1113. H3K4me3, no correction
  1114. \end_layout
  1115. \end_inset
  1116. \end_layout
  1117. \end_inset
  1118. \begin_inset space \hfill{}
  1119. \end_inset
  1120. \begin_inset Float figure
  1121. wide false
  1122. sideways false
  1123. status collapsed
  1124. \begin_layout Plain Layout
  1125. \align center
  1126. \begin_inset Graphics
  1127. filename graphics/CD4-csaw/ChIP-seq/H3K27me3-PCA-raw-CROP.png
  1128. lyxscale 25
  1129. width 30col%
  1130. groupId pcoa-subfig
  1131. \end_inset
  1132. \end_layout
  1133. \begin_layout Plain Layout
  1134. \begin_inset Caption Standard
  1135. \begin_layout Plain Layout
  1136. \series bold
  1137. \begin_inset CommandInset label
  1138. LatexCommand label
  1139. name "fig:PCoA-H3K27me3-bad"
  1140. \end_inset
  1141. H3K27me3, no correction
  1142. \end_layout
  1143. \end_inset
  1144. \end_layout
  1145. \end_inset
  1146. \end_layout
  1147. \begin_layout Plain Layout
  1148. \begin_inset Float figure
  1149. wide false
  1150. sideways false
  1151. status collapsed
  1152. \begin_layout Plain Layout
  1153. \align center
  1154. \begin_inset Graphics
  1155. filename graphics/CD4-csaw/ChIP-seq/H3K4me2-PCA-SVsub-CROP.png
  1156. lyxscale 25
  1157. width 30col%
  1158. groupId pcoa-subfig
  1159. \end_inset
  1160. \end_layout
  1161. \begin_layout Plain Layout
  1162. \begin_inset Caption Standard
  1163. \begin_layout Plain Layout
  1164. \series bold
  1165. \begin_inset CommandInset label
  1166. LatexCommand label
  1167. name "fig:PCoA-H3K4me2-good"
  1168. \end_inset
  1169. H3K4me2, SVs subtracted
  1170. \end_layout
  1171. \end_inset
  1172. \end_layout
  1173. \end_inset
  1174. \begin_inset space \hfill{}
  1175. \end_inset
  1176. \begin_inset Float figure
  1177. wide false
  1178. sideways false
  1179. status collapsed
  1180. \begin_layout Plain Layout
  1181. \align center
  1182. \begin_inset Graphics
  1183. filename graphics/CD4-csaw/ChIP-seq/H3K4me3-PCA-SVsub-CROP.png
  1184. lyxscale 25
  1185. width 30col%
  1186. groupId pcoa-subfig
  1187. \end_inset
  1188. \end_layout
  1189. \begin_layout Plain Layout
  1190. \begin_inset Caption Standard
  1191. \begin_layout Plain Layout
  1192. \series bold
  1193. \begin_inset CommandInset label
  1194. LatexCommand label
  1195. name "fig:PCoA-H3K4me3-good"
  1196. \end_inset
  1197. H3K4me3 windows, SVs subtracted
  1198. \end_layout
  1199. \end_inset
  1200. \end_layout
  1201. \end_inset
  1202. \begin_inset space \hfill{}
  1203. \end_inset
  1204. \begin_inset Float figure
  1205. wide false
  1206. sideways false
  1207. status collapsed
  1208. \begin_layout Plain Layout
  1209. \align center
  1210. \begin_inset Graphics
  1211. filename graphics/CD4-csaw/ChIP-seq/H3K27me3-PCA-SVsub-CROP.png
  1212. lyxscale 25
  1213. width 30col%
  1214. groupId pcoa-subfig
  1215. \end_inset
  1216. \end_layout
  1217. \begin_layout Plain Layout
  1218. \begin_inset Caption Standard
  1219. \begin_layout Plain Layout
  1220. \series bold
  1221. \begin_inset CommandInset label
  1222. LatexCommand label
  1223. name "fig:PCoA-H3K27me3-good"
  1224. \end_inset
  1225. H3K27me3, SVs subtracted
  1226. \end_layout
  1227. \end_inset
  1228. \end_layout
  1229. \end_inset
  1230. \end_layout
  1231. \begin_layout Plain Layout
  1232. \begin_inset Caption Standard
  1233. \begin_layout Plain Layout
  1234. \series bold
  1235. \begin_inset CommandInset label
  1236. LatexCommand label
  1237. name "fig:PCoA-ChIP"
  1238. \end_inset
  1239. PCoA plots of ChIP-seq sliding window data, before and after subtracting
  1240. surrogate variables (SVs).
  1241. \end_layout
  1242. \end_inset
  1243. \end_layout
  1244. \begin_layout Plain Layout
  1245. \end_layout
  1246. \end_inset
  1247. \end_layout
  1248. \begin_layout Itemize
  1249. Figures showing BCV plots with and without SVA for each histone mark?
  1250. \end_layout
  1251. \begin_layout Subsection
  1252. MOFA recovers biologically relevant variation from blind analysis by correlating
  1253. across datasets
  1254. \end_layout
  1255. \begin_layout Standard
  1256. \begin_inset Float figure
  1257. wide false
  1258. sideways true
  1259. status collapsed
  1260. \begin_layout Plain Layout
  1261. \begin_inset Float figure
  1262. wide false
  1263. sideways false
  1264. status open
  1265. \begin_layout Plain Layout
  1266. \align center
  1267. \begin_inset Graphics
  1268. filename graphics/CD4-csaw/MOFA-varExplaiend-matrix-CROP.png
  1269. lyxscale 25
  1270. width 45col%
  1271. groupId mofa-subfig
  1272. \end_inset
  1273. \end_layout
  1274. \begin_layout Plain Layout
  1275. \begin_inset Caption Standard
  1276. \begin_layout Plain Layout
  1277. \series bold
  1278. \begin_inset CommandInset label
  1279. LatexCommand label
  1280. name "fig:mofa-varexplained"
  1281. \end_inset
  1282. Variance explained in each data set by each latent factor estimated by MOFA.
  1283. \series default
  1284. For each latent factor (LF) learned by MOFA, the variance explained by
  1285. that factor in each data set (
  1286. \begin_inset Quotes eld
  1287. \end_inset
  1288. view
  1289. \begin_inset Quotes erd
  1290. \end_inset
  1291. ) is shown by the shading of the cells in the lower section.
  1292. The upper section shows the total fraction of each data set's variance
  1293. that is explained by all LFs combined.
  1294. \end_layout
  1295. \end_inset
  1296. \end_layout
  1297. \end_inset
  1298. \begin_inset space \hfill{}
  1299. \end_inset
  1300. \begin_inset Float figure
  1301. wide false
  1302. sideways false
  1303. status open
  1304. \begin_layout Plain Layout
  1305. \align center
  1306. \begin_inset Graphics
  1307. filename graphics/CD4-csaw/MOFA-LF-scatter-CROP.png
  1308. lyxscale 25
  1309. width 45col%
  1310. groupId mofa-subfig
  1311. \end_inset
  1312. \end_layout
  1313. \begin_layout Plain Layout
  1314. \begin_inset Caption Standard
  1315. \begin_layout Plain Layout
  1316. \series bold
  1317. \begin_inset CommandInset label
  1318. LatexCommand label
  1319. name "fig:mofa-lf-scatter"
  1320. \end_inset
  1321. Scatter plots of specific pairs of MOFA latent factors.
  1322. \series default
  1323. LFs 1, 4, and 5 explain substantial variation in all data sets, so they
  1324. are plotted against each other in order to reveal patterns of variation
  1325. that are shared across all data sets.
  1326. \end_layout
  1327. \end_inset
  1328. \end_layout
  1329. \end_inset
  1330. \end_layout
  1331. \begin_layout Plain Layout
  1332. \begin_inset Caption Standard
  1333. \begin_layout Plain Layout
  1334. \series bold
  1335. \begin_inset CommandInset label
  1336. LatexCommand label
  1337. name "fig:MOFA-master"
  1338. \end_inset
  1339. MOFA latent factors separate technical confounders from
  1340. \end_layout
  1341. \end_inset
  1342. \end_layout
  1343. \end_inset
  1344. \end_layout
  1345. \begin_layout Itemize
  1346. Figure
  1347. \begin_inset CommandInset ref
  1348. LatexCommand ref
  1349. reference "fig:mofa-varexplained"
  1350. plural "false"
  1351. caps "false"
  1352. noprefix "false"
  1353. \end_inset
  1354. shows that LF1, 4, and 5 explain substantial var in all data sets
  1355. \end_layout
  1356. \begin_layout Itemize
  1357. Figure
  1358. \begin_inset CommandInset ref
  1359. LatexCommand ref
  1360. reference "fig:mofa-lf-scatter"
  1361. plural "false"
  1362. caps "false"
  1363. noprefix "false"
  1364. \end_inset
  1365. shows that those same 3 LFs, (1, 4, & 5) also correlate best with the experimen
  1366. tal factors (cell type & time point)
  1367. \end_layout
  1368. \begin_layout Itemize
  1369. LF2 is clearly the RNA-seq batch effect
  1370. \end_layout
  1371. \begin_layout Standard
  1372. \begin_inset Float figure
  1373. wide false
  1374. sideways false
  1375. status collapsed
  1376. \begin_layout Plain Layout
  1377. \align center
  1378. \begin_inset Graphics
  1379. filename graphics/CD4-csaw/MOFA-batch-correct-CROP.png
  1380. lyxscale 25
  1381. width 100col%
  1382. groupId colwidth-raster
  1383. \end_inset
  1384. \end_layout
  1385. \begin_layout Plain Layout
  1386. \begin_inset Caption Standard
  1387. \begin_layout Plain Layout
  1388. \series bold
  1389. \begin_inset CommandInset label
  1390. LatexCommand label
  1391. name "fig:mofa-batchsub"
  1392. \end_inset
  1393. Result of RNA-seq batch-correction using MOFA latent factors
  1394. \end_layout
  1395. \end_inset
  1396. \end_layout
  1397. \end_inset
  1398. \end_layout
  1399. \begin_layout Itemize
  1400. Attempting to remove the effect of LF2 (Figure
  1401. \begin_inset CommandInset ref
  1402. LatexCommand ref
  1403. reference "fig:mofa-batchsub"
  1404. plural "false"
  1405. caps "false"
  1406. noprefix "false"
  1407. \end_inset
  1408. ) results in batch correction comparable to ComBat (Figure
  1409. \begin_inset CommandInset ref
  1410. LatexCommand ref
  1411. reference "fig:RNA-PCA-ComBat-batchsub"
  1412. plural "false"
  1413. caps "false"
  1414. noprefix "false"
  1415. \end_inset
  1416. )
  1417. \end_layout
  1418. \begin_layout Itemize
  1419. MOFA was able to do this batch subtraction without directly using the sample
  1420. labels (sample labels were used implicitly to select which factor to subtract)
  1421. \end_layout
  1422. \begin_layout Itemize
  1423. Similarity of results shows that batch correction can't get much better
  1424. than ComBat (despite ComBat ignoring time point)
  1425. \end_layout
  1426. \begin_layout Subsection
  1427. MOFA does some interesting stuff but is mostly confirmatory in this context
  1428. \end_layout
  1429. \begin_layout Standard
  1430. \begin_inset Flex TODO Note (inline)
  1431. status open
  1432. \begin_layout Plain Layout
  1433. MOFA should be a footnote to something else, not its own point
  1434. \end_layout
  1435. \end_inset
  1436. \end_layout
  1437. \begin_layout Standard
  1438. \begin_inset Flex TODO Note (inline)
  1439. status open
  1440. \begin_layout Plain Layout
  1441. Combine with previous subsection
  1442. \end_layout
  1443. \end_inset
  1444. \end_layout
  1445. \begin_layout Itemize
  1446. MOFA shows great promise for accelerating discovery of major biological
  1447. effects in multi-omics datasets
  1448. \end_layout
  1449. \begin_deeper
  1450. \begin_layout Itemize
  1451. MOFA successfully separates biologically relevant patterns of variation
  1452. from technical confounding factors without knowing the sample labels, by
  1453. finding latent factors that explain variation across multiple data sets.
  1454. \end_layout
  1455. \begin_layout Itemize
  1456. MOFA was added to this analysis late and played primarily a confirmatory
  1457. role, but it was able to confirm earlier conclusions with much less prior
  1458. information (no sample labels) and much less analyst effort/input
  1459. \end_layout
  1460. \begin_layout Itemize
  1461. Less input from analyst means less opportunity to introduce unwanted bias
  1462. into results
  1463. \end_layout
  1464. \begin_layout Itemize
  1465. MOFA confirmed that the already-implemented batch correction in the RNA-seq
  1466. data was already performing as well as possible given the limitations of
  1467. the data
  1468. \end_layout
  1469. \end_deeper
  1470. \begin_layout Section
  1471. Results
  1472. \end_layout
  1473. \begin_layout Standard
  1474. \begin_inset Note Note
  1475. status open
  1476. \begin_layout Plain Layout
  1477. Focus on what hypotheses were tested, then select figures that show how
  1478. those hypotheses were tested, even if the result is a negative.
  1479. \end_layout
  1480. \begin_layout Plain Layout
  1481. Not every interesting result needs to be in here.
  1482. Chapter should tell a story.
  1483. \end_layout
  1484. \end_inset
  1485. \end_layout
  1486. \begin_layout Standard
  1487. \begin_inset Flex TODO Note (inline)
  1488. status open
  1489. \begin_layout Plain Layout
  1490. Maybe reorder these sections to do RNA-seq, then ChIP-seq, then combined
  1491. analyses?
  1492. \end_layout
  1493. \end_inset
  1494. \end_layout
  1495. \begin_layout Subsection
  1496. H3K4 and H3K27 methylation occur in broad regions and are enriched near
  1497. promoters
  1498. \end_layout
  1499. \begin_layout Standard
  1500. \begin_inset Float table
  1501. wide false
  1502. sideways false
  1503. status open
  1504. \begin_layout Plain Layout
  1505. \align center
  1506. \begin_inset Flex TODO Note (inline)
  1507. status open
  1508. \begin_layout Plain Layout
  1509. Also get
  1510. \emph on
  1511. median
  1512. \emph default
  1513. peak width and maybe other quantiles (25%, 75%)
  1514. \end_layout
  1515. \end_inset
  1516. \end_layout
  1517. \begin_layout Plain Layout
  1518. \align center
  1519. \begin_inset Tabular
  1520. <lyxtabular version="3" rows="4" columns="5">
  1521. <features tabularvalignment="middle">
  1522. <column alignment="center" valignment="top">
  1523. <column alignment="center" valignment="top">
  1524. <column alignment="center" valignment="top">
  1525. <column alignment="center" valignment="top">
  1526. <column alignment="center" valignment="top">
  1527. <row>
  1528. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  1529. \begin_inset Text
  1530. \begin_layout Plain Layout
  1531. Histone Mark
  1532. \end_layout
  1533. \end_inset
  1534. </cell>
  1535. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  1536. \begin_inset Text
  1537. \begin_layout Plain Layout
  1538. # Peaks
  1539. \end_layout
  1540. \end_inset
  1541. </cell>
  1542. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  1543. \begin_inset Text
  1544. \begin_layout Plain Layout
  1545. Mean peak width
  1546. \end_layout
  1547. \end_inset
  1548. </cell>
  1549. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  1550. \begin_inset Text
  1551. \begin_layout Plain Layout
  1552. genome coverage
  1553. \end_layout
  1554. \end_inset
  1555. </cell>
  1556. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  1557. \begin_inset Text
  1558. \begin_layout Plain Layout
  1559. FRiP
  1560. \end_layout
  1561. \end_inset
  1562. </cell>
  1563. </row>
  1564. <row>
  1565. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  1566. \begin_inset Text
  1567. \begin_layout Plain Layout
  1568. H3K4me2
  1569. \end_layout
  1570. \end_inset
  1571. </cell>
  1572. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  1573. \begin_inset Text
  1574. \begin_layout Plain Layout
  1575. 14965
  1576. \end_layout
  1577. \end_inset
  1578. </cell>
  1579. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  1580. \begin_inset Text
  1581. \begin_layout Plain Layout
  1582. 3970
  1583. \end_layout
  1584. \end_inset
  1585. </cell>
  1586. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  1587. \begin_inset Text
  1588. \begin_layout Plain Layout
  1589. 1.92%
  1590. \end_layout
  1591. \end_inset
  1592. </cell>
  1593. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  1594. \begin_inset Text
  1595. \begin_layout Plain Layout
  1596. 14.2%
  1597. \end_layout
  1598. \end_inset
  1599. </cell>
  1600. </row>
  1601. <row>
  1602. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  1603. \begin_inset Text
  1604. \begin_layout Plain Layout
  1605. H3K4me3
  1606. \end_layout
  1607. \end_inset
  1608. </cell>
  1609. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  1610. \begin_inset Text
  1611. \begin_layout Plain Layout
  1612. 6163
  1613. \end_layout
  1614. \end_inset
  1615. </cell>
  1616. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  1617. \begin_inset Text
  1618. \begin_layout Plain Layout
  1619. 2946
  1620. \end_layout
  1621. \end_inset
  1622. </cell>
  1623. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  1624. \begin_inset Text
  1625. \begin_layout Plain Layout
  1626. 0.588%
  1627. \end_layout
  1628. \end_inset
  1629. </cell>
  1630. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  1631. \begin_inset Text
  1632. \begin_layout Plain Layout
  1633. 6.57%
  1634. \end_layout
  1635. \end_inset
  1636. </cell>
  1637. </row>
  1638. <row>
  1639. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  1640. \begin_inset Text
  1641. \begin_layout Plain Layout
  1642. H3K27me3
  1643. \end_layout
  1644. \end_inset
  1645. </cell>
  1646. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  1647. \begin_inset Text
  1648. \begin_layout Plain Layout
  1649. 18139
  1650. \end_layout
  1651. \end_inset
  1652. </cell>
  1653. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  1654. \begin_inset Text
  1655. \begin_layout Plain Layout
  1656. 18967
  1657. \end_layout
  1658. \end_inset
  1659. </cell>
  1660. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  1661. \begin_inset Text
  1662. \begin_layout Plain Layout
  1663. 11.1%
  1664. \end_layout
  1665. \end_inset
  1666. </cell>
  1667. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  1668. \begin_inset Text
  1669. \begin_layout Plain Layout
  1670. 22.5%
  1671. \end_layout
  1672. \end_inset
  1673. </cell>
  1674. </row>
  1675. </lyxtabular>
  1676. \end_inset
  1677. \end_layout
  1678. \begin_layout Plain Layout
  1679. \begin_inset Caption Standard
  1680. \begin_layout Plain Layout
  1681. \series bold
  1682. \begin_inset CommandInset label
  1683. LatexCommand label
  1684. name "tab:peak-calling-summary"
  1685. \end_inset
  1686. Peak-calling summary.
  1687. \series default
  1688. For each histone mark, the number of peaks called using SICER at an IDR
  1689. threshold of ???, the mean width of those peaks, the fraction of the genome
  1690. covered by peaks, and the fraction of reads in peaks (FRiP).
  1691. \end_layout
  1692. \end_inset
  1693. \end_layout
  1694. \end_inset
  1695. \end_layout
  1696. \begin_layout Standard
  1697. Table
  1698. \begin_inset CommandInset ref
  1699. LatexCommand ref
  1700. reference "tab:peak-calling-summary"
  1701. plural "false"
  1702. caps "false"
  1703. noprefix "false"
  1704. \end_inset
  1705. gives a summary of the peak calling statistics for each histone mark.
  1706. Consistent with previous observations [CITATION NEEDED], all 3 histone
  1707. marks occur in broad regions spanning many consecutive nucleosomes, rather
  1708. than in sharp peaks as would be expected for a transcription factor or
  1709. other molecule that binds to specific sites.
  1710. This conclusion is further supported by Figure
  1711. \begin_inset CommandInset ref
  1712. LatexCommand ref
  1713. reference "fig:CCF-with-blacklist"
  1714. plural "false"
  1715. caps "false"
  1716. noprefix "false"
  1717. \end_inset
  1718. , in which a clear nucleosome-sized periodicity is visible in the cross-correlat
  1719. ion value for each sample, indicating that each time a given mark is present
  1720. on one histone, it is also likely to be found on adjacent histones as well.
  1721. H3K27me3 enrichment in particular is substantially more broad than either
  1722. H3K4 mark, with a mean peak width of almost 19,000 bp.
  1723. This is also reflected in the periodicity observed in Figure
  1724. \begin_inset CommandInset ref
  1725. LatexCommand ref
  1726. reference "fig:CCF-with-blacklist"
  1727. plural "false"
  1728. caps "false"
  1729. noprefix "false"
  1730. \end_inset
  1731. , which remains strong much farther out for H3K27me3 than the other marks,
  1732. showing H3K27me3 especially tends to be found on long runs of consecutive
  1733. histones.
  1734. \end_layout
  1735. \begin_layout Standard
  1736. \begin_inset Float figure
  1737. wide false
  1738. sideways false
  1739. status open
  1740. \begin_layout Plain Layout
  1741. \begin_inset Flex TODO Note (inline)
  1742. status open
  1743. \begin_layout Plain Layout
  1744. Ensure this figure uses the peak calls from the new analysis.
  1745. \end_layout
  1746. \end_inset
  1747. \end_layout
  1748. \begin_layout Plain Layout
  1749. \begin_inset Flex TODO Note (inline)
  1750. status open
  1751. \begin_layout Plain Layout
  1752. Need a control: shuffle all peaks and repeat, N times.
  1753. Do real vs shuffled control both in a top/bottom arrangement.
  1754. \end_layout
  1755. \end_inset
  1756. \end_layout
  1757. \begin_layout Plain Layout
  1758. \begin_inset Flex TODO Note (inline)
  1759. status open
  1760. \begin_layout Plain Layout
  1761. Consider counting TSS inside peaks as negative number indicating how far
  1762. \emph on
  1763. inside
  1764. \emph default
  1765. the peak the TSS is (i.e.
  1766. distance to nearest non-peak area).
  1767. \end_layout
  1768. \end_inset
  1769. \end_layout
  1770. \begin_layout Plain Layout
  1771. \begin_inset Flex TODO Note (inline)
  1772. status open
  1773. \begin_layout Plain Layout
  1774. The H3K4 part of this figure is included in
  1775. \begin_inset CommandInset citation
  1776. LatexCommand cite
  1777. key "LaMere2016"
  1778. literal "false"
  1779. \end_inset
  1780. as Fig.
  1781. S2.
  1782. Do I need to do anything about that?
  1783. \end_layout
  1784. \end_inset
  1785. \end_layout
  1786. \begin_layout Plain Layout
  1787. \align center
  1788. \begin_inset Graphics
  1789. filename graphics/CD4-csaw/Promoter Peak Distance Profile-PAGE1-CROP.pdf
  1790. lyxscale 50
  1791. width 80col%
  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:near-promoter-peak-enrich"
  1801. \end_inset
  1802. Enrichment of peaks in promoter neighborhoods.
  1803. \series default
  1804. This plot shows the distribution of distances from each annotated transcription
  1805. start site in the genome to the nearest called peak.
  1806. Each line represents one combination of histone mark, cell type, and time
  1807. point.
  1808. Distributions are smoothed using kernel density estimation [CITE?].
  1809. Transcription start sites that occur
  1810. \emph on
  1811. within
  1812. \emph default
  1813. peaks were excluded from this plot to avoid a large spike at zero that
  1814. would overshadow the rest of the distribution.
  1815. \end_layout
  1816. \end_inset
  1817. \end_layout
  1818. \end_inset
  1819. \end_layout
  1820. \begin_layout Standard
  1821. \begin_inset Float table
  1822. wide false
  1823. sideways false
  1824. status open
  1825. \begin_layout Plain Layout
  1826. \align center
  1827. \begin_inset Tabular
  1828. <lyxtabular version="3" rows="4" columns="2">
  1829. <features tabularvalignment="middle">
  1830. <column alignment="center" valignment="top">
  1831. <column alignment="center" valignment="top">
  1832. <row>
  1833. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  1834. \begin_inset Text
  1835. \begin_layout Plain Layout
  1836. Histone mark
  1837. \end_layout
  1838. \end_inset
  1839. </cell>
  1840. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  1841. \begin_inset Text
  1842. \begin_layout Plain Layout
  1843. Effective promoter radius
  1844. \end_layout
  1845. \end_inset
  1846. </cell>
  1847. </row>
  1848. <row>
  1849. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  1850. \begin_inset Text
  1851. \begin_layout Plain Layout
  1852. H3K4me2
  1853. \end_layout
  1854. \end_inset
  1855. </cell>
  1856. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  1857. \begin_inset Text
  1858. \begin_layout Plain Layout
  1859. 1 kb
  1860. \end_layout
  1861. \end_inset
  1862. </cell>
  1863. </row>
  1864. <row>
  1865. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  1866. \begin_inset Text
  1867. \begin_layout Plain Layout
  1868. H3K4me3
  1869. \end_layout
  1870. \end_inset
  1871. </cell>
  1872. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  1873. \begin_inset Text
  1874. \begin_layout Plain Layout
  1875. 1 kb
  1876. \end_layout
  1877. \end_inset
  1878. </cell>
  1879. </row>
  1880. <row>
  1881. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  1882. \begin_inset Text
  1883. \begin_layout Plain Layout
  1884. H3K27me3
  1885. \end_layout
  1886. \end_inset
  1887. </cell>
  1888. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  1889. \begin_inset Text
  1890. \begin_layout Plain Layout
  1891. 2.5 kb
  1892. \end_layout
  1893. \end_inset
  1894. </cell>
  1895. </row>
  1896. </lyxtabular>
  1897. \end_inset
  1898. \end_layout
  1899. \begin_layout Plain Layout
  1900. \begin_inset Caption Standard
  1901. \begin_layout Plain Layout
  1902. \series bold
  1903. \begin_inset CommandInset label
  1904. LatexCommand label
  1905. name "tab:effective-promoter-radius"
  1906. \end_inset
  1907. Effective promoter radius for each histone mark.
  1908. \series default
  1909. These values represent the approximate distance from transcription start
  1910. site positions within which an excess of peaks are found, as shown in Figure
  1911. \begin_inset CommandInset ref
  1912. LatexCommand ref
  1913. reference "fig:near-promoter-peak-enrich"
  1914. plural "false"
  1915. caps "false"
  1916. noprefix "false"
  1917. \end_inset
  1918. .
  1919. \end_layout
  1920. \end_inset
  1921. \end_layout
  1922. \begin_layout Plain Layout
  1923. \end_layout
  1924. \end_inset
  1925. \end_layout
  1926. \begin_layout Standard
  1927. \begin_inset Flex TODO Note (inline)
  1928. status open
  1929. \begin_layout Plain Layout
  1930. Problem: the effective promoter radius concept is an interesting result
  1931. on its own, hence its placement here.
  1932. However, it is also important in the methods section, which comes first.
  1933. What do? Refer forward to this section? Move this section to Methods?
  1934. \end_layout
  1935. \end_inset
  1936. \end_layout
  1937. \begin_layout Standard
  1938. All 3 histone marks tend to occur more often near promoter regions, as shown
  1939. in Figure
  1940. \begin_inset CommandInset ref
  1941. LatexCommand ref
  1942. reference "fig:near-promoter-peak-enrich"
  1943. plural "false"
  1944. caps "false"
  1945. noprefix "false"
  1946. \end_inset
  1947. .
  1948. The majority of each density distribution is flat, representing the background
  1949. density of peaks genome-wide.
  1950. Each distribution has a peak near zero, representing an enrichment of peaks
  1951. close transcription start site (TSS) positions relative to the remainder
  1952. of the genome.
  1953. Interestingly, the
  1954. \begin_inset Quotes eld
  1955. \end_inset
  1956. radius
  1957. \begin_inset Quotes erd
  1958. \end_inset
  1959. within which this enrichment occurs is not the same for every histone mark
  1960. (Table
  1961. \begin_inset CommandInset ref
  1962. LatexCommand ref
  1963. reference "tab:effective-promoter-radius"
  1964. plural "false"
  1965. caps "false"
  1966. noprefix "false"
  1967. \end_inset
  1968. ).
  1969. For H3K4me2 and H3K4me3, peaks are most enriched within 1
  1970. \begin_inset space ~
  1971. \end_inset
  1972. kbp of TSS positions, while for H3K27me3, enrichment is broader, extending
  1973. to 2.5
  1974. \begin_inset space ~
  1975. \end_inset
  1976. kbp.
  1977. These
  1978. \begin_inset Quotes eld
  1979. \end_inset
  1980. effective promoter radii
  1981. \begin_inset Quotes erd
  1982. \end_inset
  1983. were used to define the promoter regions for all further analyses.
  1984. \end_layout
  1985. \begin_layout Standard
  1986. \begin_inset Flex TODO Note (inline)
  1987. status open
  1988. \begin_layout Plain Layout
  1989. Consider also showing figure for distance to nearest peak center, and reference
  1990. median peak size once that is known.
  1991. \end_layout
  1992. \end_inset
  1993. \end_layout
  1994. \begin_layout Subsection
  1995. H3K4 and H3K27 promoter methylation has broadly the expected correlation
  1996. with gene expression
  1997. \end_layout
  1998. \begin_layout Standard
  1999. \begin_inset Flex TODO Note (inline)
  2000. status open
  2001. \begin_layout Plain Layout
  2002. This section can easily be cut, especially if I can't find those plots.
  2003. \end_layout
  2004. \end_inset
  2005. \end_layout
  2006. \begin_layout Itemize
  2007. H3K4 is correlated with higher expression, and H3K27 is correlated with
  2008. lower expression genome-wide
  2009. \end_layout
  2010. \begin_layout Standard
  2011. \begin_inset Flex TODO Note (inline)
  2012. status open
  2013. \begin_layout Plain Layout
  2014. Grr, gotta find these figures.
  2015. Maybe in the old analysis? At least one of these plots is definitely in
  2016. Sarah's paper.
  2017. \end_layout
  2018. \end_inset
  2019. \end_layout
  2020. \begin_layout Itemize
  2021. Figures showing these correlations: box/violin plots of expression distributions
  2022. with every combination of peak presence/absence in promoter
  2023. \end_layout
  2024. \begin_layout Itemize
  2025. Appropriate statistical tests showing significant differences in expected
  2026. directions
  2027. \end_layout
  2028. \begin_layout Subsection
  2029. RNA-seq and H3K4 methylation patterns in naive and memory show convergence
  2030. at day 14
  2031. \end_layout
  2032. \begin_layout Standard
  2033. \begin_inset Float figure
  2034. wide false
  2035. sideways false
  2036. status collapsed
  2037. \begin_layout Plain Layout
  2038. \align center
  2039. \begin_inset Float figure
  2040. wide false
  2041. sideways false
  2042. status collapsed
  2043. \begin_layout Plain Layout
  2044. \align center
  2045. \begin_inset Graphics
  2046. filename graphics/CD4-csaw/ChIP-seq/H3K4me2-promoter-PCA-group-CROP.png
  2047. lyxscale 25
  2048. width 45col%
  2049. groupId pcoa-prom-subfig
  2050. \end_inset
  2051. \end_layout
  2052. \begin_layout Plain Layout
  2053. \begin_inset Caption Standard
  2054. \begin_layout Plain Layout
  2055. \series bold
  2056. \begin_inset CommandInset label
  2057. LatexCommand label
  2058. name "fig:PCoA-H3K4me2-prom"
  2059. \end_inset
  2060. PCoA plot of H3K4me2 promoters, after subtracting surrogate variables
  2061. \end_layout
  2062. \end_inset
  2063. \end_layout
  2064. \end_inset
  2065. \begin_inset space \hfill{}
  2066. \end_inset
  2067. \begin_inset Float figure
  2068. wide false
  2069. sideways false
  2070. status collapsed
  2071. \begin_layout Plain Layout
  2072. \align center
  2073. \begin_inset Graphics
  2074. filename graphics/CD4-csaw/ChIP-seq/H3K4me3-promoter-PCA-group-CROP.png
  2075. lyxscale 25
  2076. width 45col%
  2077. groupId pcoa-prom-subfig
  2078. \end_inset
  2079. \end_layout
  2080. \begin_layout Plain Layout
  2081. \begin_inset Caption Standard
  2082. \begin_layout Plain Layout
  2083. \series bold
  2084. \begin_inset CommandInset label
  2085. LatexCommand label
  2086. name "fig:PCoA-H3K4me3-prom"
  2087. \end_inset
  2088. PCoA plot of H3K4me3 promoters, after subtracting surrogate variables
  2089. \end_layout
  2090. \end_inset
  2091. \end_layout
  2092. \end_inset
  2093. \end_layout
  2094. \begin_layout Plain Layout
  2095. \align center
  2096. \begin_inset Float figure
  2097. wide false
  2098. sideways false
  2099. status collapsed
  2100. \begin_layout Plain Layout
  2101. \align center
  2102. \begin_inset Graphics
  2103. filename graphics/CD4-csaw/ChIP-seq/H3K27me3-promoter-PCA-group-CROP.png
  2104. lyxscale 25
  2105. width 45col%
  2106. groupId pcoa-prom-subfig
  2107. \end_inset
  2108. \end_layout
  2109. \begin_layout Plain Layout
  2110. \begin_inset Caption Standard
  2111. \begin_layout Plain Layout
  2112. \series bold
  2113. \begin_inset CommandInset label
  2114. LatexCommand label
  2115. name "fig:PCoA-H3K27me3-prom"
  2116. \end_inset
  2117. PCoA plot of H3K27me3 promoters, after subtracting surrogate variables
  2118. \end_layout
  2119. \end_inset
  2120. \end_layout
  2121. \end_inset
  2122. \begin_inset space \hfill{}
  2123. \end_inset
  2124. \begin_inset Float figure
  2125. wide false
  2126. sideways false
  2127. status collapsed
  2128. \begin_layout Plain Layout
  2129. \align center
  2130. \begin_inset Graphics
  2131. filename graphics/CD4-csaw/RNA-seq/PCA-final-23-CROP.png
  2132. lyxscale 25
  2133. width 45col%
  2134. groupId pcoa-prom-subfig
  2135. \end_inset
  2136. \end_layout
  2137. \begin_layout Plain Layout
  2138. \begin_inset Caption Standard
  2139. \begin_layout Plain Layout
  2140. \series bold
  2141. \begin_inset CommandInset label
  2142. LatexCommand label
  2143. name "fig:RNA-PCA-group"
  2144. \end_inset
  2145. RNA-seq PCoA showing principal coordiantes 2 and 3.
  2146. \end_layout
  2147. \end_inset
  2148. \end_layout
  2149. \end_inset
  2150. \end_layout
  2151. \begin_layout Plain Layout
  2152. \begin_inset Caption Standard
  2153. \begin_layout Plain Layout
  2154. \series bold
  2155. \begin_inset CommandInset label
  2156. LatexCommand label
  2157. name "fig:PCoA-promoters"
  2158. \end_inset
  2159. PCoA plots for promoter ChIP-seq and expression RNA-seq data
  2160. \end_layout
  2161. \end_inset
  2162. \end_layout
  2163. \end_inset
  2164. \end_layout
  2165. \begin_layout Standard
  2166. \begin_inset Float table
  2167. wide false
  2168. sideways true
  2169. status collapsed
  2170. \begin_layout Plain Layout
  2171. \align center
  2172. \begin_inset Tabular
  2173. <lyxtabular version="3" rows="6" columns="7">
  2174. <features tabularvalignment="middle">
  2175. <column alignment="center" valignment="top">
  2176. <column alignment="center" valignment="top">
  2177. <column alignment="center" valignment="top">
  2178. <column alignment="center" valignment="top">
  2179. <column alignment="center" valignment="top">
  2180. <column alignment="center" valignment="top">
  2181. <column alignment="center" valignment="top">
  2182. <row>
  2183. <cell alignment="center" valignment="top" usebox="none">
  2184. \begin_inset Text
  2185. \begin_layout Plain Layout
  2186. \end_layout
  2187. \end_inset
  2188. </cell>
  2189. <cell multicolumn="1" alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  2190. \begin_inset Text
  2191. \begin_layout Plain Layout
  2192. Number of significant promoters
  2193. \end_layout
  2194. \end_inset
  2195. </cell>
  2196. <cell multicolumn="2" alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  2197. \begin_inset Text
  2198. \begin_layout Plain Layout
  2199. \end_layout
  2200. \end_inset
  2201. </cell>
  2202. <cell multicolumn="2" alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  2203. \begin_inset Text
  2204. \begin_layout Plain Layout
  2205. \end_layout
  2206. \end_inset
  2207. </cell>
  2208. <cell multicolumn="1" alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  2209. \begin_inset Text
  2210. \begin_layout Plain Layout
  2211. Est.
  2212. differentially modified promoters
  2213. \end_layout
  2214. \end_inset
  2215. </cell>
  2216. <cell multicolumn="2" alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  2217. \begin_inset Text
  2218. \begin_layout Plain Layout
  2219. \end_layout
  2220. \end_inset
  2221. </cell>
  2222. <cell multicolumn="2" alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  2223. \begin_inset Text
  2224. \begin_layout Plain Layout
  2225. \end_layout
  2226. \end_inset
  2227. </cell>
  2228. </row>
  2229. <row>
  2230. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  2231. \begin_inset Text
  2232. \begin_layout Plain Layout
  2233. Time Point
  2234. \end_layout
  2235. \end_inset
  2236. </cell>
  2237. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  2238. \begin_inset Text
  2239. \begin_layout Plain Layout
  2240. H3K4me2
  2241. \end_layout
  2242. \end_inset
  2243. </cell>
  2244. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  2245. \begin_inset Text
  2246. \begin_layout Plain Layout
  2247. H3K4me3
  2248. \end_layout
  2249. \end_inset
  2250. </cell>
  2251. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  2252. \begin_inset Text
  2253. \begin_layout Plain Layout
  2254. H3K27me3
  2255. \end_layout
  2256. \end_inset
  2257. </cell>
  2258. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  2259. \begin_inset Text
  2260. \begin_layout Plain Layout
  2261. H3K4me2
  2262. \end_layout
  2263. \end_inset
  2264. </cell>
  2265. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  2266. \begin_inset Text
  2267. \begin_layout Plain Layout
  2268. H3K4me3
  2269. \end_layout
  2270. \end_inset
  2271. </cell>
  2272. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  2273. \begin_inset Text
  2274. \begin_layout Plain Layout
  2275. H3K27me3
  2276. \end_layout
  2277. \end_inset
  2278. </cell>
  2279. </row>
  2280. <row>
  2281. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  2282. \begin_inset Text
  2283. \begin_layout Plain Layout
  2284. Day 0
  2285. \end_layout
  2286. \end_inset
  2287. </cell>
  2288. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  2289. \begin_inset Text
  2290. \begin_layout Plain Layout
  2291. 4553
  2292. \end_layout
  2293. \end_inset
  2294. </cell>
  2295. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  2296. \begin_inset Text
  2297. \begin_layout Plain Layout
  2298. 927
  2299. \end_layout
  2300. \end_inset
  2301. </cell>
  2302. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  2303. \begin_inset Text
  2304. \begin_layout Plain Layout
  2305. 6
  2306. \end_layout
  2307. \end_inset
  2308. </cell>
  2309. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  2310. \begin_inset Text
  2311. \begin_layout Plain Layout
  2312. 9967
  2313. \end_layout
  2314. \end_inset
  2315. </cell>
  2316. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  2317. \begin_inset Text
  2318. \begin_layout Plain Layout
  2319. 4149
  2320. \end_layout
  2321. \end_inset
  2322. </cell>
  2323. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  2324. \begin_inset Text
  2325. \begin_layout Plain Layout
  2326. 2404
  2327. \end_layout
  2328. \end_inset
  2329. </cell>
  2330. </row>
  2331. <row>
  2332. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  2333. \begin_inset Text
  2334. \begin_layout Plain Layout
  2335. Day 1
  2336. \end_layout
  2337. \end_inset
  2338. </cell>
  2339. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  2340. \begin_inset Text
  2341. \begin_layout Plain Layout
  2342. 567
  2343. \end_layout
  2344. \end_inset
  2345. </cell>
  2346. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  2347. \begin_inset Text
  2348. \begin_layout Plain Layout
  2349. 278
  2350. \end_layout
  2351. \end_inset
  2352. </cell>
  2353. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  2354. \begin_inset Text
  2355. \begin_layout Plain Layout
  2356. 1570
  2357. \end_layout
  2358. \end_inset
  2359. </cell>
  2360. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  2361. \begin_inset Text
  2362. \begin_layout Plain Layout
  2363. 4370
  2364. \end_layout
  2365. \end_inset
  2366. </cell>
  2367. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  2368. \begin_inset Text
  2369. \begin_layout Plain Layout
  2370. 2145
  2371. \end_layout
  2372. \end_inset
  2373. </cell>
  2374. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  2375. \begin_inset Text
  2376. \begin_layout Plain Layout
  2377. 6598
  2378. \end_layout
  2379. \end_inset
  2380. </cell>
  2381. </row>
  2382. <row>
  2383. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  2384. \begin_inset Text
  2385. \begin_layout Plain Layout
  2386. Day 5
  2387. \end_layout
  2388. \end_inset
  2389. </cell>
  2390. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  2391. \begin_inset Text
  2392. \begin_layout Plain Layout
  2393. 2313
  2394. \end_layout
  2395. \end_inset
  2396. </cell>
  2397. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  2398. \begin_inset Text
  2399. \begin_layout Plain Layout
  2400. 139
  2401. \end_layout
  2402. \end_inset
  2403. </cell>
  2404. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  2405. \begin_inset Text
  2406. \begin_layout Plain Layout
  2407. 490
  2408. \end_layout
  2409. \end_inset
  2410. </cell>
  2411. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  2412. \begin_inset Text
  2413. \begin_layout Plain Layout
  2414. 9450
  2415. \end_layout
  2416. \end_inset
  2417. </cell>
  2418. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  2419. \begin_inset Text
  2420. \begin_layout Plain Layout
  2421. 1148
  2422. \end_layout
  2423. \end_inset
  2424. </cell>
  2425. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  2426. \begin_inset Text
  2427. \begin_layout Plain Layout
  2428. 4141
  2429. \end_layout
  2430. \end_inset
  2431. </cell>
  2432. </row>
  2433. <row>
  2434. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  2435. \begin_inset Text
  2436. \begin_layout Plain Layout
  2437. Day 14
  2438. \end_layout
  2439. \end_inset
  2440. </cell>
  2441. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  2442. \begin_inset Text
  2443. \begin_layout Plain Layout
  2444. 0
  2445. \end_layout
  2446. \end_inset
  2447. </cell>
  2448. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  2449. \begin_inset Text
  2450. \begin_layout Plain Layout
  2451. 0
  2452. \end_layout
  2453. \end_inset
  2454. </cell>
  2455. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  2456. \begin_inset Text
  2457. \begin_layout Plain Layout
  2458. 0
  2459. \end_layout
  2460. \end_inset
  2461. </cell>
  2462. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  2463. \begin_inset Text
  2464. \begin_layout Plain Layout
  2465. 0
  2466. \end_layout
  2467. \end_inset
  2468. </cell>
  2469. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  2470. \begin_inset Text
  2471. \begin_layout Plain Layout
  2472. 0
  2473. \end_layout
  2474. \end_inset
  2475. </cell>
  2476. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  2477. \begin_inset Text
  2478. \begin_layout Plain Layout
  2479. 0
  2480. \end_layout
  2481. \end_inset
  2482. </cell>
  2483. </row>
  2484. </lyxtabular>
  2485. \end_inset
  2486. \end_layout
  2487. \begin_layout Plain Layout
  2488. \begin_inset Caption Standard
  2489. \begin_layout Plain Layout
  2490. \series bold
  2491. \begin_inset CommandInset label
  2492. LatexCommand label
  2493. name "tab:Number-signif-promoters"
  2494. \end_inset
  2495. Number of differentially modified promoters between naive and memory cells
  2496. at each time point after activation.
  2497. \series default
  2498. This table shows both the number of differentially modified promoters detected
  2499. at a 10% FDR threshold (left half), and the total number of differentially
  2500. modified promoters as estimated using the method of
  2501. \begin_inset CommandInset citation
  2502. LatexCommand cite
  2503. key "Phipson2013"
  2504. literal "false"
  2505. \end_inset
  2506. (right half).
  2507. \end_layout
  2508. \end_inset
  2509. \end_layout
  2510. \end_inset
  2511. \end_layout
  2512. \begin_layout Standard
  2513. \begin_inset Flex TODO Note (inline)
  2514. status open
  2515. \begin_layout Plain Layout
  2516. Check up on figure refs in this paragraph
  2517. \end_layout
  2518. \end_inset
  2519. \end_layout
  2520. \begin_layout Standard
  2521. Figure
  2522. \begin_inset CommandInset ref
  2523. LatexCommand ref
  2524. reference "fig:PCoA-promoters"
  2525. plural "false"
  2526. caps "false"
  2527. noprefix "false"
  2528. \end_inset
  2529. shows the patterns of variation in all 3 histone marks in the promoter
  2530. regions of the genome using principal coordinate analysis.
  2531. All 3 marks show a noticeable convergence between the naive and memory
  2532. samples at day 14, visible as an overlapping of the day 14 groups on each
  2533. plot.
  2534. This is consistent with the counts of significantly differentially modified
  2535. promoters and estimates of the total numbers of differentially modified
  2536. promoters shown in Table
  2537. \begin_inset CommandInset ref
  2538. LatexCommand ref
  2539. reference "tab:Number-signif-promoters"
  2540. plural "false"
  2541. caps "false"
  2542. noprefix "false"
  2543. \end_inset
  2544. .
  2545. For all histone marks, evidence of differential modification between naive
  2546. and memory samples was detected at every time point except day 14.
  2547. The day 14 convergence pattern is also present in the RNA-seq data (Figure
  2548. \begin_inset CommandInset ref
  2549. LatexCommand ref
  2550. reference "fig:RNA-PCA-group"
  2551. plural "false"
  2552. caps "false"
  2553. noprefix "false"
  2554. \end_inset
  2555. ), albiet in the 2nd and 3rd principal coordinates, indicating that it is
  2556. not the most dominant pattern driving gene expression.
  2557. Taken together, the data show that promoter histone methylation for these
  2558. 3 histone marks and RNA expression for naive and memory cells are most
  2559. similar at day 14, the furthest time point after activation.
  2560. MOFA was also able to capture this day 14 convergence pattern in latent
  2561. factor 5 (Figure
  2562. \begin_inset CommandInset ref
  2563. LatexCommand ref
  2564. reference "fig:mofa-lf-scatter"
  2565. plural "false"
  2566. caps "false"
  2567. noprefix "false"
  2568. \end_inset
  2569. ), which accounts for shared variation across all 3 histone marks and the
  2570. RNA-seq data, confirming that this is a coordinated pattern across all
  2571. 4 data sets.
  2572. \end_layout
  2573. \begin_layout Standard
  2574. \begin_inset Flex TODO Note (inline)
  2575. status collapsed
  2576. \begin_layout Plain Layout
  2577. This result feels shallow, somehow.
  2578. Am I oversimplifying the observation, or trivializing the amount of work
  2579. it took to get here? Shouldn't this section be more than one paragraph?
  2580. Am I just forgetting some supporting evidence that should also go here
  2581. in order to build up to the result? Or is it good that I have a simple
  2582. relatively straightforward result that doesn't take to long to explain,
  2583. and I'm just overthinking it?
  2584. \end_layout
  2585. \end_inset
  2586. \end_layout
  2587. \begin_layout Subsection
  2588. Effect of promoter coverage upstream vs downstream of TSS
  2589. \end_layout
  2590. \begin_layout Standard
  2591. \begin_inset Flex TODO Note (inline)
  2592. status open
  2593. \begin_layout Plain Layout
  2594. For the figures in this section, the group labels are arbitrary, so if time
  2595. allows, it would be good to manually reorder them in a logical way, e.g.
  2596. most upstream to most downstream.
  2597. \end_layout
  2598. \end_inset
  2599. \end_layout
  2600. \begin_layout Standard
  2601. \begin_inset Float figure
  2602. wide false
  2603. sideways true
  2604. status collapsed
  2605. \begin_layout Plain Layout
  2606. \align center
  2607. \begin_inset Float figure
  2608. wide false
  2609. sideways false
  2610. status collapsed
  2611. \begin_layout Plain Layout
  2612. \align center
  2613. \begin_inset Graphics
  2614. filename graphics/CD4-csaw/ChIP-seq/H3K4me2-neighborhood-clusters-CROP.png
  2615. lyxscale 25
  2616. width 30col%
  2617. groupId covprof-subfig
  2618. \end_inset
  2619. \end_layout
  2620. \begin_layout Plain Layout
  2621. \begin_inset Caption Standard
  2622. \begin_layout Plain Layout
  2623. \series bold
  2624. \begin_inset CommandInset label
  2625. LatexCommand label
  2626. name "fig:H3K4me2-neighborhood-clusters"
  2627. \end_inset
  2628. Average relative coverage for each bin in each cluster
  2629. \end_layout
  2630. \end_inset
  2631. \end_layout
  2632. \end_inset
  2633. \begin_inset space \hfill{}
  2634. \end_inset
  2635. \begin_inset Float figure
  2636. wide false
  2637. sideways false
  2638. status collapsed
  2639. \begin_layout Plain Layout
  2640. \align center
  2641. \begin_inset Graphics
  2642. filename graphics/CD4-csaw/ChIP-seq/H3K4me2-neighborhood-PCA-CROP.png
  2643. lyxscale 25
  2644. width 30col%
  2645. groupId covprof-subfig
  2646. \end_inset
  2647. \end_layout
  2648. \begin_layout Plain Layout
  2649. \begin_inset Caption Standard
  2650. \begin_layout Plain Layout
  2651. \series bold
  2652. \begin_inset CommandInset label
  2653. LatexCommand label
  2654. name "fig:H3K4me2-neighborhood-pca"
  2655. \end_inset
  2656. PCA of relative coverage depth, colored by K-means cluster membership.
  2657. \end_layout
  2658. \end_inset
  2659. \end_layout
  2660. \end_inset
  2661. \begin_inset space \hfill{}
  2662. \end_inset
  2663. \begin_inset Float figure
  2664. wide false
  2665. sideways false
  2666. status collapsed
  2667. \begin_layout Plain Layout
  2668. \align center
  2669. \begin_inset Graphics
  2670. filename graphics/CD4-csaw/ChIP-seq/H3K4me2-neighborhood-expression-CROP.png
  2671. lyxscale 25
  2672. width 30col%
  2673. groupId covprof-subfig
  2674. \end_inset
  2675. \end_layout
  2676. \begin_layout Plain Layout
  2677. \begin_inset Caption Standard
  2678. \begin_layout Plain Layout
  2679. \series bold
  2680. \begin_inset CommandInset label
  2681. LatexCommand label
  2682. name "fig:H3K4me2-neighborhood-expression"
  2683. \end_inset
  2684. Gene expression grouped by promoter coverage clusters.
  2685. \end_layout
  2686. \end_inset
  2687. \end_layout
  2688. \end_inset
  2689. \end_layout
  2690. \begin_layout Plain Layout
  2691. \begin_inset Caption Standard
  2692. \begin_layout Plain Layout
  2693. \series bold
  2694. K-means clustering of promoter H3K4me2 relative coverage depth in naive
  2695. day 0 samples.
  2696. \series default
  2697. H3K4me2 ChIP-seq reads were binned into 500-bp windows tiled across each
  2698. promoter from 5
  2699. \begin_inset space ~
  2700. \end_inset
  2701. kbp upstream to 5
  2702. \begin_inset space ~
  2703. \end_inset
  2704. kbp downstream, and the logCPM values were normalized within each promoter
  2705. to an average of 0, yielding relative coverage depths.
  2706. These were then grouped using K-means clustering with
  2707. \begin_inset Formula $K=6$
  2708. \end_inset
  2709. ,
  2710. \series bold
  2711. \series default
  2712. and the average bin values were plotted for each cluster (a).
  2713. The
  2714. \begin_inset Formula $x$
  2715. \end_inset
  2716. -axis is the genomic coordinate of each bin relative to the the transcription
  2717. start site, and the
  2718. \begin_inset Formula $y$
  2719. \end_inset
  2720. -axis is the mean relative coverage depth of that bin across all promoters
  2721. in the cluster.
  2722. Each line represents the average
  2723. \begin_inset Quotes eld
  2724. \end_inset
  2725. shape
  2726. \begin_inset Quotes erd
  2727. \end_inset
  2728. of the promoter coverage for promoters in that cluster.
  2729. PCA was performed on the same data, and the first two principal components
  2730. were plotted, coloring each point by its K-means cluster identity (b).
  2731. For each cluster, the distribution of gene expression values was plotted
  2732. (c).
  2733. \end_layout
  2734. \end_inset
  2735. \end_layout
  2736. \end_inset
  2737. \end_layout
  2738. \begin_layout Standard
  2739. \begin_inset Float figure
  2740. wide false
  2741. sideways true
  2742. status collapsed
  2743. \begin_layout Plain Layout
  2744. \align center
  2745. \begin_inset Float figure
  2746. wide false
  2747. sideways false
  2748. status collapsed
  2749. \begin_layout Plain Layout
  2750. \align center
  2751. \begin_inset Graphics
  2752. filename graphics/CD4-csaw/ChIP-seq/H3K27me3-neighborhood-clusters-CROP.png
  2753. lyxscale 25
  2754. width 30col%
  2755. groupId covprof-subfig
  2756. \end_inset
  2757. \end_layout
  2758. \begin_layout Plain Layout
  2759. \begin_inset Caption Standard
  2760. \begin_layout Plain Layout
  2761. \series bold
  2762. \begin_inset CommandInset label
  2763. LatexCommand label
  2764. name "fig:H3K27me3-neighborhood-clusters"
  2765. \end_inset
  2766. Average relative coverage for each bin in each cluster
  2767. \end_layout
  2768. \end_inset
  2769. \end_layout
  2770. \end_inset
  2771. \begin_inset space \hfill{}
  2772. \end_inset
  2773. \begin_inset Float figure
  2774. wide false
  2775. sideways false
  2776. status collapsed
  2777. \begin_layout Plain Layout
  2778. \align center
  2779. \begin_inset Graphics
  2780. filename graphics/CD4-csaw/ChIP-seq/H3K27me3-neighborhood-PCA-CROP.png
  2781. lyxscale 25
  2782. width 30col%
  2783. groupId covprof-subfig
  2784. \end_inset
  2785. \end_layout
  2786. \begin_layout Plain Layout
  2787. \begin_inset Caption Standard
  2788. \begin_layout Plain Layout
  2789. \series bold
  2790. \begin_inset CommandInset label
  2791. LatexCommand label
  2792. name "fig:H3K27me3-neighborhood-pca"
  2793. \end_inset
  2794. PCA of relative coverage depth, colored by K-means cluster membership.
  2795. \end_layout
  2796. \end_inset
  2797. \end_layout
  2798. \end_inset
  2799. \begin_inset space \hfill{}
  2800. \end_inset
  2801. \begin_inset Float figure
  2802. wide false
  2803. sideways false
  2804. status collapsed
  2805. \begin_layout Plain Layout
  2806. \align center
  2807. \begin_inset Graphics
  2808. filename graphics/CD4-csaw/ChIP-seq/H3K27me3-neighborhood-expression-CROP.png
  2809. lyxscale 25
  2810. width 30col%
  2811. groupId covprof-subfig
  2812. \end_inset
  2813. \end_layout
  2814. \begin_layout Plain Layout
  2815. \begin_inset Caption Standard
  2816. \begin_layout Plain Layout
  2817. \series bold
  2818. \begin_inset CommandInset label
  2819. LatexCommand label
  2820. name "fig:H3K27me3-neighborhood-expression"
  2821. \end_inset
  2822. Gene expression grouped by promoter coverage clusters.
  2823. \end_layout
  2824. \end_inset
  2825. \end_layout
  2826. \end_inset
  2827. \end_layout
  2828. \begin_layout Plain Layout
  2829. \begin_inset Caption Standard
  2830. \begin_layout Plain Layout
  2831. \series bold
  2832. K-means clustering of promoter H3K27me3 relative coverage depth in naive
  2833. day 0 samples.
  2834. \series default
  2835. H3K27me3 ChIP-seq reads were binned into 500-bp windows tiled across each
  2836. promoter from 5
  2837. \begin_inset space ~
  2838. \end_inset
  2839. kbp upstream to 5
  2840. \begin_inset space ~
  2841. \end_inset
  2842. kbp downstream, and the logCPM values were normalized within each promoter
  2843. to an average of 0, yielding relative coverage depths.
  2844. These were then grouped using K-means clustering with
  2845. \begin_inset Formula $K=6$
  2846. \end_inset
  2847. ,
  2848. \series bold
  2849. \series default
  2850. and the average bin values were plotted for each cluster (a).
  2851. The
  2852. \begin_inset Formula $x$
  2853. \end_inset
  2854. -axis is the genomic coordinate of each bin relative to the the transcription
  2855. start site, and the
  2856. \begin_inset Formula $y$
  2857. \end_inset
  2858. -axis is the mean relative coverage depth of that bin across all promoters
  2859. in the cluster.
  2860. Each line represents the average
  2861. \begin_inset Quotes eld
  2862. \end_inset
  2863. shape
  2864. \begin_inset Quotes erd
  2865. \end_inset
  2866. of the promoter coverage for promoters in that cluster.
  2867. PCA was performed on the same data, and the first two principal components
  2868. were plotted, coloring each point by its K-means cluster identity (b).
  2869. For each cluster, the distribution of gene expression values was plotted
  2870. (c).
  2871. \end_layout
  2872. \end_inset
  2873. \end_layout
  2874. \end_inset
  2875. \end_layout
  2876. \begin_layout Itemize
  2877. H3K4me peaks seem to correlate with increased expression as long as they
  2878. are anywhere near the TSS
  2879. \end_layout
  2880. \begin_layout Itemize
  2881. H3K27me3 peaks can have different correlations to gene expression depending
  2882. on their position relative to TSS (e.g.
  2883. upstream vs downstream) Results consistent with
  2884. \begin_inset CommandInset citation
  2885. LatexCommand cite
  2886. key "Young2011"
  2887. literal "false"
  2888. \end_inset
  2889. \end_layout
  2890. \begin_layout Standard
  2891. \begin_inset Flex TODO Note (inline)
  2892. status open
  2893. \begin_layout Plain Layout
  2894. Show the figures where the negative result ended this line of inquiry
  2895. \end_layout
  2896. \end_inset
  2897. \end_layout
  2898. \begin_layout Section
  2899. Discussion
  2900. \end_layout
  2901. \begin_layout Standard
  2902. \begin_inset Flex TODO Note (inline)
  2903. status open
  2904. \begin_layout Plain Layout
  2905. Try to boil it down to 3 main messages to get across
  2906. \end_layout
  2907. \end_inset
  2908. \end_layout
  2909. \begin_layout Itemize
  2910. 3 Main points
  2911. \end_layout
  2912. \begin_deeper
  2913. \begin_layout Itemize
  2914. Naive-to-memory convergence in certain data sets but not others, implies
  2915. which marks are involved in memory differentiation
  2916. \end_layout
  2917. \end_deeper
  2918. \begin_layout Subsection
  2919. Effective promoter radius
  2920. \end_layout
  2921. \begin_layout Itemize
  2922. "Promoter radius" is not constant and must be defined empirically for a
  2923. given data set.
  2924. Coverage within promoter radius has an expression correlation as well
  2925. \end_layout
  2926. \begin_layout Itemize
  2927. Further study required to demonstarte functional consequences of effective
  2928. promoter radius (e.g.
  2929. show diminished association with gene expression outside radius)
  2930. \end_layout
  2931. \begin_layout Subsection
  2932. Convergence
  2933. \end_layout
  2934. \begin_layout Standard
  2935. \begin_inset Float figure
  2936. wide false
  2937. sideways false
  2938. status collapsed
  2939. \begin_layout Plain Layout
  2940. \align center
  2941. \begin_inset Graphics
  2942. filename graphics/CD4-csaw/LaMere2016_fig8.pdf
  2943. lyxscale 50
  2944. width 100col%
  2945. groupId colwidth
  2946. \end_inset
  2947. \end_layout
  2948. \begin_layout Plain Layout
  2949. \begin_inset Caption Standard
  2950. \begin_layout Plain Layout
  2951. \series bold
  2952. LaMere 2016 Figure 8, reproduced with permission.
  2953. \end_layout
  2954. \end_inset
  2955. \end_layout
  2956. \end_inset
  2957. \end_layout
  2958. \begin_layout Itemize
  2959. Naive-to-memory convergence implies that naive cells are differentiating
  2960. into memory cells, and that gene expression and H3K4 methylation are involved
  2961. in this differentiation while H3K27me3 is less involved
  2962. \end_layout
  2963. \begin_deeper
  2964. \begin_layout Itemize
  2965. Convergence is consistent with Lamere2016 fig 8
  2966. \begin_inset CommandInset citation
  2967. LatexCommand cite
  2968. key "LaMere2016"
  2969. literal "false"
  2970. \end_inset
  2971. (which was created without the benefit of SVA)
  2972. \end_layout
  2973. \begin_layout Itemize
  2974. H3K27me3, canonically regarded as a deactivating mark, seems to have a more
  2975. complex effect
  2976. \end_layout
  2977. \end_deeper
  2978. \begin_layout Subsection
  2979. Positional
  2980. \end_layout
  2981. \begin_layout Itemize
  2982. TSS positional coverage, hints of something interesting but no clear conclusions
  2983. \end_layout
  2984. \begin_layout Subsection
  2985. Workflow
  2986. \end_layout
  2987. \begin_layout Standard
  2988. \begin_inset Float figure
  2989. wide false
  2990. sideways true
  2991. status open
  2992. \begin_layout Plain Layout
  2993. \align center
  2994. \begin_inset Graphics
  2995. filename graphics/CD4-csaw/rulegraphs/rulegraph-all.pdf
  2996. lyxscale 50
  2997. width 100theight%
  2998. \end_inset
  2999. \end_layout
  3000. \begin_layout Plain Layout
  3001. \begin_inset Caption Standard
  3002. \begin_layout Plain Layout
  3003. \begin_inset CommandInset label
  3004. LatexCommand label
  3005. name "fig:rulegraph"
  3006. \end_inset
  3007. \series bold
  3008. Dependency graph of steps in reproducible workflow
  3009. \end_layout
  3010. \end_inset
  3011. \end_layout
  3012. \end_inset
  3013. \end_layout
  3014. \begin_layout Itemize
  3015. Discuss advantages of developing using a reproducible workflow
  3016. \end_layout
  3017. \begin_deeper
  3018. \begin_layout Itemize
  3019. Decision-making based on trying every option and running the workflow downstream
  3020. to see the effects
  3021. \end_layout
  3022. \end_deeper
  3023. \begin_layout Subsection
  3024. Data quality issues limit conclusions
  3025. \end_layout
  3026. \begin_layout Chapter
  3027. Improving array-based analyses of transplant rejection by optimizing data
  3028. preprocessing
  3029. \end_layout
  3030. \begin_layout Standard
  3031. \begin_inset Note Note
  3032. status open
  3033. \begin_layout Plain Layout
  3034. Chapter author list: Me, Sunil, Tom, Padma, Dan
  3035. \end_layout
  3036. \end_inset
  3037. \end_layout
  3038. \begin_layout Section
  3039. Approach
  3040. \end_layout
  3041. \begin_layout Subsection
  3042. Proper pre-processing is essential for array data
  3043. \end_layout
  3044. \begin_layout Standard
  3045. \begin_inset Flex TODO Note (inline)
  3046. status open
  3047. \begin_layout Plain Layout
  3048. This section could probably use some citations
  3049. \end_layout
  3050. \end_inset
  3051. \end_layout
  3052. \begin_layout Standard
  3053. Microarrays, bead arrays, and similar assays produce raw data in the form
  3054. of fluorescence intensity measurements, with the each intensity measurement
  3055. proportional to the abundance of some fluorescently-labelled target DNA
  3056. or RNA sequence that base pairs to a specific probe sequence.
  3057. However, these measurements for each probe are also affected my many technical
  3058. confounding factors, such as the concentration of target material, strength
  3059. of off-target binding, and the sensitivity of the imaging sensor.
  3060. Some array designs also use multiple probe sequences for each target.
  3061. Hence, extensive pre-processing of array data is necessary to normalize
  3062. out the effects of these technical factors and summarize the information
  3063. from multiple probes to arrive at a single usable estimate of abundance
  3064. or other relevant quantity, such as a ratio of two abundances, for each
  3065. target.
  3066. \end_layout
  3067. \begin_layout Standard
  3068. The choice of pre-processing algorithms used in the analysis of an array
  3069. data set can have a large effect on the results of that analysis.
  3070. However, despite their importance, these steps are often neglected or rushed
  3071. in order to get to the more scientifically interesting analysis steps involving
  3072. the actual biology of the system under study.
  3073. Hence, it is often possible to achieve substantial gains in statistical
  3074. power, model goodness-of-fit, or other relevant performance measures, by
  3075. checking the assumptions made by each preprocessing step and choosing specific
  3076. normalization methods tailored to the specific goals of the current analysis.
  3077. \end_layout
  3078. \begin_layout Subsection
  3079. Normalization for clinical microarray classifiers must be single-channel
  3080. \end_layout
  3081. \begin_layout Subsubsection
  3082. Standard normalization methods are unsuitable for clinical application
  3083. \end_layout
  3084. \begin_layout Standard
  3085. As the cost of performing microarray assays falls, there is increasing interest
  3086. in using genomic assays for diagnostic purposes, such as distinguishing
  3087. healthy transplants (TX) from transplants undergoing acute rejection (AR)
  3088. or acute dysfunction with no rejection (ADNR).
  3089. However, the the standard normalization algorithm used for microarray data,
  3090. Robust Multi-chip Average (RMA)
  3091. \begin_inset CommandInset citation
  3092. LatexCommand cite
  3093. key "Irizarry2003a"
  3094. literal "false"
  3095. \end_inset
  3096. , is not applicable in a clinical setting.
  3097. Two of the steps in RMA, quantile normalization and probe summarization
  3098. by median polish, depend on every array in the data set being normalized.
  3099. This means that adding or removing any arrays from a data set changes the
  3100. normalized values for all arrays, and data sets that have been normalized
  3101. separately cannot be compared to each other.
  3102. Hence, when using RMA, any arrays to be analyzed together must also be
  3103. normalized together, and the set of arrays included in the data set must
  3104. be held constant throughout an analysis.
  3105. \end_layout
  3106. \begin_layout Standard
  3107. These limitations present serious impediments to the use of arrays as a
  3108. diagnostic tool.
  3109. When training a classifier, the samples to be classified must not be involved
  3110. in any step of the training process, lest their inclusion bias the training
  3111. process.
  3112. Once a classifier is deployed in a clinical setting, the samples to be
  3113. classified will not even
  3114. \emph on
  3115. exist
  3116. \emph default
  3117. at the time of training, so including them would be impossible even if
  3118. it were statistically justifiable.
  3119. Therefore, any machine learning application for microarrays demands that
  3120. the normalized expression values computed for an array must depend only
  3121. on information contained within that array.
  3122. This would ensure that each array's normalization is independent of every
  3123. other array, and that arrays normalized separately can still be compared
  3124. to each other without bias.
  3125. Such a normalization is commonly referred to as
  3126. \begin_inset Quotes eld
  3127. \end_inset
  3128. single-channel normalization
  3129. \begin_inset Quotes erd
  3130. \end_inset
  3131. .
  3132. \end_layout
  3133. \begin_layout Subsubsection
  3134. Several strategies are available to meet clinical normalization requirements
  3135. \end_layout
  3136. \begin_layout Standard
  3137. Frozen RMA (fRMA) addresses these concerns by replacing the quantile normalizati
  3138. on and median polish with alternatives that do not introduce inter-array
  3139. dependence, allowing each array to be normalized independently of all others
  3140. \begin_inset CommandInset citation
  3141. LatexCommand cite
  3142. key "McCall2010"
  3143. literal "false"
  3144. \end_inset
  3145. .
  3146. Quantile normalization is performed against a pre-generated set of quantiles
  3147. learned from a collection of 850 publically available arrays sampled from
  3148. a wide variety of tissues in the Gene Expression Omnibus (GEO).
  3149. Each array's probe intensity distribution is normalized against these pre-gener
  3150. ated quantiles.
  3151. The median polish step is replaced with a robust weighted average of probe
  3152. intensities, using inverse variance weights learned from the same public
  3153. GEO data.
  3154. The result is a normalization that satisfies the requirements mentioned
  3155. above: each array is normalized independently of all others, and any two
  3156. normalized arrays can be compared directly to each other.
  3157. \end_layout
  3158. \begin_layout Standard
  3159. One important limitation of fRMA is that it requires a separate reference
  3160. data set from which to learn the parameters (reference quantiles and probe
  3161. weights) that will be used to normalize each array.
  3162. These parameters are specific to a given array platform, and pre-generated
  3163. parameters are only provided for the most common platforms, such as Affymetrix
  3164. hgu133plus2.
  3165. For a less common platform, such as hthgu133pluspm, is is necessary to
  3166. learn custom parameters from in-house data before fRMA can be used to normalize
  3167. samples on that platform
  3168. \begin_inset CommandInset citation
  3169. LatexCommand cite
  3170. key "McCall2011"
  3171. literal "false"
  3172. \end_inset
  3173. .
  3174. \end_layout
  3175. \begin_layout Standard
  3176. One other option is the aptly-named Single Channel Array Normalization (SCAN),
  3177. which adapts a normalization method originally designed for tiling arrays
  3178. \begin_inset CommandInset citation
  3179. LatexCommand cite
  3180. key "Piccolo2012"
  3181. literal "false"
  3182. \end_inset
  3183. .
  3184. SCAN is truly single-channel in that it does not require a set of normalization
  3185. paramters estimated from an external set of reference samples like fRMA
  3186. does.
  3187. \end_layout
  3188. \begin_layout Subsection
  3189. Heteroskedasticity must be accounted for in methylation array data
  3190. \end_layout
  3191. \begin_layout Subsubsection
  3192. Methylation array preprocessing induces heteroskedasticity
  3193. \end_layout
  3194. \begin_layout Standard
  3195. DNA methylation arrays are a relatively new kind of assay that uses microarrays
  3196. to measure the degree of methylation on cytosines in specific regions arrayed
  3197. across the genome.
  3198. First, bisulfite treatment converts all unmethylated cytosines to uracil
  3199. (which then become thymine after amplication) while leaving methylated
  3200. cytosines unaffected.
  3201. Then, each target region is interrogated with two probes: one binds to
  3202. the original genomic sequence and interrogates the level of methylated
  3203. DNA, and the other binds to the same sequence with all cytosines replaced
  3204. by thymidines and interrogates the level of unmethylated DNA.
  3205. \end_layout
  3206. \begin_layout Standard
  3207. \begin_inset Float figure
  3208. wide false
  3209. sideways false
  3210. status collapsed
  3211. \begin_layout Plain Layout
  3212. \align center
  3213. \begin_inset Graphics
  3214. filename graphics/methylvoom/sigmoid.pdf
  3215. \end_inset
  3216. \end_layout
  3217. \begin_layout Plain Layout
  3218. \begin_inset Caption Standard
  3219. \begin_layout Plain Layout
  3220. \begin_inset CommandInset label
  3221. LatexCommand label
  3222. name "fig:Sigmoid-beta-m-mapping"
  3223. \end_inset
  3224. \series bold
  3225. Sigmoid shape of the mapping between β and M values
  3226. \end_layout
  3227. \end_inset
  3228. \end_layout
  3229. \end_inset
  3230. \end_layout
  3231. \begin_layout Standard
  3232. After normalization, these two probe intensities are summarized in one of
  3233. two ways, each with advantages and disadvantages.
  3234. β
  3235. \series bold
  3236. \series default
  3237. values, interpreted as fraction of DNA copies methylated, range from 0 to
  3238. 1.
  3239. β
  3240. \series bold
  3241. \series default
  3242. values are conceptually easy to interpret, but the constrained range makes
  3243. them unsuitable for linear modeling, and their error distributions are
  3244. highly non-normal, which also frustrates linear modeling.
  3245. M-values, interpreted as the log ratio of methylated to unmethylated copies,
  3246. are computed by mapping the beta values from
  3247. \begin_inset Formula $[0,1]$
  3248. \end_inset
  3249. onto
  3250. \begin_inset Formula $(-\infty,+\infty)$
  3251. \end_inset
  3252. using a sigmoid curve (Figure
  3253. \begin_inset CommandInset ref
  3254. LatexCommand ref
  3255. reference "fig:Sigmoid-beta-m-mapping"
  3256. plural "false"
  3257. caps "false"
  3258. noprefix "false"
  3259. \end_inset
  3260. ).
  3261. This transformation results in values with better statistical perperties:
  3262. the unconstrained range is suitable for linear modeling, and the error
  3263. distributions are more normal.
  3264. Hence, most linear modeling and other statistical testing on methylation
  3265. arrays is performed using M-values.
  3266. \end_layout
  3267. \begin_layout Standard
  3268. However, the steep slope of the sigmoid transformation near 0 and 1 tends
  3269. to over-exaggerate small differences in β values near those extremes, which
  3270. in turn amplifies the error in those values, leading to a U-shaped trend
  3271. in the mean-variance curve: extreme values have higher variances than values
  3272. near the middle.
  3273. This mean-variance dependency must be accounted for when fitting the linear
  3274. model for differential methylation, or else the variance will be systematically
  3275. overestimated for probes with moderate M-values and underestimated for
  3276. probes with extreme M-values.
  3277. \end_layout
  3278. \begin_layout Subsubsection
  3279. The voom method for RNA-seq data can model M-value heteroskedasticity
  3280. \end_layout
  3281. \begin_layout Standard
  3282. RNA-seq read count data are also known to show heteroskedasticity, and the
  3283. voom method was developed for modeling this heteroskedasticity by estimating
  3284. the mean-variance trend in the data and using this trend to assign precision
  3285. weights to each observation
  3286. \begin_inset CommandInset citation
  3287. LatexCommand cite
  3288. key "Law2013"
  3289. literal "false"
  3290. \end_inset
  3291. .
  3292. While methylation array data are not derived from counts and have a very
  3293. different mean-variance relationship from that of typical RNA-seq data,
  3294. the voom method makes no specific assumptions on the shape of the mean-variance
  3295. relationship - it only assumes that the relationship is smooth enough to
  3296. model using a lowess curve.
  3297. Hence, the method is sufficiently general to model the mean-variance relationsh
  3298. ip in methylation array data.
  3299. However, the standard implementation of voom assumes that the input is
  3300. given in raw read counts, and it must be adapted to run on methylation
  3301. M-values.
  3302. \end_layout
  3303. \begin_layout Section
  3304. Methods
  3305. \end_layout
  3306. \begin_layout Subsection
  3307. Evaluation of classifier performance with different normalization methods
  3308. \end_layout
  3309. \begin_layout Standard
  3310. For testing different expression microarray normalizations, a data set of
  3311. 157 hgu133plus2 arrays was used, consisting of blood samples from kidney
  3312. transplant patients whose grafts had been graded as TX, AR, or ADNR via
  3313. biopsy and histology (46 TX, 69 AR, 42 ADNR)
  3314. \begin_inset CommandInset citation
  3315. LatexCommand cite
  3316. key "Kurian2014"
  3317. literal "true"
  3318. \end_inset
  3319. .
  3320. Additionally, an external validation set of 75 samples was gathered from
  3321. public GEO data (37 TX, 38 AR, no ADNR).
  3322. \end_layout
  3323. \begin_layout Standard
  3324. \begin_inset Flex TODO Note (inline)
  3325. status open
  3326. \begin_layout Plain Layout
  3327. Find appropriate GEO identifiers if possible.
  3328. Kurian 2014 says GSE15296, but this seems to be different data.
  3329. I also need to look up the GEO accession for the external validation set.
  3330. \end_layout
  3331. \end_inset
  3332. \end_layout
  3333. \begin_layout Standard
  3334. To evaluate the effect of each normalization on classifier performance,
  3335. the same classifier training and validation procedure was used after each
  3336. normalization method.
  3337. The PAM package was used to train a nearest shrunken centroid classifier
  3338. on the training set and select the appropriate threshold for centroid shrinking.
  3339. Then the trained classifier was used to predict the class probabilities
  3340. of each validation sample.
  3341. From these class probabilities, ROC curves and area-under-curve (AUC) values
  3342. were generated
  3343. \begin_inset CommandInset citation
  3344. LatexCommand cite
  3345. key "Turck2011"
  3346. literal "false"
  3347. \end_inset
  3348. .
  3349. Each normalization was tested on two different sets of training and validation
  3350. samples.
  3351. For internal validation, the 115 TX and AR arrays in the internal set were
  3352. split at random into two equal sized sets, one for training and one for
  3353. validation, each containing the same numbers of TX and AR samples as the
  3354. other set.
  3355. For external validation, the full set of 115 TX and AR samples were used
  3356. as a training set, and the 75 external TX and AR samples were used as the
  3357. validation set.
  3358. Thus, 2 ROC curves and AUC values were generated for each normalization
  3359. method: one internal and one external.
  3360. Because the external validation set contains no ADNR samples, only classificati
  3361. on of TX and AR samples was considered.
  3362. The ADNR samples were included during normalization but excluded from all
  3363. classifier training and validation.
  3364. This ensures that the performance on internal and external validation sets
  3365. is directly comparable, since both are performing the same task: distinguising
  3366. TX from AR.
  3367. \end_layout
  3368. \begin_layout Standard
  3369. \begin_inset Flex TODO Note (inline)
  3370. status open
  3371. \begin_layout Plain Layout
  3372. Summarize the get.best.threshold algorithm for PAM threshold selection, or
  3373. just put the code online?
  3374. \end_layout
  3375. \end_inset
  3376. \end_layout
  3377. \begin_layout Standard
  3378. Six different normalization strategies were evaluated.
  3379. First, 2 well-known non-single-channel normalization methods were considered:
  3380. RMA and dChip
  3381. \begin_inset CommandInset citation
  3382. LatexCommand cite
  3383. key "Li2001,Irizarry2003a"
  3384. literal "false"
  3385. \end_inset
  3386. .
  3387. Since RMA produces expression values on a log2 scale and dChip does not,
  3388. the values from dChip were log2 transformed after normalization.
  3389. Next, RMA and dChip followed by Global Rank-invariant Set Normalization
  3390. (GRSN) were tested
  3391. \begin_inset CommandInset citation
  3392. LatexCommand cite
  3393. key "Pelz2008"
  3394. literal "false"
  3395. \end_inset
  3396. .
  3397. Post-processing with GRSN does not turn RMA or dChip into single-channel
  3398. methods, but it may help mitigate batch effects and is therefore useful
  3399. as a benchmark.
  3400. Lastly, the two single-channel normalization methods, fRMA and SCAN, were
  3401. tested
  3402. \begin_inset CommandInset citation
  3403. LatexCommand cite
  3404. key "McCall2010,Piccolo2012"
  3405. literal "false"
  3406. \end_inset
  3407. .
  3408. When evaluting internal validation performance, only the 157 internal samples
  3409. were normalized; when evaluating external validation performance, all 157
  3410. internal samples and 75 external samples were normalized together.
  3411. \end_layout
  3412. \begin_layout Standard
  3413. For demonstrating the problem with separate normalization of training and
  3414. validation data, one additional normalization was performed: the internal
  3415. and external sets were each normalized separately using RMA, and the normalized
  3416. data for each set were combined into a single set with no further attempts
  3417. at normalizing between the two sets.
  3418. The represents approximately how RMA would have to be used in a clinical
  3419. setting, where the samples to be classified are not available at the time
  3420. the classifier is trained.
  3421. \end_layout
  3422. \begin_layout Subsection
  3423. Generating custom fRMA vectors for hthgu133pluspm array platform
  3424. \end_layout
  3425. \begin_layout Standard
  3426. In order to enable fRMA normalization for the hthgu133pluspm array platform,
  3427. custom fRMA normalization vectors were trained using the frmaTools package
  3428. \begin_inset CommandInset citation
  3429. LatexCommand cite
  3430. key "McCall2011"
  3431. literal "false"
  3432. \end_inset
  3433. .
  3434. Separate vectors were created for two types of samples: kidney graft biopsy
  3435. samples and blood samples from graft recipients.
  3436. For training, a 341 kidney biopsy samples from 2 data sets and 965 blood
  3437. samples from 5 data sets were used as the reference set.
  3438. Arrays were groups into batches based on unique combinations of sample
  3439. type (blood or biopsy), diagnosis (TX, AR, etc.), data set, and scan date.
  3440. Thus, each batch represents arrays of the same kind that were run together
  3441. on the same day.
  3442. For estimating the probe inverse variance weights, frmaTools requires equal-siz
  3443. ed batches, which means a batch size must be chosen, and then batches smaller
  3444. than that size must be ignored, while batches larger than the chosen size
  3445. must be downsampled.
  3446. This downsampling is performed randomly, so the sampling process is repeated
  3447. 5 times and the resulting normalizations are compared to each other.
  3448. \end_layout
  3449. \begin_layout Standard
  3450. To evaluate the consistency of the generated normalization vectors, the
  3451. 5 fRMA vector sets generated from 5 random batch samplings were each used
  3452. to normalize the same 20 randomly selected samples from each tissue.
  3453. Then the normalized expression values for each probe on each array were
  3454. compared across all normalizations.
  3455. Each fRMA normalization was also compared against the normalized expression
  3456. values obtained by normalizing the same 20 samples with ordinary RMA.
  3457. \end_layout
  3458. \begin_layout Subsection
  3459. Modeling methylation array M-value heteroskedasticy in linear models with
  3460. modified voom implementation
  3461. \end_layout
  3462. \begin_layout Standard
  3463. \begin_inset Flex TODO Note (inline)
  3464. status open
  3465. \begin_layout Plain Layout
  3466. Put code on Github and reference it.
  3467. \end_layout
  3468. \end_inset
  3469. \end_layout
  3470. \begin_layout Standard
  3471. To investigate the whether DNA methylation could be used to distinguish
  3472. between healthy and dysfunctional transplants, a data set of 78 Illumina
  3473. 450k methylation arrays from human kidney graft biopsies was analyzed for
  3474. differential metylation between 4 transplant statuses: healthy transplant
  3475. (TX), transplants undergoing acute rejection (AR), acute dysfunction with
  3476. no rejection (ADNR), and chronic allograpft nephropathy (CAN).
  3477. The data consisted of 33 TX, 9 AR, 8 ADNR, and 28 CAN samples.
  3478. The uneven group sizes are a result of taking the biopsy samples before
  3479. the eventual fate of the transplant was known.
  3480. Each sample was additionally annotated with a donor ID (anonymized), Sex,
  3481. Age, Ethnicity, Creatinine Level, and Diabetes diagnosois (all samples
  3482. in this data set came from patients with either Type 1 or Type 2 diabetes).
  3483. \end_layout
  3484. \begin_layout Standard
  3485. The intensity data were first normalized using subset-quantile within array
  3486. normalization (SWAN)
  3487. \begin_inset CommandInset citation
  3488. LatexCommand cite
  3489. key "Maksimovic2012"
  3490. literal "false"
  3491. \end_inset
  3492. , then converted to intensity ratios (beta values)
  3493. \begin_inset CommandInset citation
  3494. LatexCommand cite
  3495. key "Aryee2014"
  3496. literal "false"
  3497. \end_inset
  3498. .
  3499. Any probes binding to loci that overlapped annotated SNPs were dropped,
  3500. and the annotated sex of each sample was verified against the sex inferred
  3501. from the ratio of median probe intensities for the X and Y chromosomes.
  3502. Then, the ratios were transformed to M-values.
  3503. \end_layout
  3504. \begin_layout Standard
  3505. \begin_inset Float table
  3506. wide false
  3507. sideways false
  3508. status open
  3509. \begin_layout Plain Layout
  3510. \begin_inset Tabular
  3511. <lyxtabular version="3" rows="4" columns="6">
  3512. <features tabularvalignment="middle">
  3513. <column alignment="center" valignment="top">
  3514. <column alignment="center" valignment="top">
  3515. <column alignment="center" valignment="top">
  3516. <column alignment="center" valignment="top">
  3517. <column alignment="center" valignment="top">
  3518. <column alignment="center" valignment="top">
  3519. <row>
  3520. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  3521. \begin_inset Text
  3522. \begin_layout Plain Layout
  3523. Analysis
  3524. \end_layout
  3525. \end_inset
  3526. </cell>
  3527. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  3528. \begin_inset Text
  3529. \begin_layout Plain Layout
  3530. patient random effect
  3531. \end_layout
  3532. \end_inset
  3533. </cell>
  3534. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  3535. \begin_inset Text
  3536. \begin_layout Plain Layout
  3537. empirical Bayes
  3538. \end_layout
  3539. \end_inset
  3540. </cell>
  3541. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  3542. \begin_inset Text
  3543. \begin_layout Plain Layout
  3544. SVA
  3545. \end_layout
  3546. \end_inset
  3547. </cell>
  3548. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  3549. \begin_inset Text
  3550. \begin_layout Plain Layout
  3551. sample weights
  3552. \end_layout
  3553. \end_inset
  3554. </cell>
  3555. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  3556. \begin_inset Text
  3557. \begin_layout Plain Layout
  3558. voom
  3559. \end_layout
  3560. \end_inset
  3561. </cell>
  3562. </row>
  3563. <row>
  3564. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  3565. \begin_inset Text
  3566. \begin_layout Plain Layout
  3567. A
  3568. \end_layout
  3569. \end_inset
  3570. </cell>
  3571. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  3572. \begin_inset Text
  3573. \begin_layout Plain Layout
  3574. Yes
  3575. \end_layout
  3576. \end_inset
  3577. </cell>
  3578. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  3579. \begin_inset Text
  3580. \begin_layout Plain Layout
  3581. Yes
  3582. \end_layout
  3583. \end_inset
  3584. </cell>
  3585. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  3586. \begin_inset Text
  3587. \begin_layout Plain Layout
  3588. No
  3589. \end_layout
  3590. \end_inset
  3591. </cell>
  3592. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  3593. \begin_inset Text
  3594. \begin_layout Plain Layout
  3595. No
  3596. \end_layout
  3597. \end_inset
  3598. </cell>
  3599. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  3600. \begin_inset Text
  3601. \begin_layout Plain Layout
  3602. No
  3603. \end_layout
  3604. \end_inset
  3605. </cell>
  3606. </row>
  3607. <row>
  3608. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  3609. \begin_inset Text
  3610. \begin_layout Plain Layout
  3611. B
  3612. \end_layout
  3613. \end_inset
  3614. </cell>
  3615. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  3616. \begin_inset Text
  3617. \begin_layout Plain Layout
  3618. Yes
  3619. \end_layout
  3620. \end_inset
  3621. </cell>
  3622. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  3623. \begin_inset Text
  3624. \begin_layout Plain Layout
  3625. Yes
  3626. \end_layout
  3627. \end_inset
  3628. </cell>
  3629. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  3630. \begin_inset Text
  3631. \begin_layout Plain Layout
  3632. Yes
  3633. \end_layout
  3634. \end_inset
  3635. </cell>
  3636. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  3637. \begin_inset Text
  3638. \begin_layout Plain Layout
  3639. Yes
  3640. \end_layout
  3641. \end_inset
  3642. </cell>
  3643. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  3644. \begin_inset Text
  3645. \begin_layout Plain Layout
  3646. No
  3647. \end_layout
  3648. \end_inset
  3649. </cell>
  3650. </row>
  3651. <row>
  3652. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  3653. \begin_inset Text
  3654. \begin_layout Plain Layout
  3655. C
  3656. \end_layout
  3657. \end_inset
  3658. </cell>
  3659. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  3660. \begin_inset Text
  3661. \begin_layout Plain Layout
  3662. Yes
  3663. \end_layout
  3664. \end_inset
  3665. </cell>
  3666. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  3667. \begin_inset Text
  3668. \begin_layout Plain Layout
  3669. Yes
  3670. \end_layout
  3671. \end_inset
  3672. </cell>
  3673. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  3674. \begin_inset Text
  3675. \begin_layout Plain Layout
  3676. Yes
  3677. \end_layout
  3678. \end_inset
  3679. </cell>
  3680. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  3681. \begin_inset Text
  3682. \begin_layout Plain Layout
  3683. Yes
  3684. \end_layout
  3685. \end_inset
  3686. </cell>
  3687. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  3688. \begin_inset Text
  3689. \begin_layout Plain Layout
  3690. Yes
  3691. \end_layout
  3692. \end_inset
  3693. </cell>
  3694. </row>
  3695. </lyxtabular>
  3696. \end_inset
  3697. \end_layout
  3698. \begin_layout Plain Layout
  3699. \begin_inset Caption Standard
  3700. \begin_layout Plain Layout
  3701. \series bold
  3702. \begin_inset CommandInset label
  3703. LatexCommand label
  3704. name "tab:Summary-of-meth-analysis"
  3705. \end_inset
  3706. Summary of analysis variants for methylation array data.
  3707. \series default
  3708. Each analysis included a different set of steps to adjust or account for
  3709. various systematic features of the data.
  3710. See the text for a more detailed explanation of each step.
  3711. \end_layout
  3712. \end_inset
  3713. \end_layout
  3714. \end_inset
  3715. \end_layout
  3716. \begin_layout Standard
  3717. From the M-values, a series of parallel analyses was performed, each adding
  3718. additional steps into the model fit to accomodate a feature of the data
  3719. (see Table
  3720. \begin_inset CommandInset ref
  3721. LatexCommand ref
  3722. reference "tab:Summary-of-meth-analysis"
  3723. plural "false"
  3724. caps "false"
  3725. noprefix "false"
  3726. \end_inset
  3727. ).
  3728. For analysis A, a
  3729. \begin_inset Quotes eld
  3730. \end_inset
  3731. basic
  3732. \begin_inset Quotes erd
  3733. \end_inset
  3734. linear modeling analysis was performed, compensating for known confounders
  3735. by including terms for the factor of interest (transplant status) as well
  3736. as the known biological confounders: sex, age, ethnicity, and diabetes.
  3737. Since some samples came from the same patients at different times, the
  3738. intra-patient correlation was modeled as a random effect, estimating a
  3739. shared correlation value across all probes
  3740. \begin_inset CommandInset citation
  3741. LatexCommand cite
  3742. key "Smyth2005a"
  3743. literal "false"
  3744. \end_inset
  3745. .
  3746. Then the linear model was fit, and the variance was modeled using empirical
  3747. Bayes squeezing toward the mean-variance trend
  3748. \begin_inset CommandInset citation
  3749. LatexCommand cite
  3750. key "Ritchie2015"
  3751. literal "false"
  3752. \end_inset
  3753. .
  3754. Finally, t-tests or F-tests were performed as appropriate for each test:
  3755. t-tests for single contrasts, and F-tests for multiple contrasts.
  3756. P-values were corrected for multiple testing using the Benjamini-Hochberg
  3757. procedure for FDR control
  3758. \begin_inset CommandInset citation
  3759. LatexCommand cite
  3760. key "Benjamini1995"
  3761. literal "false"
  3762. \end_inset
  3763. .
  3764. \end_layout
  3765. \begin_layout Standard
  3766. For the analysis B, surrogate variable analysis (SVA) was used to infer
  3767. additional unobserved sources of heterogeneity in the data
  3768. \begin_inset CommandInset citation
  3769. LatexCommand cite
  3770. key "Leek2007"
  3771. literal "false"
  3772. \end_inset
  3773. .
  3774. These surrogate variables were added to the design matrix before fitting
  3775. the linear model.
  3776. In addition, sample quality weights were estimated from the data and used
  3777. during linear modeling to down-weight the contribution of highly variable
  3778. arrays while increasing the weight to arrays with lower variability
  3779. \begin_inset CommandInset citation
  3780. LatexCommand cite
  3781. key "Ritchie2006"
  3782. literal "false"
  3783. \end_inset
  3784. .
  3785. The remainder of the analysis proceeded as in analysis A.
  3786. For analysis C, the voom method was adapted to run on methylation array
  3787. data and used to model and correct for the mean-variance trend using individual
  3788. observation weights
  3789. \begin_inset CommandInset citation
  3790. LatexCommand cite
  3791. key "Law2013"
  3792. literal "false"
  3793. \end_inset
  3794. , which were combined with the sample weights
  3795. \begin_inset CommandInset citation
  3796. LatexCommand cite
  3797. key "Liu2015"
  3798. literal "false"
  3799. \end_inset
  3800. .
  3801. Each time weights were used, they were estimated once before estimating
  3802. the random effect correlation value, and then the weights were re-estimated
  3803. taking the random effect into account.
  3804. The remainder of the analysis proceeded as in analysis B.
  3805. \end_layout
  3806. \begin_layout Section
  3807. Results
  3808. \end_layout
  3809. \begin_layout Standard
  3810. \begin_inset Flex TODO Note (inline)
  3811. status open
  3812. \begin_layout Plain Layout
  3813. Improve subsection titles in this section
  3814. \end_layout
  3815. \end_inset
  3816. \end_layout
  3817. \begin_layout Subsection
  3818. fRMA eliminates unwanted dependence of classifier training on normalization
  3819. strategy caused by RMA
  3820. \end_layout
  3821. \begin_layout Standard
  3822. \begin_inset Flex TODO Note (inline)
  3823. status open
  3824. \begin_layout Plain Layout
  3825. Write figure legends
  3826. \end_layout
  3827. \end_inset
  3828. \end_layout
  3829. \begin_layout Subsubsection
  3830. Separate normalization with RMA introduces unwanted biases in classification
  3831. \end_layout
  3832. \begin_layout Standard
  3833. \begin_inset Float figure
  3834. wide false
  3835. sideways false
  3836. status collapsed
  3837. \begin_layout Plain Layout
  3838. \align center
  3839. \begin_inset Graphics
  3840. filename graphics/PAM/predplot.pdf
  3841. lyxscale 50
  3842. width 100col%
  3843. groupId colwidth
  3844. \end_inset
  3845. \end_layout
  3846. \begin_layout Plain Layout
  3847. \begin_inset Caption Standard
  3848. \begin_layout Plain Layout
  3849. \begin_inset CommandInset label
  3850. LatexCommand label
  3851. name "fig:Classifier-probabilities-RMA"
  3852. \end_inset
  3853. \series bold
  3854. Classifier probabilities on validation samples when normalized with RMA
  3855. together vs.
  3856. separately.
  3857. \end_layout
  3858. \end_inset
  3859. \end_layout
  3860. \end_inset
  3861. \end_layout
  3862. \begin_layout Standard
  3863. To demonstrate the problem with non-single-channel normalization methods,
  3864. we considered the problem of training a classifier to distinguish TX from
  3865. AR using the samples from the internal set as training data, evaluating
  3866. performance on the external set.
  3867. First, training and evaluation were performed after normalizing all array
  3868. samples together as a single set using RMA, and second, the internal samples
  3869. were normalized separately from the external samples and the training and
  3870. evaluation were repeated.
  3871. For each sample in the validation set, the classifier probabilities from
  3872. both classifiers were plotted against each other (Fig.
  3873. \begin_inset CommandInset ref
  3874. LatexCommand ref
  3875. reference "fig:Classifier-probabilities-RMA"
  3876. plural "false"
  3877. caps "false"
  3878. noprefix "false"
  3879. \end_inset
  3880. ).
  3881. As expected, separate normalization biases the classifier probabilities,
  3882. resulting in several misclassifications.
  3883. In this case, the bias from separate normalization causes the classifier
  3884. to assign a lower probability of AR to every sample.
  3885. \end_layout
  3886. \begin_layout Subsubsection
  3887. fRMA and SCAN achieve maintain classification performance while eliminating
  3888. dependence on normalization strategy
  3889. \end_layout
  3890. \begin_layout Standard
  3891. \begin_inset Float figure
  3892. placement tb
  3893. wide false
  3894. sideways false
  3895. status collapsed
  3896. \begin_layout Plain Layout
  3897. \align center
  3898. \begin_inset Graphics
  3899. filename graphics/PAM/ROC-TXvsAR-internal.pdf
  3900. lyxscale 50
  3901. width 100col%
  3902. groupId colwidth
  3903. \end_inset
  3904. \end_layout
  3905. \begin_layout Plain Layout
  3906. \begin_inset Caption Standard
  3907. \begin_layout Plain Layout
  3908. \series bold
  3909. \begin_inset CommandInset label
  3910. LatexCommand label
  3911. name "fig:ROC-PAM-int"
  3912. \end_inset
  3913. ROC curves for PAM on internal validation data using different normalization
  3914. strategies
  3915. \end_layout
  3916. \end_inset
  3917. \end_layout
  3918. \end_inset
  3919. \end_layout
  3920. \begin_layout Standard
  3921. \begin_inset Float table
  3922. wide false
  3923. sideways false
  3924. status collapsed
  3925. \begin_layout Plain Layout
  3926. \align center
  3927. \begin_inset Tabular
  3928. <lyxtabular version="3" rows="7" columns="4">
  3929. <features tabularvalignment="middle">
  3930. <column alignment="center" valignment="top">
  3931. <column alignment="center" valignment="top">
  3932. <column alignment="center" valignment="top">
  3933. <column alignment="center" valignment="top">
  3934. <row>
  3935. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  3936. \begin_inset Text
  3937. \begin_layout Plain Layout
  3938. \family roman
  3939. \series medium
  3940. \shape up
  3941. \size normal
  3942. \emph off
  3943. \bar no
  3944. \strikeout off
  3945. \xout off
  3946. \uuline off
  3947. \uwave off
  3948. \noun off
  3949. \color none
  3950. Normalization
  3951. \end_layout
  3952. \end_inset
  3953. </cell>
  3954. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  3955. \begin_inset Text
  3956. \begin_layout Plain Layout
  3957. Single-channel?
  3958. \end_layout
  3959. \end_inset
  3960. </cell>
  3961. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  3962. \begin_inset Text
  3963. \begin_layout Plain Layout
  3964. \family roman
  3965. \series medium
  3966. \shape up
  3967. \size normal
  3968. \emph off
  3969. \bar no
  3970. \strikeout off
  3971. \xout off
  3972. \uuline off
  3973. \uwave off
  3974. \noun off
  3975. \color none
  3976. Internal Val.
  3977. AUC
  3978. \end_layout
  3979. \end_inset
  3980. </cell>
  3981. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  3982. \begin_inset Text
  3983. \begin_layout Plain Layout
  3984. External Val.
  3985. AUC
  3986. \end_layout
  3987. \end_inset
  3988. </cell>
  3989. </row>
  3990. <row>
  3991. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  3992. \begin_inset Text
  3993. \begin_layout Plain Layout
  3994. \family roman
  3995. \series medium
  3996. \shape up
  3997. \size normal
  3998. \emph off
  3999. \bar no
  4000. \strikeout off
  4001. \xout off
  4002. \uuline off
  4003. \uwave off
  4004. \noun off
  4005. \color none
  4006. RMA
  4007. \end_layout
  4008. \end_inset
  4009. </cell>
  4010. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  4011. \begin_inset Text
  4012. \begin_layout Plain Layout
  4013. No
  4014. \end_layout
  4015. \end_inset
  4016. </cell>
  4017. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  4018. \begin_inset Text
  4019. \begin_layout Plain Layout
  4020. \family roman
  4021. \series medium
  4022. \shape up
  4023. \size normal
  4024. \emph off
  4025. \bar no
  4026. \strikeout off
  4027. \xout off
  4028. \uuline off
  4029. \uwave off
  4030. \noun off
  4031. \color none
  4032. 0.852
  4033. \end_layout
  4034. \end_inset
  4035. </cell>
  4036. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  4037. \begin_inset Text
  4038. \begin_layout Plain Layout
  4039. \family roman
  4040. \series medium
  4041. \shape up
  4042. \size normal
  4043. \emph off
  4044. \bar no
  4045. \strikeout off
  4046. \xout off
  4047. \uuline off
  4048. \uwave off
  4049. \noun off
  4050. \color none
  4051. 0.713
  4052. \end_layout
  4053. \end_inset
  4054. </cell>
  4055. </row>
  4056. <row>
  4057. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  4058. \begin_inset Text
  4059. \begin_layout Plain Layout
  4060. \family roman
  4061. \series medium
  4062. \shape up
  4063. \size normal
  4064. \emph off
  4065. \bar no
  4066. \strikeout off
  4067. \xout off
  4068. \uuline off
  4069. \uwave off
  4070. \noun off
  4071. \color none
  4072. dChip
  4073. \end_layout
  4074. \end_inset
  4075. </cell>
  4076. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  4077. \begin_inset Text
  4078. \begin_layout Plain Layout
  4079. No
  4080. \end_layout
  4081. \end_inset
  4082. </cell>
  4083. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  4084. \begin_inset Text
  4085. \begin_layout Plain Layout
  4086. \family roman
  4087. \series medium
  4088. \shape up
  4089. \size normal
  4090. \emph off
  4091. \bar no
  4092. \strikeout off
  4093. \xout off
  4094. \uuline off
  4095. \uwave off
  4096. \noun off
  4097. \color none
  4098. 0.891
  4099. \end_layout
  4100. \end_inset
  4101. </cell>
  4102. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  4103. \begin_inset Text
  4104. \begin_layout Plain Layout
  4105. \family roman
  4106. \series medium
  4107. \shape up
  4108. \size normal
  4109. \emph off
  4110. \bar no
  4111. \strikeout off
  4112. \xout off
  4113. \uuline off
  4114. \uwave off
  4115. \noun off
  4116. \color none
  4117. 0.657
  4118. \end_layout
  4119. \end_inset
  4120. </cell>
  4121. </row>
  4122. <row>
  4123. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  4124. \begin_inset Text
  4125. \begin_layout Plain Layout
  4126. \family roman
  4127. \series medium
  4128. \shape up
  4129. \size normal
  4130. \emph off
  4131. \bar no
  4132. \strikeout off
  4133. \xout off
  4134. \uuline off
  4135. \uwave off
  4136. \noun off
  4137. \color none
  4138. RMA + GRSN
  4139. \end_layout
  4140. \end_inset
  4141. </cell>
  4142. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  4143. \begin_inset Text
  4144. \begin_layout Plain Layout
  4145. No
  4146. \end_layout
  4147. \end_inset
  4148. </cell>
  4149. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  4150. \begin_inset Text
  4151. \begin_layout Plain Layout
  4152. \family roman
  4153. \series medium
  4154. \shape up
  4155. \size normal
  4156. \emph off
  4157. \bar no
  4158. \strikeout off
  4159. \xout off
  4160. \uuline off
  4161. \uwave off
  4162. \noun off
  4163. \color none
  4164. 0.816
  4165. \end_layout
  4166. \end_inset
  4167. </cell>
  4168. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  4169. \begin_inset Text
  4170. \begin_layout Plain Layout
  4171. \family roman
  4172. \series medium
  4173. \shape up
  4174. \size normal
  4175. \emph off
  4176. \bar no
  4177. \strikeout off
  4178. \xout off
  4179. \uuline off
  4180. \uwave off
  4181. \noun off
  4182. \color none
  4183. 0.750
  4184. \end_layout
  4185. \end_inset
  4186. </cell>
  4187. </row>
  4188. <row>
  4189. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  4190. \begin_inset Text
  4191. \begin_layout Plain Layout
  4192. \family roman
  4193. \series medium
  4194. \shape up
  4195. \size normal
  4196. \emph off
  4197. \bar no
  4198. \strikeout off
  4199. \xout off
  4200. \uuline off
  4201. \uwave off
  4202. \noun off
  4203. \color none
  4204. dChip + GRSN
  4205. \end_layout
  4206. \end_inset
  4207. </cell>
  4208. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  4209. \begin_inset Text
  4210. \begin_layout Plain Layout
  4211. No
  4212. \end_layout
  4213. \end_inset
  4214. </cell>
  4215. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  4216. \begin_inset Text
  4217. \begin_layout Plain Layout
  4218. \family roman
  4219. \series medium
  4220. \shape up
  4221. \size normal
  4222. \emph off
  4223. \bar no
  4224. \strikeout off
  4225. \xout off
  4226. \uuline off
  4227. \uwave off
  4228. \noun off
  4229. \color none
  4230. 0.875
  4231. \end_layout
  4232. \end_inset
  4233. </cell>
  4234. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  4235. \begin_inset Text
  4236. \begin_layout Plain Layout
  4237. \family roman
  4238. \series medium
  4239. \shape up
  4240. \size normal
  4241. \emph off
  4242. \bar no
  4243. \strikeout off
  4244. \xout off
  4245. \uuline off
  4246. \uwave off
  4247. \noun off
  4248. \color none
  4249. 0.642
  4250. \end_layout
  4251. \end_inset
  4252. </cell>
  4253. </row>
  4254. <row>
  4255. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  4256. \begin_inset Text
  4257. \begin_layout Plain Layout
  4258. \family roman
  4259. \series medium
  4260. \shape up
  4261. \size normal
  4262. \emph off
  4263. \bar no
  4264. \strikeout off
  4265. \xout off
  4266. \uuline off
  4267. \uwave off
  4268. \noun off
  4269. \color none
  4270. fRMA
  4271. \end_layout
  4272. \end_inset
  4273. </cell>
  4274. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  4275. \begin_inset Text
  4276. \begin_layout Plain Layout
  4277. Yes
  4278. \end_layout
  4279. \end_inset
  4280. </cell>
  4281. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  4282. \begin_inset Text
  4283. \begin_layout Plain Layout
  4284. \family roman
  4285. \series medium
  4286. \shape up
  4287. \size normal
  4288. \emph off
  4289. \bar no
  4290. \strikeout off
  4291. \xout off
  4292. \uuline off
  4293. \uwave off
  4294. \noun off
  4295. \color none
  4296. 0.863
  4297. \end_layout
  4298. \end_inset
  4299. </cell>
  4300. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  4301. \begin_inset Text
  4302. \begin_layout Plain Layout
  4303. \family roman
  4304. \series medium
  4305. \shape up
  4306. \size normal
  4307. \emph off
  4308. \bar no
  4309. \strikeout off
  4310. \xout off
  4311. \uuline off
  4312. \uwave off
  4313. \noun off
  4314. \color none
  4315. 0.718
  4316. \end_layout
  4317. \end_inset
  4318. </cell>
  4319. </row>
  4320. <row>
  4321. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  4322. \begin_inset Text
  4323. \begin_layout Plain Layout
  4324. \family roman
  4325. \series medium
  4326. \shape up
  4327. \size normal
  4328. \emph off
  4329. \bar no
  4330. \strikeout off
  4331. \xout off
  4332. \uuline off
  4333. \uwave off
  4334. \noun off
  4335. \color none
  4336. SCAN
  4337. \end_layout
  4338. \end_inset
  4339. </cell>
  4340. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  4341. \begin_inset Text
  4342. \begin_layout Plain Layout
  4343. Yes
  4344. \end_layout
  4345. \end_inset
  4346. </cell>
  4347. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  4348. \begin_inset Text
  4349. \begin_layout Plain Layout
  4350. \family roman
  4351. \series medium
  4352. \shape up
  4353. \size normal
  4354. \emph off
  4355. \bar no
  4356. \strikeout off
  4357. \xout off
  4358. \uuline off
  4359. \uwave off
  4360. \noun off
  4361. \color none
  4362. 0.853
  4363. \end_layout
  4364. \end_inset
  4365. </cell>
  4366. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  4367. \begin_inset Text
  4368. \begin_layout Plain Layout
  4369. \family roman
  4370. \series medium
  4371. \shape up
  4372. \size normal
  4373. \emph off
  4374. \bar no
  4375. \strikeout off
  4376. \xout off
  4377. \uuline off
  4378. \uwave off
  4379. \noun off
  4380. \color none
  4381. 0.689
  4382. \end_layout
  4383. \end_inset
  4384. </cell>
  4385. </row>
  4386. </lyxtabular>
  4387. \end_inset
  4388. \end_layout
  4389. \begin_layout Plain Layout
  4390. \begin_inset Caption Standard
  4391. \begin_layout Plain Layout
  4392. \begin_inset CommandInset label
  4393. LatexCommand label
  4394. name "tab:AUC-PAM"
  4395. \end_inset
  4396. \series bold
  4397. AUC values for internal and external validation with 6 different normalization
  4398. strategies.
  4399. \series default
  4400. Only fRMA and SCAN are single-channel normalizations.
  4401. The other 4 normalizations are for comparison.
  4402. \end_layout
  4403. \end_inset
  4404. \end_layout
  4405. \end_inset
  4406. \end_layout
  4407. \begin_layout Standard
  4408. For internal validation, the 6 methods' AUC values ranged from 0.816 to 0.891,
  4409. as shown in Table
  4410. \begin_inset CommandInset ref
  4411. LatexCommand ref
  4412. reference "tab:AUC-PAM"
  4413. plural "false"
  4414. caps "false"
  4415. noprefix "false"
  4416. \end_inset
  4417. .
  4418. Among the non-single-channel normalizations, dChip outperformed RMA, while
  4419. GRSN reduced the AUC values for both dChip and RMA.
  4420. Both single-channel methods, fRMA and SCAN, slightly outperformed RMA,
  4421. with fRMA ahead of SCAN.
  4422. However, the difference between RMA and fRMA is still quite small.
  4423. Figure
  4424. \begin_inset CommandInset ref
  4425. LatexCommand ref
  4426. reference "fig:ROC-PAM-int"
  4427. plural "false"
  4428. caps "false"
  4429. noprefix "false"
  4430. \end_inset
  4431. shows that the ROC curves for RMA, dChip, and fRMA look very similar and
  4432. relatively smooth, while both GRSN curves and the curve for SCAN have a
  4433. more jagged appearance.
  4434. \end_layout
  4435. \begin_layout Standard
  4436. \begin_inset Float figure
  4437. placement tb
  4438. wide false
  4439. sideways false
  4440. status collapsed
  4441. \begin_layout Plain Layout
  4442. \align center
  4443. \begin_inset Graphics
  4444. filename graphics/PAM/ROC-TXvsAR-external.pdf
  4445. lyxscale 50
  4446. width 100col%
  4447. groupId colwidth
  4448. \end_inset
  4449. \end_layout
  4450. \begin_layout Plain Layout
  4451. \begin_inset Caption Standard
  4452. \begin_layout Plain Layout
  4453. \series bold
  4454. \begin_inset CommandInset label
  4455. LatexCommand label
  4456. name "fig:ROC-PAM-ext"
  4457. \end_inset
  4458. ROC curve for PAM on external validation data using different normalization
  4459. strategies
  4460. \end_layout
  4461. \end_inset
  4462. \end_layout
  4463. \end_inset
  4464. \end_layout
  4465. \begin_layout Standard
  4466. For external validation, as expected, all the AUC values are lower than
  4467. the internal validations, ranging from 0.642 to 0.750 (Table
  4468. \begin_inset CommandInset ref
  4469. LatexCommand ref
  4470. reference "tab:AUC-PAM"
  4471. plural "false"
  4472. caps "false"
  4473. noprefix "false"
  4474. \end_inset
  4475. ).
  4476. With or without GRSN, RMA shows its dominance over dChip in this more challengi
  4477. ng test.
  4478. Unlike in the internal validation, GRSN actually improves the classifier
  4479. performance for RMA, although it does not for dChip.
  4480. Once again, both single-channel methods perform about on par with RMA,
  4481. with fRMA performing slightly better and SCAN performing a bit worse.
  4482. Figure
  4483. \begin_inset CommandInset ref
  4484. LatexCommand ref
  4485. reference "fig:ROC-PAM-ext"
  4486. plural "false"
  4487. caps "false"
  4488. noprefix "false"
  4489. \end_inset
  4490. shows the ROC curves for the external validation test.
  4491. As expected, none of them are as clean-looking as the internal validation
  4492. ROC curves.
  4493. The curves for RMA, RMA+GRSN, and fRMA all look similar, while the other
  4494. curves look more divergent.
  4495. \end_layout
  4496. \begin_layout Subsection
  4497. fRMA with custom-generated vectors enables normalization on hthgu133pluspm
  4498. \end_layout
  4499. \begin_layout Standard
  4500. \begin_inset Float figure
  4501. wide false
  4502. sideways false
  4503. status open
  4504. \begin_layout Plain Layout
  4505. \align center
  4506. \begin_inset Float figure
  4507. placement tb
  4508. wide false
  4509. sideways false
  4510. status collapsed
  4511. \begin_layout Plain Layout
  4512. \align center
  4513. \begin_inset Graphics
  4514. filename graphics/frma-pax-bx/batchsize_batches.pdf
  4515. lyxscale 50
  4516. height 35theight%
  4517. groupId frmatools-subfig
  4518. \end_inset
  4519. \end_layout
  4520. \begin_layout Plain Layout
  4521. \begin_inset Caption Standard
  4522. \begin_layout Plain Layout
  4523. \begin_inset CommandInset label
  4524. LatexCommand label
  4525. name "fig:batch-size-batches"
  4526. \end_inset
  4527. \series bold
  4528. Number of batches usable in fRMA probe weight learning as a function of
  4529. batch size.
  4530. \end_layout
  4531. \end_inset
  4532. \end_layout
  4533. \end_inset
  4534. \end_layout
  4535. \begin_layout Plain Layout
  4536. \align center
  4537. \begin_inset Float figure
  4538. placement tb
  4539. wide false
  4540. sideways false
  4541. status collapsed
  4542. \begin_layout Plain Layout
  4543. \align center
  4544. \begin_inset Graphics
  4545. filename graphics/frma-pax-bx/batchsize_samples.pdf
  4546. lyxscale 50
  4547. height 35theight%
  4548. groupId frmatools-subfig
  4549. \end_inset
  4550. \end_layout
  4551. \begin_layout Plain Layout
  4552. \begin_inset Caption Standard
  4553. \begin_layout Plain Layout
  4554. \begin_inset CommandInset label
  4555. LatexCommand label
  4556. name "fig:batch-size-samples"
  4557. \end_inset
  4558. \series bold
  4559. Number of samples usable in fRMA probe weight learning as a function of
  4560. batch size.
  4561. \end_layout
  4562. \end_inset
  4563. \end_layout
  4564. \end_inset
  4565. \end_layout
  4566. \begin_layout Plain Layout
  4567. \begin_inset Caption Standard
  4568. \begin_layout Plain Layout
  4569. \series bold
  4570. \begin_inset CommandInset label
  4571. LatexCommand label
  4572. name "fig:frmatools-batch-size"
  4573. \end_inset
  4574. Effect of batch size selection on number of batches and number of samples
  4575. included in fRMA probe weight learning.
  4576. \series default
  4577. For batch sizes ranging from 3 to 15, the number of batches (a) and samples
  4578. (b) included in probe weight training were plotted for biopsy (BX) and
  4579. blood (PAX) samples.
  4580. The selected batch size, 5, is marked with a dotted vertical line.
  4581. \end_layout
  4582. \end_inset
  4583. \end_layout
  4584. \end_inset
  4585. \end_layout
  4586. \begin_layout Standard
  4587. In order to enable use of fRMA to normalize hthgu133pluspm, a custom set
  4588. of fRMA vectors was created.
  4589. First, an appropriate batch size was chosen by looking at the number of
  4590. batches and number of samples included as a function of batch size (Figure
  4591. \begin_inset CommandInset ref
  4592. LatexCommand ref
  4593. reference "fig:frmatools-batch-size"
  4594. plural "false"
  4595. caps "false"
  4596. noprefix "false"
  4597. \end_inset
  4598. ).
  4599. For a given batch size, all batches with fewer samples that the chosen
  4600. size must be ignored during training, while larger batches must be randomly
  4601. downsampled to the chosen size.
  4602. Hence, the number of samples included for a given batch size equals the
  4603. batch size times the number of batches with at least that many samples.
  4604. From Figure
  4605. \begin_inset CommandInset ref
  4606. LatexCommand ref
  4607. reference "fig:batch-size-samples"
  4608. plural "false"
  4609. caps "false"
  4610. noprefix "false"
  4611. \end_inset
  4612. , it is apparent that that a batch size of 8 maximizes the number of samples
  4613. included in training.
  4614. Increasing the batch size beyond this causes too many smaller batches to
  4615. be excluded, reducing the total number of samples for both tissue types.
  4616. However, a batch size of 8 is not necessarily optimal.
  4617. The article introducing frmaTools concluded that it was highly advantageous
  4618. to use a smaller batch size in order to include more batches, even at the
  4619. expense of including fewer total samples in training
  4620. \begin_inset CommandInset citation
  4621. LatexCommand cite
  4622. key "McCall2011"
  4623. literal "false"
  4624. \end_inset
  4625. .
  4626. To strike an appropriate balance between more batches and more samples,
  4627. a batch size of 5 was chosen.
  4628. For both blood and biopsy samples, this increased the number of batches
  4629. included by 10, with only a modest reduction in the number of samples compared
  4630. to a batch size of 8.
  4631. With a batch size of 5, 26 batches of biopsy samples and 46 batches of
  4632. blood samples were available.
  4633. \end_layout
  4634. \begin_layout Standard
  4635. \begin_inset Float figure
  4636. wide false
  4637. sideways false
  4638. status open
  4639. \begin_layout Plain Layout
  4640. \begin_inset Float figure
  4641. wide false
  4642. sideways false
  4643. status collapsed
  4644. \begin_layout Plain Layout
  4645. \align center
  4646. \begin_inset Graphics
  4647. filename graphics/frma-pax-bx/M-BX-violin.pdf
  4648. lyxscale 40
  4649. width 45col%
  4650. groupId m-violin
  4651. \end_inset
  4652. \end_layout
  4653. \begin_layout Plain Layout
  4654. \begin_inset Caption Standard
  4655. \begin_layout Plain Layout
  4656. \begin_inset CommandInset label
  4657. LatexCommand label
  4658. name "fig:m-bx-violin"
  4659. \end_inset
  4660. \series bold
  4661. Violin plot of inter-normalization log ratios for biopsy samples.
  4662. \end_layout
  4663. \end_inset
  4664. \end_layout
  4665. \end_inset
  4666. \begin_inset space \hfill{}
  4667. \end_inset
  4668. \begin_inset Float figure
  4669. wide false
  4670. sideways false
  4671. status collapsed
  4672. \begin_layout Plain Layout
  4673. \align center
  4674. \begin_inset Graphics
  4675. filename graphics/frma-pax-bx/M-PAX-violin.pdf
  4676. lyxscale 40
  4677. width 45col%
  4678. groupId m-violin
  4679. \end_inset
  4680. \end_layout
  4681. \begin_layout Plain Layout
  4682. \begin_inset Caption Standard
  4683. \begin_layout Plain Layout
  4684. \begin_inset CommandInset label
  4685. LatexCommand label
  4686. name "fig:m-pax-violin"
  4687. \end_inset
  4688. \series bold
  4689. Violin plot of inter-normalization log ratios for blood samples.
  4690. \end_layout
  4691. \end_inset
  4692. \end_layout
  4693. \end_inset
  4694. \end_layout
  4695. \begin_layout Plain Layout
  4696. \begin_inset Caption Standard
  4697. \begin_layout Plain Layout
  4698. \series bold
  4699. Violin plot of log ratios between normalizations for 20 biopsy samples.
  4700. \series default
  4701. Each of 20 randomly selected samples was normalized with RMA and with 5
  4702. different sets of fRMA vectors.
  4703. The distribution of log ratios between normalized expression values, aggregated
  4704. across all 20 arrays, was plotted for each pair of normalizations.
  4705. \end_layout
  4706. \end_inset
  4707. \end_layout
  4708. \end_inset
  4709. \end_layout
  4710. \begin_layout Standard
  4711. Since fRMA training requires equal-size batches, larger batches are downsampled
  4712. randomly.
  4713. This introduces a nondeterministic step in the generation of normalization
  4714. vectors.
  4715. To show that this randomness does not substantially change the outcome,
  4716. the random downsampling and subsequent vector learning was repeated 5 times,
  4717. with a different random seed each time.
  4718. 20 samples were selected at random as a test set and normalized with each
  4719. of the 5 sets of fRMA normalization vectors as well as ordinary RMA, and
  4720. the normalized expression values were compared across normalizations.
  4721. Figure
  4722. \begin_inset CommandInset ref
  4723. LatexCommand ref
  4724. reference "fig:m-bx-violin"
  4725. plural "false"
  4726. caps "false"
  4727. noprefix "false"
  4728. \end_inset
  4729. shows a summary of these comparisons for biopsy samples.
  4730. Comparing RMA to each of the 5 fRMA normalizations, the distribution of
  4731. log ratios is somewhat wide, indicating that the normalizations disagree
  4732. on the expression values of a fair number of probe sets.
  4733. In contrast, comparisons of fRMA against fRMA, the vast mojority of probe
  4734. sets have very small log ratios, indicating a very high agreement between
  4735. the normalized values generated by the two normalizations.
  4736. This shows that the fRMA normalization's behavior is not very sensitive
  4737. to the random downsampling of larger batches during training.
  4738. \end_layout
  4739. \begin_layout Standard
  4740. \begin_inset Float figure
  4741. wide false
  4742. sideways false
  4743. status open
  4744. \begin_layout Plain Layout
  4745. \align center
  4746. \begin_inset Float figure
  4747. wide false
  4748. sideways false
  4749. status collapsed
  4750. \begin_layout Plain Layout
  4751. \align center
  4752. \begin_inset Graphics
  4753. filename graphics/frma-pax-bx/MA-BX-RMA.fRMA-RASTER.png
  4754. lyxscale 10
  4755. width 45col%
  4756. groupId ma-frma
  4757. \end_inset
  4758. \end_layout
  4759. \begin_layout Plain Layout
  4760. \begin_inset Caption Standard
  4761. \begin_layout Plain Layout
  4762. \begin_inset CommandInset label
  4763. LatexCommand label
  4764. name "fig:ma-bx-rma-frma"
  4765. \end_inset
  4766. \series bold
  4767. RMA vs.
  4768. fRMA for biopsy samples.
  4769. \end_layout
  4770. \end_inset
  4771. \end_layout
  4772. \end_inset
  4773. \begin_inset space \hfill{}
  4774. \end_inset
  4775. \begin_inset Float figure
  4776. wide false
  4777. sideways false
  4778. status collapsed
  4779. \begin_layout Plain Layout
  4780. \align center
  4781. \begin_inset Graphics
  4782. filename graphics/frma-pax-bx/MA-BX-fRMA.fRMA-RASTER.png
  4783. lyxscale 10
  4784. width 45col%
  4785. groupId ma-frma
  4786. \end_inset
  4787. \end_layout
  4788. \begin_layout Plain Layout
  4789. \begin_inset Caption Standard
  4790. \begin_layout Plain Layout
  4791. \begin_inset CommandInset label
  4792. LatexCommand label
  4793. name "fig:ma-bx-frma-frma"
  4794. \end_inset
  4795. \series bold
  4796. fRMA vs fRMA for biopsy samples.
  4797. \series default
  4798. Two different fRMA normalizations using vectors from two different batch
  4799. samplings were compared.
  4800. \end_layout
  4801. \end_inset
  4802. \end_layout
  4803. \end_inset
  4804. \end_layout
  4805. \begin_layout Plain Layout
  4806. \align center
  4807. \begin_inset Float figure
  4808. wide false
  4809. sideways false
  4810. status collapsed
  4811. \begin_layout Plain Layout
  4812. \align center
  4813. \begin_inset Graphics
  4814. filename graphics/frma-pax-bx/MA-PAX-RMA.fRMA-RASTER.png
  4815. lyxscale 10
  4816. width 45col%
  4817. groupId ma-frma
  4818. \end_inset
  4819. \end_layout
  4820. \begin_layout Plain Layout
  4821. \begin_inset Caption Standard
  4822. \begin_layout Plain Layout
  4823. \begin_inset CommandInset label
  4824. LatexCommand label
  4825. name "fig:MA-PAX-rma-frma"
  4826. \end_inset
  4827. \series bold
  4828. RMA vs.
  4829. fRMA for blood samples.
  4830. \end_layout
  4831. \end_inset
  4832. \end_layout
  4833. \end_inset
  4834. \begin_inset space \hfill{}
  4835. \end_inset
  4836. \begin_inset Float figure
  4837. wide false
  4838. sideways false
  4839. status collapsed
  4840. \begin_layout Plain Layout
  4841. \align center
  4842. \begin_inset Graphics
  4843. filename graphics/frma-pax-bx/MA-PAX-fRMA.fRMA-RASTER.png
  4844. lyxscale 10
  4845. width 45col%
  4846. groupId ma-frma
  4847. \end_inset
  4848. \end_layout
  4849. \begin_layout Plain Layout
  4850. \begin_inset Caption Standard
  4851. \begin_layout Plain Layout
  4852. \begin_inset CommandInset label
  4853. LatexCommand label
  4854. name "fig:MA-PAX-frma-frma"
  4855. \end_inset
  4856. \series bold
  4857. fRMA vs fRMA for blood samples.
  4858. \series default
  4859. Two different fRMA normalizations using vectors from two different batch
  4860. samplings were compared.
  4861. \end_layout
  4862. \end_inset
  4863. \end_layout
  4864. \end_inset
  4865. \end_layout
  4866. \begin_layout Plain Layout
  4867. \begin_inset Caption Standard
  4868. \begin_layout Plain Layout
  4869. \series bold
  4870. \begin_inset CommandInset label
  4871. LatexCommand label
  4872. name "fig:Representative-MA-plots"
  4873. \end_inset
  4874. Representative MA plots comparing RMA and custom fRMA normalizations.
  4875. \series default
  4876. For each plot, 20 samples were normalized using 2 different normalizations,
  4877. and then averages and log ratios were computed between the two different
  4878. normalizations for every probe.
  4879. Density of points is represented by darkness of shading, and individual
  4880. outlier points are plotted.
  4881. \end_layout
  4882. \end_inset
  4883. \end_layout
  4884. \end_inset
  4885. \end_layout
  4886. \begin_layout Standard
  4887. Figure
  4888. \begin_inset CommandInset ref
  4889. LatexCommand ref
  4890. reference "fig:ma-bx-rma-frma"
  4891. plural "false"
  4892. caps "false"
  4893. noprefix "false"
  4894. \end_inset
  4895. shows an MA plot of the RMA-normalized values against the fRMA-normalized
  4896. values for the same probe sets and arrays, corresponding to the first row
  4897. of Figure
  4898. \begin_inset CommandInset ref
  4899. LatexCommand ref
  4900. reference "fig:m-bx-violin"
  4901. plural "false"
  4902. caps "false"
  4903. noprefix "false"
  4904. \end_inset
  4905. .
  4906. This MA plot shows that not only is there a wide distribution of M-values,
  4907. but the trend of M-values is dependent on the average normalized intensity.
  4908. This is expected, since the overall trend represents the differences in
  4909. the quantile normalization step.
  4910. When running RMA, only the quantiles for these specific 20 arrays are used,
  4911. while for fRMA the quantile distribution is taking from all arrays used
  4912. in training.
  4913. Figure
  4914. \begin_inset CommandInset ref
  4915. LatexCommand ref
  4916. reference "fig:ma-bx-frma-frma"
  4917. plural "false"
  4918. caps "false"
  4919. noprefix "false"
  4920. \end_inset
  4921. shows a similar MA plot comparing 2 different fRMA normalizations, correspondin
  4922. g to the 6th row of Figure
  4923. \begin_inset CommandInset ref
  4924. LatexCommand ref
  4925. reference "fig:m-bx-violin"
  4926. plural "false"
  4927. caps "false"
  4928. noprefix "false"
  4929. \end_inset
  4930. .
  4931. The MA plot is very tightly centered around zero with no visible trend.
  4932. Figures
  4933. \begin_inset CommandInset ref
  4934. LatexCommand ref
  4935. reference "fig:m-pax-violin"
  4936. plural "false"
  4937. caps "false"
  4938. noprefix "false"
  4939. \end_inset
  4940. ,
  4941. \begin_inset CommandInset ref
  4942. LatexCommand ref
  4943. reference "fig:MA-PAX-rma-frma"
  4944. plural "false"
  4945. caps "false"
  4946. noprefix "false"
  4947. \end_inset
  4948. , and
  4949. \begin_inset CommandInset ref
  4950. LatexCommand ref
  4951. reference "fig:ma-bx-frma-frma"
  4952. plural "false"
  4953. caps "false"
  4954. noprefix "false"
  4955. \end_inset
  4956. show exactly the same information for the blood samples, once again comparing
  4957. the normalized expression values between normalizations for all probe sets
  4958. across 20 randomly selected test arrays.
  4959. Once again, there is a wider distribution of log ratios between RMA-normalized
  4960. values and fRMA-normalized, and a much tighter distribution when comparing
  4961. different fRMA normalizations to each other, indicating that the fRMA training
  4962. process is robust to random batch downsampling for the blood samples as
  4963. well.
  4964. \end_layout
  4965. \begin_layout Subsection
  4966. SVA, voom, and array weights improve model fit for methylation array data
  4967. \end_layout
  4968. \begin_layout Standard
  4969. \begin_inset Float figure
  4970. wide false
  4971. sideways true
  4972. status open
  4973. \begin_layout Plain Layout
  4974. \begin_inset Flex TODO Note (inline)
  4975. status open
  4976. \begin_layout Plain Layout
  4977. Fix axis labels:
  4978. \begin_inset Quotes eld
  4979. \end_inset
  4980. log2 M-value
  4981. \begin_inset Quotes erd
  4982. \end_inset
  4983. is redundant because M-values are already log scale
  4984. \end_layout
  4985. \end_inset
  4986. \end_layout
  4987. \begin_layout Plain Layout
  4988. \begin_inset Float figure
  4989. wide false
  4990. sideways false
  4991. status collapsed
  4992. \begin_layout Plain Layout
  4993. \align center
  4994. \begin_inset Graphics
  4995. filename graphics/methylvoom/unadj.dupcor/meanvar-trends-PAGE1-CROP-RASTER.png
  4996. lyxscale 15
  4997. width 30col%
  4998. groupId voomaw-subfig
  4999. \end_inset
  5000. \end_layout
  5001. \begin_layout Plain Layout
  5002. \begin_inset Caption Standard
  5003. \begin_layout Plain Layout
  5004. \series bold
  5005. \begin_inset CommandInset label
  5006. LatexCommand label
  5007. name "fig:meanvar-basic"
  5008. \end_inset
  5009. Mean-variance trend for analysis A.
  5010. \end_layout
  5011. \end_inset
  5012. \end_layout
  5013. \end_inset
  5014. \begin_inset space \hfill{}
  5015. \end_inset
  5016. \begin_inset Float figure
  5017. wide false
  5018. sideways false
  5019. status collapsed
  5020. \begin_layout Plain Layout
  5021. \align center
  5022. \begin_inset Graphics
  5023. filename graphics/methylvoom/unadj.dupcor.sva.aw/meanvar-trends-PAGE1-CROP-RASTER.png
  5024. lyxscale 15
  5025. width 30col%
  5026. groupId voomaw-subfig
  5027. \end_inset
  5028. \end_layout
  5029. \begin_layout Plain Layout
  5030. \begin_inset Caption Standard
  5031. \begin_layout Plain Layout
  5032. \series bold
  5033. \begin_inset CommandInset label
  5034. LatexCommand label
  5035. name "fig:meanvar-sva-aw"
  5036. \end_inset
  5037. Mean-variance trend for analysis B.
  5038. \end_layout
  5039. \end_inset
  5040. \end_layout
  5041. \end_inset
  5042. \begin_inset space \hfill{}
  5043. \end_inset
  5044. \begin_inset Float figure
  5045. wide false
  5046. sideways false
  5047. status collapsed
  5048. \begin_layout Plain Layout
  5049. \align center
  5050. \begin_inset Graphics
  5051. filename graphics/methylvoom/unadj.dupcor.sva.voomaw/meanvar-trends-PAGE2-CROP-RASTER.png
  5052. lyxscale 15
  5053. width 30col%
  5054. groupId voomaw-subfig
  5055. \end_inset
  5056. \end_layout
  5057. \begin_layout Plain Layout
  5058. \begin_inset Caption Standard
  5059. \begin_layout Plain Layout
  5060. \series bold
  5061. \begin_inset CommandInset label
  5062. LatexCommand label
  5063. name "fig:meanvar-sva-voomaw"
  5064. \end_inset
  5065. Mean-variance trend after voom modeling in analysis C.
  5066. \end_layout
  5067. \end_inset
  5068. \end_layout
  5069. \end_inset
  5070. \end_layout
  5071. \begin_layout Plain Layout
  5072. \begin_inset Caption Standard
  5073. \begin_layout Plain Layout
  5074. \series bold
  5075. Mean-variance trend modeling in methylation array data.
  5076. \series default
  5077. The log2(standard deviation) for each probe is plotted against the probe's
  5078. average M-value across all samples as a black point, with some transparency
  5079. to make overplotting more visible, since there are about 450,000 points.
  5080. Density of points is also indicated by the dark blue contour lines.
  5081. The prior variance trend estimated by eBayes is shown in light blue, while
  5082. the lowess trend of the points is shown in red.
  5083. \end_layout
  5084. \end_inset
  5085. \end_layout
  5086. \end_inset
  5087. \end_layout
  5088. \begin_layout Standard
  5089. Figure
  5090. \begin_inset CommandInset ref
  5091. LatexCommand ref
  5092. reference "fig:meanvar-basic"
  5093. plural "false"
  5094. caps "false"
  5095. noprefix "false"
  5096. \end_inset
  5097. shows the relationship between the mean M-value and the standard deviation
  5098. calculated for each probe in the methylation array data set.
  5099. A few features of the data are apparent.
  5100. First, the data are very strongly bimodal, with peaks in the density around
  5101. M-values of +4 and -4.
  5102. These modes correspond to methylation sites that are nearly 100% methylated
  5103. and nearly 100% unmethylated, respectively.
  5104. The strong bomodality indicates that a majority of probes interrogate sites
  5105. that fall into one of these two categories.
  5106. The points in between these modes represent sites that are either partially
  5107. methylated in many samples, or are fully methylated in some samples and
  5108. fully unmethylated in other samples, or some combination.
  5109. The next visible feature of the data is the W-shaped variance trend.
  5110. The upticks in the variance trend on either side are expected, based on
  5111. the sigmoid transformation exaggerating small differences at extreme M-values
  5112. (Figure
  5113. \begin_inset CommandInset ref
  5114. LatexCommand ref
  5115. reference "fig:Sigmoid-beta-m-mapping"
  5116. plural "false"
  5117. caps "false"
  5118. noprefix "false"
  5119. \end_inset
  5120. ).
  5121. However, the uptick in the center is interesting: it indicates that sites
  5122. that are not constitutitively methylated or unmethylated have a higher
  5123. variance.
  5124. This could be a genuine biological effect, or it could be spurious noise
  5125. that is only observable at sites with varying methylation.
  5126. \end_layout
  5127. \begin_layout Standard
  5128. In Figure
  5129. \begin_inset CommandInset ref
  5130. LatexCommand ref
  5131. reference "fig:meanvar-sva-aw"
  5132. plural "false"
  5133. caps "false"
  5134. noprefix "false"
  5135. \end_inset
  5136. , we see the mean-variance trend for the same methylation array data, this
  5137. time with surrogate variables and sample quality weights estimated from
  5138. the data and included in the model.
  5139. As expected, the overall average variance is smaller, since the surrogate
  5140. variables account for some of the variance.
  5141. In addition, the uptick in variance in the middle of the M-value range
  5142. has disappeared, turning the W shape into a wide U shape.
  5143. This indicates that the excess variance in the probes with intermediate
  5144. M-values was explained by systematic variations not correlated with known
  5145. covariates, and these variations were modeled by the surrogate variables.
  5146. The result is a nearly flat variance trend for the entire intermediate
  5147. M-value range from about -3 to +3.
  5148. In contrast, the excess variance at the extremes was not
  5149. \begin_inset Quotes eld
  5150. \end_inset
  5151. absorbed
  5152. \begin_inset Quotes erd
  5153. \end_inset
  5154. by the surrogate variables and remains in the plot, indicating that this
  5155. variation has no systematic component: probes with extreme M-values are
  5156. uniformly more variable across all samples, as expected.
  5157. \end_layout
  5158. \begin_layout Standard
  5159. Figure
  5160. \begin_inset CommandInset ref
  5161. LatexCommand ref
  5162. reference "fig:meanvar-sva-voomaw"
  5163. plural "false"
  5164. caps "false"
  5165. noprefix "false"
  5166. \end_inset
  5167. shows the mean-variance trend after fitting the model with the observation
  5168. weights assigned by voom based on the mean-variance trend shown in Figure
  5169. \begin_inset CommandInset ref
  5170. LatexCommand ref
  5171. reference "fig:meanvar-sva-aw"
  5172. plural "false"
  5173. caps "false"
  5174. noprefix "false"
  5175. \end_inset
  5176. .
  5177. As expected, the weights exactly counteract the trend in the data, resulting
  5178. in a nearly flat trend centered vertically at 1 (i.e.
  5179. 0 on the log scale).
  5180. This shows that the observations with extreme M-values have been appropriately
  5181. down-weighted to account for the fact that the noise in those observations
  5182. has been amplified by the non-linear M-value transformation.
  5183. In turn, this gives relatively more weight to observervations in the middle
  5184. region, which are more likely to correspond to probes measuring interesting
  5185. biology (not constitutively methylated or unmethylated).
  5186. \end_layout
  5187. \begin_layout Standard
  5188. \begin_inset Float table
  5189. wide false
  5190. sideways false
  5191. status collapsed
  5192. \begin_layout Plain Layout
  5193. \align center
  5194. \begin_inset Tabular
  5195. <lyxtabular version="3" rows="5" columns="3">
  5196. <features tabularvalignment="middle">
  5197. <column alignment="center" valignment="top">
  5198. <column alignment="center" valignment="top">
  5199. <column alignment="center" valignment="top">
  5200. <row>
  5201. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  5202. \begin_inset Text
  5203. \begin_layout Plain Layout
  5204. Covariate
  5205. \end_layout
  5206. \end_inset
  5207. </cell>
  5208. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  5209. \begin_inset Text
  5210. \begin_layout Plain Layout
  5211. Test used
  5212. \end_layout
  5213. \end_inset
  5214. </cell>
  5215. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  5216. \begin_inset Text
  5217. \begin_layout Plain Layout
  5218. p-value
  5219. \end_layout
  5220. \end_inset
  5221. </cell>
  5222. </row>
  5223. <row>
  5224. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  5225. \begin_inset Text
  5226. \begin_layout Plain Layout
  5227. Transplant Status
  5228. \end_layout
  5229. \end_inset
  5230. </cell>
  5231. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  5232. \begin_inset Text
  5233. \begin_layout Plain Layout
  5234. F-test
  5235. \end_layout
  5236. \end_inset
  5237. </cell>
  5238. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  5239. \begin_inset Text
  5240. \begin_layout Plain Layout
  5241. 0.404
  5242. \end_layout
  5243. \end_inset
  5244. </cell>
  5245. </row>
  5246. <row>
  5247. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  5248. \begin_inset Text
  5249. \begin_layout Plain Layout
  5250. Diabetes Diagnosis
  5251. \end_layout
  5252. \end_inset
  5253. </cell>
  5254. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  5255. \begin_inset Text
  5256. \begin_layout Plain Layout
  5257. t-test
  5258. \end_layout
  5259. \end_inset
  5260. </cell>
  5261. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  5262. \begin_inset Text
  5263. \begin_layout Plain Layout
  5264. 0.00106
  5265. \end_layout
  5266. \end_inset
  5267. </cell>
  5268. </row>
  5269. <row>
  5270. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  5271. \begin_inset Text
  5272. \begin_layout Plain Layout
  5273. Sex
  5274. \end_layout
  5275. \end_inset
  5276. </cell>
  5277. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  5278. \begin_inset Text
  5279. \begin_layout Plain Layout
  5280. t-test
  5281. \end_layout
  5282. \end_inset
  5283. </cell>
  5284. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  5285. \begin_inset Text
  5286. \begin_layout Plain Layout
  5287. 0.148
  5288. \end_layout
  5289. \end_inset
  5290. </cell>
  5291. </row>
  5292. <row>
  5293. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  5294. \begin_inset Text
  5295. \begin_layout Plain Layout
  5296. Age
  5297. \end_layout
  5298. \end_inset
  5299. </cell>
  5300. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  5301. \begin_inset Text
  5302. \begin_layout Plain Layout
  5303. linear regression
  5304. \end_layout
  5305. \end_inset
  5306. </cell>
  5307. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  5308. \begin_inset Text
  5309. \begin_layout Plain Layout
  5310. 0.212
  5311. \end_layout
  5312. \end_inset
  5313. </cell>
  5314. </row>
  5315. </lyxtabular>
  5316. \end_inset
  5317. \end_layout
  5318. \begin_layout Plain Layout
  5319. \begin_inset Caption Standard
  5320. \begin_layout Plain Layout
  5321. \series bold
  5322. \begin_inset CommandInset label
  5323. LatexCommand label
  5324. name "tab:weight-covariate-tests"
  5325. \end_inset
  5326. Association of sample weights with clinical covariates in methylation array
  5327. data.
  5328. \series default
  5329. Computed sample quality log weights were tested for significant association
  5330. with each of the variables in the model (1st column).
  5331. An appropriate test was selected for each variable (2nd column).
  5332. P-values for significant association are shown in the 3rd column.
  5333. \end_layout
  5334. \end_inset
  5335. \end_layout
  5336. \end_inset
  5337. \end_layout
  5338. \begin_layout Standard
  5339. \begin_inset Flex TODO Note (inline)
  5340. status open
  5341. \begin_layout Plain Layout
  5342. Redo the sample weight boxplot with notches and without fill colors (and
  5343. update the legend)
  5344. \end_layout
  5345. \end_inset
  5346. \end_layout
  5347. \begin_layout Standard
  5348. \begin_inset Float figure
  5349. wide false
  5350. sideways false
  5351. status collapsed
  5352. \begin_layout Plain Layout
  5353. \align center
  5354. \begin_inset Graphics
  5355. filename graphics/methylvoom/unadj.dupcor.sva.voomaw/sample-weights-PAGE3-CROP.pdf
  5356. lyxscale 50
  5357. width 100col%
  5358. groupId colwidth
  5359. \end_inset
  5360. \end_layout
  5361. \begin_layout Plain Layout
  5362. \begin_inset Caption Standard
  5363. \begin_layout Plain Layout
  5364. \begin_inset CommandInset label
  5365. LatexCommand label
  5366. name "fig:diabetes-sample-weights"
  5367. \end_inset
  5368. \series bold
  5369. Boxplot of sample quality weights grouped by diabetes diagnosis.
  5370. \series default
  5371. Sample were grouped based on diabetes diagnosis, and the distribution of
  5372. sample quality weights for each diagnosis was plotted.
  5373. \end_layout
  5374. \end_inset
  5375. \end_layout
  5376. \begin_layout Plain Layout
  5377. \end_layout
  5378. \end_inset
  5379. \end_layout
  5380. \begin_layout Standard
  5381. To determine whether any of the known experimental factors had an impact
  5382. on data quality, the sample quality weights estimated from the data were
  5383. tested for association with each of the experimental factors (Table
  5384. \begin_inset CommandInset ref
  5385. LatexCommand ref
  5386. reference "tab:weight-covariate-tests"
  5387. plural "false"
  5388. caps "false"
  5389. noprefix "false"
  5390. \end_inset
  5391. ).
  5392. Diabetes diagnosis was found to have a potentially significant association
  5393. with the sample weights, with a t-test p-value of
  5394. \begin_inset Formula $1.06\times10^{-3}$
  5395. \end_inset
  5396. .
  5397. Figure
  5398. \begin_inset CommandInset ref
  5399. LatexCommand ref
  5400. reference "fig:diabetes-sample-weights"
  5401. plural "false"
  5402. caps "false"
  5403. noprefix "false"
  5404. \end_inset
  5405. shows the distribution of sample weights grouped by diabetes diagnosis.
  5406. The samples from patients with Type 2 diabetes were assigned significantly
  5407. lower weights than those from patients with Type 1 diabetes.
  5408. This indicates that the type 2 diabetes samples had an overall higher variance
  5409. on average across all probes.
  5410. \end_layout
  5411. \begin_layout Standard
  5412. \begin_inset Float table
  5413. wide false
  5414. sideways false
  5415. status collapsed
  5416. \begin_layout Plain Layout
  5417. \align center
  5418. \begin_inset Flex TODO Note (inline)
  5419. status open
  5420. \begin_layout Plain Layout
  5421. Consider transposing this table and the next one
  5422. \end_layout
  5423. \end_inset
  5424. \end_layout
  5425. \begin_layout Plain Layout
  5426. \align center
  5427. \begin_inset Tabular
  5428. <lyxtabular version="3" rows="5" columns="4">
  5429. <features tabularvalignment="middle">
  5430. <column alignment="center" valignment="top">
  5431. <column alignment="center" valignment="top">
  5432. <column alignment="center" valignment="top">
  5433. <column alignment="center" valignment="top">
  5434. <row>
  5435. <cell alignment="center" valignment="top" usebox="none">
  5436. \begin_inset Text
  5437. \begin_layout Plain Layout
  5438. \end_layout
  5439. \end_inset
  5440. </cell>
  5441. <cell multicolumn="1" alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  5442. \begin_inset Text
  5443. \begin_layout Plain Layout
  5444. Analysis
  5445. \end_layout
  5446. \end_inset
  5447. </cell>
  5448. <cell multicolumn="2" alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  5449. \begin_inset Text
  5450. \begin_layout Plain Layout
  5451. \end_layout
  5452. \end_inset
  5453. </cell>
  5454. <cell multicolumn="2" alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  5455. \begin_inset Text
  5456. \begin_layout Plain Layout
  5457. \end_layout
  5458. \end_inset
  5459. </cell>
  5460. </row>
  5461. <row>
  5462. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  5463. \begin_inset Text
  5464. \begin_layout Plain Layout
  5465. Contrast
  5466. \end_layout
  5467. \end_inset
  5468. </cell>
  5469. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  5470. \begin_inset Text
  5471. \begin_layout Plain Layout
  5472. A
  5473. \end_layout
  5474. \end_inset
  5475. </cell>
  5476. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  5477. \begin_inset Text
  5478. \begin_layout Plain Layout
  5479. B
  5480. \end_layout
  5481. \end_inset
  5482. </cell>
  5483. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  5484. \begin_inset Text
  5485. \begin_layout Plain Layout
  5486. C
  5487. \end_layout
  5488. \end_inset
  5489. </cell>
  5490. </row>
  5491. <row>
  5492. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  5493. \begin_inset Text
  5494. \begin_layout Plain Layout
  5495. TX vs AR
  5496. \end_layout
  5497. \end_inset
  5498. </cell>
  5499. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  5500. \begin_inset Text
  5501. \begin_layout Plain Layout
  5502. 0
  5503. \end_layout
  5504. \end_inset
  5505. </cell>
  5506. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  5507. \begin_inset Text
  5508. \begin_layout Plain Layout
  5509. 25
  5510. \end_layout
  5511. \end_inset
  5512. </cell>
  5513. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  5514. \begin_inset Text
  5515. \begin_layout Plain Layout
  5516. 22
  5517. \end_layout
  5518. \end_inset
  5519. </cell>
  5520. </row>
  5521. <row>
  5522. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  5523. \begin_inset Text
  5524. \begin_layout Plain Layout
  5525. TX vs ADNR
  5526. \end_layout
  5527. \end_inset
  5528. </cell>
  5529. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  5530. \begin_inset Text
  5531. \begin_layout Plain Layout
  5532. 7
  5533. \end_layout
  5534. \end_inset
  5535. </cell>
  5536. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  5537. \begin_inset Text
  5538. \begin_layout Plain Layout
  5539. 338
  5540. \end_layout
  5541. \end_inset
  5542. </cell>
  5543. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  5544. \begin_inset Text
  5545. \begin_layout Plain Layout
  5546. 369
  5547. \end_layout
  5548. \end_inset
  5549. </cell>
  5550. </row>
  5551. <row>
  5552. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  5553. \begin_inset Text
  5554. \begin_layout Plain Layout
  5555. TX vs CAN
  5556. \end_layout
  5557. \end_inset
  5558. </cell>
  5559. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  5560. \begin_inset Text
  5561. \begin_layout Plain Layout
  5562. 0
  5563. \end_layout
  5564. \end_inset
  5565. </cell>
  5566. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  5567. \begin_inset Text
  5568. \begin_layout Plain Layout
  5569. 231
  5570. \end_layout
  5571. \end_inset
  5572. </cell>
  5573. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  5574. \begin_inset Text
  5575. \begin_layout Plain Layout
  5576. 278
  5577. \end_layout
  5578. \end_inset
  5579. </cell>
  5580. </row>
  5581. </lyxtabular>
  5582. \end_inset
  5583. \end_layout
  5584. \begin_layout Plain Layout
  5585. \begin_inset Caption Standard
  5586. \begin_layout Plain Layout
  5587. \begin_inset CommandInset label
  5588. LatexCommand label
  5589. name "tab:methyl-num-signif"
  5590. \end_inset
  5591. \series bold
  5592. Number of probes significant at 10% FDR for each contrast in each analysis.
  5593. \series default
  5594. For each of the analyses in Table
  5595. \begin_inset CommandInset ref
  5596. LatexCommand ref
  5597. reference "tab:Summary-of-meth-analysis"
  5598. plural "false"
  5599. caps "false"
  5600. noprefix "false"
  5601. \end_inset
  5602. , the table shows the number of probes called significantly differentially
  5603. methylated at a threshold of 10% FDR for each comparison between TX and
  5604. the other 3 transplant statuses.
  5605. \end_layout
  5606. \end_inset
  5607. \end_layout
  5608. \end_inset
  5609. \end_layout
  5610. \begin_layout Standard
  5611. \begin_inset Float table
  5612. wide false
  5613. sideways false
  5614. status collapsed
  5615. \begin_layout Plain Layout
  5616. \align center
  5617. \begin_inset Tabular
  5618. <lyxtabular version="3" rows="5" columns="4">
  5619. <features tabularvalignment="middle">
  5620. <column alignment="center" valignment="top">
  5621. <column alignment="center" valignment="top">
  5622. <column alignment="center" valignment="top">
  5623. <column alignment="center" valignment="top">
  5624. <row>
  5625. <cell alignment="center" valignment="top" usebox="none">
  5626. \begin_inset Text
  5627. \begin_layout Plain Layout
  5628. \end_layout
  5629. \end_inset
  5630. </cell>
  5631. <cell multicolumn="1" alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  5632. \begin_inset Text
  5633. \begin_layout Plain Layout
  5634. Analysis
  5635. \end_layout
  5636. \end_inset
  5637. </cell>
  5638. <cell multicolumn="2" alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  5639. \begin_inset Text
  5640. \begin_layout Plain Layout
  5641. \end_layout
  5642. \end_inset
  5643. </cell>
  5644. <cell multicolumn="2" alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  5645. \begin_inset Text
  5646. \begin_layout Plain Layout
  5647. \end_layout
  5648. \end_inset
  5649. </cell>
  5650. </row>
  5651. <row>
  5652. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  5653. \begin_inset Text
  5654. \begin_layout Plain Layout
  5655. Contrast
  5656. \end_layout
  5657. \end_inset
  5658. </cell>
  5659. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  5660. \begin_inset Text
  5661. \begin_layout Plain Layout
  5662. A
  5663. \end_layout
  5664. \end_inset
  5665. </cell>
  5666. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  5667. \begin_inset Text
  5668. \begin_layout Plain Layout
  5669. B
  5670. \end_layout
  5671. \end_inset
  5672. </cell>
  5673. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  5674. \begin_inset Text
  5675. \begin_layout Plain Layout
  5676. C
  5677. \end_layout
  5678. \end_inset
  5679. </cell>
  5680. </row>
  5681. <row>
  5682. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  5683. \begin_inset Text
  5684. \begin_layout Plain Layout
  5685. TX vs AR
  5686. \end_layout
  5687. \end_inset
  5688. </cell>
  5689. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  5690. \begin_inset Text
  5691. \begin_layout Plain Layout
  5692. 0
  5693. \end_layout
  5694. \end_inset
  5695. </cell>
  5696. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  5697. \begin_inset Text
  5698. \begin_layout Plain Layout
  5699. 10,063
  5700. \end_layout
  5701. \end_inset
  5702. </cell>
  5703. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  5704. \begin_inset Text
  5705. \begin_layout Plain Layout
  5706. 11,225
  5707. \end_layout
  5708. \end_inset
  5709. </cell>
  5710. </row>
  5711. <row>
  5712. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  5713. \begin_inset Text
  5714. \begin_layout Plain Layout
  5715. TX vs ADNR
  5716. \end_layout
  5717. \end_inset
  5718. </cell>
  5719. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  5720. \begin_inset Text
  5721. \begin_layout Plain Layout
  5722. 27
  5723. \end_layout
  5724. \end_inset
  5725. </cell>
  5726. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  5727. \begin_inset Text
  5728. \begin_layout Plain Layout
  5729. 12,674
  5730. \end_layout
  5731. \end_inset
  5732. </cell>
  5733. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  5734. \begin_inset Text
  5735. \begin_layout Plain Layout
  5736. 13,086
  5737. \end_layout
  5738. \end_inset
  5739. </cell>
  5740. </row>
  5741. <row>
  5742. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  5743. \begin_inset Text
  5744. \begin_layout Plain Layout
  5745. TX vs CAN
  5746. \end_layout
  5747. \end_inset
  5748. </cell>
  5749. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  5750. \begin_inset Text
  5751. \begin_layout Plain Layout
  5752. 966
  5753. \end_layout
  5754. \end_inset
  5755. </cell>
  5756. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  5757. \begin_inset Text
  5758. \begin_layout Plain Layout
  5759. 20,039
  5760. \end_layout
  5761. \end_inset
  5762. </cell>
  5763. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  5764. \begin_inset Text
  5765. \begin_layout Plain Layout
  5766. 20,955
  5767. \end_layout
  5768. \end_inset
  5769. </cell>
  5770. </row>
  5771. </lyxtabular>
  5772. \end_inset
  5773. \end_layout
  5774. \begin_layout Plain Layout
  5775. \begin_inset Caption Standard
  5776. \begin_layout Plain Layout
  5777. \begin_inset CommandInset label
  5778. LatexCommand label
  5779. name "tab:methyl-est-nonnull"
  5780. \end_inset
  5781. \series bold
  5782. Estimated number of non-null tests for each contrast in each analysis.
  5783. \series default
  5784. For each of the analyses in Table
  5785. \begin_inset CommandInset ref
  5786. LatexCommand ref
  5787. reference "tab:Summary-of-meth-analysis"
  5788. plural "false"
  5789. caps "false"
  5790. noprefix "false"
  5791. \end_inset
  5792. , the table shows the number of probes estimated to be differentially methylated
  5793. between TX and the other 3 transplant statuses using the method of
  5794. \begin_inset CommandInset citation
  5795. LatexCommand cite
  5796. key "Phipson2013"
  5797. literal "false"
  5798. \end_inset
  5799. .
  5800. \end_layout
  5801. \end_inset
  5802. \end_layout
  5803. \end_inset
  5804. \end_layout
  5805. \begin_layout Standard
  5806. \begin_inset Float figure
  5807. wide false
  5808. sideways false
  5809. status collapsed
  5810. \begin_layout Plain Layout
  5811. \begin_inset Flex TODO Note (inline)
  5812. status open
  5813. \begin_layout Plain Layout
  5814. Re-generate p-value histograms for all relevant contrasts in a single page,
  5815. then write an appropriate legend.
  5816. \end_layout
  5817. \end_inset
  5818. \end_layout
  5819. \begin_layout Plain Layout
  5820. \align center
  5821. \series bold
  5822. [Figure goes here]
  5823. \end_layout
  5824. \begin_layout Plain Layout
  5825. \begin_inset Caption Standard
  5826. \begin_layout Plain Layout
  5827. \series bold
  5828. \begin_inset CommandInset label
  5829. LatexCommand label
  5830. name "fig:meth-p-value-histograms"
  5831. \end_inset
  5832. Probe p-value histograms for each contrast in each analysis.
  5833. \end_layout
  5834. \end_inset
  5835. \end_layout
  5836. \begin_layout Plain Layout
  5837. \end_layout
  5838. \end_inset
  5839. \end_layout
  5840. \begin_layout Standard
  5841. Table
  5842. \begin_inset CommandInset ref
  5843. LatexCommand ref
  5844. reference "tab:methyl-num-signif"
  5845. plural "false"
  5846. caps "false"
  5847. noprefix "false"
  5848. \end_inset
  5849. shows the number of significantly differentially methylated probes reported
  5850. by each analysis for each comparison of interest at an FDR of 10%.
  5851. As expected, the more elaborate analyses, B and C, report more significant
  5852. probes than the more basic analysis A, consistent with the conclusions
  5853. above that the data contain hidden systematic variations that must be modeled.
  5854. Table
  5855. \begin_inset CommandInset ref
  5856. LatexCommand ref
  5857. reference "tab:methyl-est-nonnull"
  5858. plural "false"
  5859. caps "false"
  5860. noprefix "false"
  5861. \end_inset
  5862. shows the estimated number differentially methylated probes for each test
  5863. from each analysis.
  5864. This was computed by estimating the proportion of null hypotheses that
  5865. were true using the method of
  5866. \begin_inset CommandInset citation
  5867. LatexCommand cite
  5868. key "Phipson2013"
  5869. literal "false"
  5870. \end_inset
  5871. and subtracting that fraction from the total number of probes, yielding
  5872. an estimate of the number of null hypotheses that are false based on the
  5873. distribution of p-values across the entire dataset.
  5874. Note that this does not identify which null hypotheses should be rejected
  5875. (i.e.
  5876. which probes are significant); it only estimates the true number of such
  5877. probes.
  5878. Once again, analyses B and C result it much larger estimates for the number
  5879. of differentially methylated probes.
  5880. In this case, analysis C, the only analysis that includes voom, estimates
  5881. the largest number of differentially methylated probes for all 3 contrasts.
  5882. If the assumptions of all the methods employed hold, then this represents
  5883. a gain in statistical power over the simpler analysis A.
  5884. Figure
  5885. \begin_inset CommandInset ref
  5886. LatexCommand ref
  5887. reference "fig:meth-p-value-histograms"
  5888. plural "false"
  5889. caps "false"
  5890. noprefix "false"
  5891. \end_inset
  5892. shows the p-value distributions for each test, from which the numbers in
  5893. Table
  5894. \begin_inset CommandInset ref
  5895. LatexCommand ref
  5896. reference "tab:methyl-est-nonnull"
  5897. plural "false"
  5898. caps "false"
  5899. noprefix "false"
  5900. \end_inset
  5901. were generated.
  5902. The distributions for analysis A all have a dip in density near zero, which
  5903. is a strong sign of a poor model fit.
  5904. The histograms for analyses B and C are more well-behaved, with a uniform
  5905. component stretching all the way from 0 to 1 representing the probes for
  5906. which the null hypotheses is true (no differential methylation), and a
  5907. zero-biased component representing the probes for which the null hypothesis
  5908. is false (differentially methylated).
  5909. These histograms do not indicate any major issues with the model fit.
  5910. \end_layout
  5911. \begin_layout Standard
  5912. \begin_inset Flex TODO Note (inline)
  5913. status open
  5914. \begin_layout Plain Layout
  5915. Maybe include the PCA plots before/after SVA effect subtraction?
  5916. \end_layout
  5917. \end_inset
  5918. \end_layout
  5919. \begin_layout Section
  5920. Discussion
  5921. \end_layout
  5922. \begin_layout Subsection
  5923. fRMA achieves clinically applicable normalization without sacrificing classifica
  5924. tion performance
  5925. \end_layout
  5926. \begin_layout Standard
  5927. As shown in Figure
  5928. \begin_inset CommandInset ref
  5929. LatexCommand ref
  5930. reference "fig:Classifier-probabilities-RMA"
  5931. plural "false"
  5932. caps "false"
  5933. noprefix "false"
  5934. \end_inset
  5935. , improper normalization, particularly separate normalization of training
  5936. and test samples, leads to unwanted biases in classification.
  5937. In a controlled experimental context, it is always possible to correct
  5938. this issue by normalizing all experimental samples together.
  5939. However, because it is not feasible to normalize all samples together in
  5940. a clinical context, a single-channel normalization is required is required.
  5941. \end_layout
  5942. \begin_layout Standard
  5943. The major concern in using a single-channel normalization is that non-single-cha
  5944. nnel methods can share information between arrays to improve the normalization,
  5945. and single-channel methods risk sacrificing the gains in normalization
  5946. accuracy that come from this information sharing.
  5947. In the case of RMA, this information sharing is accomplished through quantile
  5948. normalization and median polish steps.
  5949. The need for information sharing in quantile normalization can easily be
  5950. removed by learning a fixed set of quantiles from external data and normalizing
  5951. each array to these fixed quantiles, instead of the quantiles of the data
  5952. itself.
  5953. As long as the fixed quantiles are reasonable, the result will be similar
  5954. to standard RMA.
  5955. However, there is no analogous way to eliminate cross-array information
  5956. sharing in the median polish step, so fRMA replaces this with a weighted
  5957. average of probes on each array, with the weights learned from external
  5958. data.
  5959. This step of fRMA has the greatest potential to diverge from RMA un undesirable
  5960. ways.
  5961. \end_layout
  5962. \begin_layout Standard
  5963. However, when run on real data, fRMA performed at least as well as RMA in
  5964. both the internal validation and external validation tests.
  5965. This shows that fRMA can be used to normalize individual clinical samples
  5966. in a class prediction context without sacrificing the classifier performance
  5967. that would be obtained by using the more well-established RMA for normalization.
  5968. The other single-channel normalization method considered, SCAN, showed
  5969. some loss of AUC in the external validation test.
  5970. Based on these results, fRMA is the preferred normalization for clinical
  5971. samples in a class prediction context.
  5972. \end_layout
  5973. \begin_layout Subsection
  5974. Robust fRMA vectors can be generated for new array platforms
  5975. \end_layout
  5976. \begin_layout Standard
  5977. \begin_inset Flex TODO Note (inline)
  5978. status open
  5979. \begin_layout Plain Layout
  5980. Look up the exact numbers, do a find & replace for
  5981. \begin_inset Quotes eld
  5982. \end_inset
  5983. 850
  5984. \begin_inset Quotes erd
  5985. \end_inset
  5986. \end_layout
  5987. \end_inset
  5988. \end_layout
  5989. \begin_layout Standard
  5990. The published fRMA normalization vectors for the hgu133plus2 platform were
  5991. generated from a set of about 850 samples chosen from a wide range of tissues,
  5992. which the authors determined was sufficient to generate a robust set of
  5993. normalization vectors that could be applied across all tissues
  5994. \begin_inset CommandInset citation
  5995. LatexCommand cite
  5996. key "McCall2010"
  5997. literal "false"
  5998. \end_inset
  5999. .
  6000. Since we only had hthgu133pluspm for 2 tissues of interest, our needs were
  6001. more modest.
  6002. Even using only 130 samples in 26 batches of 5 samples each for kidney
  6003. biopsies, we were able to train a robust set of fRMA normalization vectors
  6004. that were not meaningfully affected by the random selection of 5 samples
  6005. from each batch.
  6006. As expected, the training process was just as robust for the blood samples
  6007. with 230 samples in 46 batches of 5 samples each.
  6008. Because these vectors were each generated using training samples from a
  6009. single tissue, they are not suitable for general use, unlike the vectors
  6010. provided with fRMA itself.
  6011. They are purpose-built for normalizing a specific type of sample on a specific
  6012. platform.
  6013. This is a mostly acceptable limitation in the context of developing a machine
  6014. learning classifier for diagnosing a disease based on samples of a specific
  6015. tissue.
  6016. \end_layout
  6017. \begin_layout Standard
  6018. \begin_inset Flex TODO Note (inline)
  6019. status open
  6020. \begin_layout Plain Layout
  6021. How to bring up that these custom vectors were used in another project by
  6022. someone else that was never published?
  6023. \end_layout
  6024. \end_inset
  6025. \end_layout
  6026. \begin_layout Subsection
  6027. Methylation array data can be successfully analyzed using existing techniques,
  6028. but machine learning poses additional challenges
  6029. \end_layout
  6030. \begin_layout Standard
  6031. Both analysis strategies B and C both yield a reasonable analysis, with
  6032. a mean-variance trend that matches the expected behavior for the non-linear
  6033. M-value transformation (Figure
  6034. \begin_inset CommandInset ref
  6035. LatexCommand ref
  6036. reference "fig:meanvar-sva-aw"
  6037. plural "false"
  6038. caps "false"
  6039. noprefix "false"
  6040. \end_inset
  6041. ) and well-behaved p-value distributions (Figure
  6042. \begin_inset CommandInset ref
  6043. LatexCommand ref
  6044. reference "fig:meth-p-value-histograms"
  6045. plural "false"
  6046. caps "false"
  6047. noprefix "false"
  6048. \end_inset
  6049. ).
  6050. These two analyses also yield similar numbers of significant probes (Table
  6051. \begin_inset CommandInset ref
  6052. LatexCommand ref
  6053. reference "tab:methyl-num-signif"
  6054. plural "false"
  6055. caps "false"
  6056. noprefix "false"
  6057. \end_inset
  6058. ) and similar estimates of the number of differentially methylated probes
  6059. (Table
  6060. \begin_inset CommandInset ref
  6061. LatexCommand ref
  6062. reference "tab:methyl-est-nonnull"
  6063. plural "false"
  6064. caps "false"
  6065. noprefix "false"
  6066. \end_inset
  6067. ).
  6068. The main difference between these two analyses is the method used to account
  6069. for the mean-variance trend.
  6070. In analysis B, the trend is estimated and applied at the probe level: each
  6071. probe's estimated variance is squeezed toward the trend using an empirical
  6072. Bayes procedure (Figure
  6073. \begin_inset CommandInset ref
  6074. LatexCommand ref
  6075. reference "fig:meanvar-sva-aw"
  6076. plural "false"
  6077. caps "false"
  6078. noprefix "false"
  6079. \end_inset
  6080. ).
  6081. In analysis C, the trend is still estimated at the probe level, but instead
  6082. of estimating a single variance value shared across all observations for
  6083. a given probe, the voom method computes an initial estiamte of the variance
  6084. for each observation individually based on where its model-fitted M-value
  6085. falls on the trend line and then assigns inverse-variance weights to model
  6086. the difference in variance between observations.
  6087. An overall variance is still estimated for each probe using the same empirical
  6088. Bayes method, but now the residual trend is flat (Figure
  6089. \begin_inset CommandInset ref
  6090. LatexCommand ref
  6091. reference "fig:meanvar-sva-voomaw"
  6092. plural "false"
  6093. caps "false"
  6094. noprefix "false"
  6095. \end_inset
  6096. ), and the mean-variance trend is modeled by scaling the probe's estimated
  6097. variance for each observation using the weights computed by voom.
  6098. The difference between these two methods is analogous to the difference
  6099. between a t-test with equal variance and a t-test with unequal variance,
  6100. except that the unequal group variances used in the latter test are estimated
  6101. based on the mean-variance trend from all the probes rather than the data
  6102. for the specific probe being tested, thus stabilizing the group variance
  6103. estimates by sharing information between probes.
  6104. In practice, allowing voom to model the variance using observation weights
  6105. in this manner allows the linear model fit to concentrate statistical power
  6106. where it will do the most good.
  6107. For example, if a particular probe's M-values are always at the extreme
  6108. of the M-value range (e.g.
  6109. less than -4) for ADNR samples, but the M-values for that probe in TX and
  6110. CAN samples are within the flat region of the mean-variance trend (between
  6111. -3 and +3), voom is able to down-weight the contribution of the high-variance
  6112. M-values from the ADNR samples in order to gain more statistical power
  6113. while testing for differential methylation between TX and CAN.
  6114. In contrast, modeling the mean-variance trend only at the probe level would
  6115. combine the high-variance ADNR samples and lower-variance samples from
  6116. other conditions and estimate an intermediate variance for this probe.
  6117. In practice, analysis B shows that this approach is adequate, but the voom
  6118. approach in analysis C is at least as good on all model fit criteria and
  6119. yields a larger estimate for the number of differentially methylated genes.
  6120. \end_layout
  6121. \begin_layout Standard
  6122. The significant association of diebetes diagnosis with sample quality is
  6123. interesting.
  6124. The samples with Type 2 diabetes tended to have more variation, averaged
  6125. across all probes, than those with Type 1 diabetes.
  6126. This is consistent with the consensus that type 2 disbetes and the associated
  6127. metabolic syndrome represent a broad dysregulation of the body's endocrine
  6128. signalling related to metabolism [citation needed].
  6129. This dysregulation could easily manifest as a greater degree of variation
  6130. in the DNA methylation patterns of affected tissues.
  6131. In contrast, Type 1 disbetes has a more specific cause and effect, so a
  6132. less variable methylation signature is expected.
  6133. \end_layout
  6134. \begin_layout Standard
  6135. This preliminary anlaysis suggests that some degree of differential methylation
  6136. exists between TX and each of the three types of transplant disfunction
  6137. studied.
  6138. Hence, it may be feasible to train a classifier to diagnose transplant
  6139. disfunction from DNA methylation array data.
  6140. However, the major importance of both SVA and sample quality weighting
  6141. for proper modeling of this data poses significant challenges for any attempt
  6142. at a machine learning on data of similar quality.
  6143. While these are easily used in a modeling context with full sample information,
  6144. neither of these methods is directly applicable in a machine learning context,
  6145. where the diagnosis is not known ahead of time.
  6146. If a machine learning approach for methylation-based diagnosis is to be
  6147. pursued, it will either require machine-learning-friendly methods to address
  6148. the same systematic trends in the data that SVA and sample quality weighting
  6149. address, or it will require higher quality data with substantially less
  6150. systematic perturbation of the data.
  6151. \end_layout
  6152. \begin_layout Chapter
  6153. Globin-blocking for more effective blood RNA-seq analysis in primate animal
  6154. model
  6155. \end_layout
  6156. \begin_layout Standard
  6157. \begin_inset Flex TODO Note (inline)
  6158. status open
  6159. \begin_layout Plain Layout
  6160. Choose between above and the paper title: Optimizing yield of deep RNA sequencin
  6161. g for gene expression profiling by globin reduction of peripheral blood
  6162. samples from cynomolgus monkeys (Macaca fascicularis).
  6163. \end_layout
  6164. \end_inset
  6165. \end_layout
  6166. \begin_layout Standard
  6167. \begin_inset Flex TODO Note (inline)
  6168. status open
  6169. \begin_layout Plain Layout
  6170. Chapter author list: https://tex.stackexchange.com/questions/156862/displaying-aut
  6171. hor-for-each-chapter-in-book Every chapter gets an author list, which may
  6172. or may not be part of a citation to a published/preprinted paper.
  6173. \end_layout
  6174. \end_inset
  6175. \end_layout
  6176. \begin_layout Standard
  6177. \begin_inset Flex TODO Note (inline)
  6178. status open
  6179. \begin_layout Plain Layout
  6180. Preprint then cite the paper
  6181. \end_layout
  6182. \end_inset
  6183. \end_layout
  6184. \begin_layout Section*
  6185. Abstract
  6186. \end_layout
  6187. \begin_layout Paragraph
  6188. Background
  6189. \end_layout
  6190. \begin_layout Standard
  6191. Primate blood contains high concentrations of globin messenger RNA.
  6192. Globin reduction is a standard technique used to improve the expression
  6193. results obtained by DNA microarrays on RNA from blood samples.
  6194. However, with whole transcriptome RNA-sequencing (RNA-seq) quickly replacing
  6195. microarrays for many applications, the impact of globin reduction for RNA-seq
  6196. has not been previously studied.
  6197. Moreover, no off-the-shelf kits are available for globin reduction in nonhuman
  6198. primates.
  6199. \end_layout
  6200. \begin_layout Paragraph
  6201. Results
  6202. \end_layout
  6203. \begin_layout Standard
  6204. Here we report a protocol for RNA-seq in primate blood samples that uses
  6205. complimentary oligonucleotides to block reverse transcription of the alpha
  6206. and beta globin genes.
  6207. In test samples from cynomolgus monkeys (Macaca fascicularis), this globin
  6208. blocking protocol approximately doubles the yield of informative (non-globin)
  6209. reads by greatly reducing the fraction of globin reads, while also improving
  6210. the consistency in sequencing depth between samples.
  6211. The increased yield enables detection of about 2000 more genes, significantly
  6212. increases the correlation in measured gene expression levels between samples,
  6213. and increases the sensitivity of differential gene expression tests.
  6214. \end_layout
  6215. \begin_layout Paragraph
  6216. Conclusions
  6217. \end_layout
  6218. \begin_layout Standard
  6219. These results show that globin blocking significantly improves the cost-effectiv
  6220. eness of mRNA sequencing in primate blood samples by doubling the yield
  6221. of useful reads, allowing detection of more genes, and improving the precision
  6222. of gene expression measurements.
  6223. Based on these results, a globin reducing or blocking protocol is recommended
  6224. for all RNA-seq studies of primate blood samples.
  6225. \end_layout
  6226. \begin_layout Section
  6227. Approach
  6228. \end_layout
  6229. \begin_layout Standard
  6230. \begin_inset Note Note
  6231. status open
  6232. \begin_layout Plain Layout
  6233. Consider putting some of this in the Intro chapter
  6234. \end_layout
  6235. \begin_layout Itemize
  6236. Cynomolgus monkeys as a model organism
  6237. \end_layout
  6238. \begin_deeper
  6239. \begin_layout Itemize
  6240. Highly related to humans
  6241. \end_layout
  6242. \begin_layout Itemize
  6243. Small size and short life cycle - good research animal
  6244. \end_layout
  6245. \begin_layout Itemize
  6246. Genomics resources still in development
  6247. \end_layout
  6248. \end_deeper
  6249. \begin_layout Itemize
  6250. Inadequacy of existing blood RNA-seq protocols
  6251. \end_layout
  6252. \begin_deeper
  6253. \begin_layout Itemize
  6254. Existing protocols use a separate globin pulldown step, slowing down processing
  6255. \end_layout
  6256. \end_deeper
  6257. \end_inset
  6258. \end_layout
  6259. \begin_layout Standard
  6260. Increasingly, researchers are turning to high-throughput mRNA sequencing
  6261. technologies (RNA-seq) in preference to expression microarrays for analysis
  6262. of gene expression
  6263. \begin_inset CommandInset citation
  6264. LatexCommand cite
  6265. key "Mutz2012"
  6266. literal "false"
  6267. \end_inset
  6268. .
  6269. The advantages are even greater for study of model organisms with no well-estab
  6270. lished array platforms available, such as the cynomolgus monkey (Macaca
  6271. fascicularis).
  6272. High fractions of globin mRNA are naturally present in mammalian peripheral
  6273. blood samples (up to 70% of total mRNA) and these are known to interfere
  6274. with the results of array-based expression profiling
  6275. \begin_inset CommandInset citation
  6276. LatexCommand cite
  6277. key "Winn2010"
  6278. literal "false"
  6279. \end_inset
  6280. .
  6281. The importance of globin reduction for RNA-seq of blood has only been evaluated
  6282. for a deepSAGE protocol on human samples
  6283. \begin_inset CommandInset citation
  6284. LatexCommand cite
  6285. key "Mastrokolias2012"
  6286. literal "false"
  6287. \end_inset
  6288. .
  6289. In the present report, we evaluated globin reduction using custom blocking
  6290. oligonucleotides for deep RNA-seq of peripheral blood samples from a nonhuman
  6291. primate, cynomolgus monkey, using the Illumina technology platform.
  6292. We demonstrate that globin reduction significantly improves the cost-effectiven
  6293. ess of RNA-seq in blood samples.
  6294. Thus, our protocol offers a significant advantage to any investigator planning
  6295. to use RNA-seq for gene expression profiling of nonhuman primate blood
  6296. samples.
  6297. Our method can be generally applied to any species by designing complementary
  6298. oligonucleotide blocking probes to the globin gene sequences of that species.
  6299. Indeed, any highly expressed but biologically uninformative transcripts
  6300. can also be blocked to further increase sequencing efficiency and value
  6301. \begin_inset CommandInset citation
  6302. LatexCommand cite
  6303. key "Arnaud2016"
  6304. literal "false"
  6305. \end_inset
  6306. .
  6307. \end_layout
  6308. \begin_layout Section
  6309. Methods
  6310. \end_layout
  6311. \begin_layout Subsection
  6312. Sample collection
  6313. \end_layout
  6314. \begin_layout Standard
  6315. All research reported here was done under IACUC-approved protocols at the
  6316. University of Miami and complied with all applicable federal and state
  6317. regulations and ethical principles for nonhuman primate research.
  6318. Blood draws occurred between 16 April 2012 and 18 June 2015.
  6319. The experimental system involved intrahepatic pancreatic islet transplantation
  6320. into Cynomolgus monkeys with induced diabetes mellitus with or without
  6321. concomitant infusion of mesenchymal stem cells.
  6322. Blood was collected at serial time points before and after transplantation
  6323. into PAXgene Blood RNA tubes (PreAnalytiX/Qiagen, Valencia, CA) at the
  6324. precise volume:volume ratio of 2.5 ml whole blood into 6.9 ml of PAX gene
  6325. additive.
  6326. \end_layout
  6327. \begin_layout Subsection
  6328. Globin Blocking
  6329. \end_layout
  6330. \begin_layout Standard
  6331. Four oligonucleotides were designed to hybridize to the 3’ end of the transcript
  6332. s for Cynomolgus HBA1, HBA2 and HBB, with two hybridization sites for HBB
  6333. and 2 sites for HBA (the chosen sites were identical in both HBA genes).
  6334. All oligos were purchased from Sigma and were entirely composed of 2’O-Me
  6335. bases with a C3 spacer positioned at the 3’ ends to prevent any polymerase
  6336. mediated primer extension.
  6337. \end_layout
  6338. \begin_layout Quote
  6339. HBA1/2 site 1: GCCCACUCAGACUUUAUUCAAAG-C3spacer
  6340. \end_layout
  6341. \begin_layout Quote
  6342. HBA1/2 site 2: GGUGCAAGGAGGGGAGGAG-C3spacer
  6343. \end_layout
  6344. \begin_layout Quote
  6345. HBB site 1: AAUGAAAAUAAAUGUUUUUUAUUAG-C3spacer
  6346. \end_layout
  6347. \begin_layout Quote
  6348. HBB site 2: CUCAAGGCCCUUCAUAAUAUCCC-C3spacer
  6349. \end_layout
  6350. \begin_layout Subsection
  6351. RNA-seq Library Preparation
  6352. \end_layout
  6353. \begin_layout Standard
  6354. Sequencing libraries were prepared with 200ng total RNA from each sample.
  6355. Polyadenylated mRNA was selected from 200 ng aliquots of cynomologus blood-deri
  6356. ved total RNA using Ambion Dynabeads Oligo(dT)25 beads (Invitrogen) following
  6357. manufacturer’s recommended protocol.
  6358. PolyA selected RNA was then combined with 8 pmol of HBA1/2 (site 1), 8
  6359. pmol of HBA1/2 (site 2), 12 pmol of HBB (site 1) and 12 pmol of HBB (site
  6360. 2) oligonucleotides.
  6361. In addition, 20 pmol of RT primer containing a portion of the Illumina
  6362. adapter sequence (B-oligo-dTV: GAGTTCCTTGGCACCCGAGAATTCCATTTTTTTTTTTTTTTTTTTV)
  6363. and 4 µL of 5X First Strand buffer (250 mM Tris-HCl pH 8.3, 375 mM KCl,
  6364. 15mM MgCl2) were added in a total volume of 15 µL.
  6365. The RNA was fragmented by heating this cocktail for 3 minutes at 95°C and
  6366. then placed on ice.
  6367. This was followed by the addition of 2 µL 0.1 M DTT, 1 µL RNaseOUT, 1 µL
  6368. 10mM dNTPs 10% biotin-16 aminoallyl-2’- dUTP and 10% biotin-16 aminoallyl-2’-
  6369. dCTP (TriLink Biotech, San Diego, CA), 1 µL Superscript II (200U/ µL, Thermo-Fi
  6370. sher).
  6371. A second “unblocked” library was prepared in the same way for each sample
  6372. but replacing the blocking oligos with an equivalent volume of water.
  6373. The reaction was carried out at 25°C for 15 minutes and 42°C for 40 minutes,
  6374. followed by incubation at 75°C for 10 minutes to inactivate the reverse
  6375. transcriptase.
  6376. \end_layout
  6377. \begin_layout Standard
  6378. The cDNA/RNA hybrid molecules were purified using 1.8X Ampure XP beads (Agencourt
  6379. ) following supplier’s recommended protocol.
  6380. The cDNA/RNA hybrid was eluted in 25 µL of 10 mM Tris-HCl pH 8.0, and then
  6381. bound to 25 µL of M280 Magnetic Streptavidin beads washed per recommended
  6382. protocol (Thermo-Fisher).
  6383. After 30 minutes of binding, beads were washed one time in 100 µL 0.1N NaOH
  6384. to denature and remove the bound RNA, followed by two 100 µL washes with
  6385. 1X TE buffer.
  6386. \end_layout
  6387. \begin_layout Standard
  6388. Subsequent attachment of the 5-prime Illumina A adapter was performed by
  6389. on-bead random primer extension of the following sequence (A-N8 primer:
  6390. TTCAGAGTTCTACAGTCCGACGATCNNNNNNNN).
  6391. Briefly, beads were resuspended in a 20 µL reaction containing 5 µM A-N8
  6392. primer, 40mM Tris-HCl pH 7.5, 20mM MgCl2, 50mM NaCl, 0.325U/µL Sequenase
  6393. 2.0 (Affymetrix, Santa Clara, CA), 0.0025U/µL inorganic pyrophosphatase (Affymetr
  6394. ix) and 300 µM each dNTP.
  6395. Reaction was incubated at 22°C for 30 minutes, then beads were washed 2
  6396. times with 1X TE buffer (200µL).
  6397. \end_layout
  6398. \begin_layout Standard
  6399. The magnetic streptavidin beads were resuspended in 34 µL nuclease-free
  6400. water and added directly to a PCR tube.
  6401. The two Illumina protocol-specified PCR primers were added at 0.53 µM (Illumina
  6402. TruSeq Universal Primer 1 and Illumina TruSeq barcoded PCR primer 2), along
  6403. with 40 µL 2X KAPA HiFi Hotstart ReadyMix (KAPA, Willmington MA) and thermocycl
  6404. ed as follows: starting with 98°C (2 min-hold); 15 cycles of 98°C, 20sec;
  6405. 60°C, 30sec; 72°C, 30sec; and finished with a 72°C (2 min-hold).
  6406. \end_layout
  6407. \begin_layout Standard
  6408. PCR products were purified with 1X Ampure Beads following manufacturer’s
  6409. recommended protocol.
  6410. Libraries were then analyzed using the Agilent TapeStation and quantitation
  6411. of desired size range was performed by “smear analysis”.
  6412. Samples were pooled in equimolar batches of 16 samples.
  6413. Pooled libraries were size selected on 2% agarose gels (E-Gel EX Agarose
  6414. Gels; Thermo-Fisher).
  6415. Products were cut between 250 and 350 bp (corresponding to insert sizes
  6416. of 130 to 230 bps).
  6417. Finished library pools were then sequenced on the Illumina NextSeq500 instrumen
  6418. t with 75 base read lengths.
  6419. \end_layout
  6420. \begin_layout Subsection
  6421. Read alignment and counting
  6422. \end_layout
  6423. \begin_layout Standard
  6424. Reads were aligned to the cynomolgus genome using STAR
  6425. \begin_inset CommandInset citation
  6426. LatexCommand cite
  6427. key "Dobin2013,Wilson2013"
  6428. literal "false"
  6429. \end_inset
  6430. .
  6431. Counts of uniquely mapped reads were obtained for every gene in each sample
  6432. with the “featureCounts” function from the Rsubread package, using each
  6433. of the three possibilities for the “strandSpecific” option: sense, antisense,
  6434. and unstranded
  6435. \begin_inset CommandInset citation
  6436. LatexCommand cite
  6437. key "Liao2014"
  6438. literal "false"
  6439. \end_inset
  6440. .
  6441. A few artifacts in the cynomolgus genome annotation complicated read counting.
  6442. First, no ortholog is annotated for alpha globin in the cynomolgus genome,
  6443. presumably because the human genome has two alpha globin genes with nearly
  6444. identical sequences, making the orthology relationship ambiguous.
  6445. However, two loci in the cynomolgus genome are as “hemoglobin subunit alpha-lik
  6446. e” (LOC102136192 and LOC102136846).
  6447. LOC102136192 is annotated as a pseudogene while LOC102136846 is annotated
  6448. as protein-coding.
  6449. Our globin reduction protocol was designed to include blocking of these
  6450. two genes.
  6451. Indeed, these two genes have almost the same read counts in each library
  6452. as the properly-annotated HBB gene and much larger counts than any other
  6453. gene in the unblocked libraries, giving confidence that reads derived from
  6454. the real alpha globin are mapping to both genes.
  6455. Thus, reads from both of these loci were counted as alpha globin reads
  6456. in all further analyses.
  6457. The second artifact is a small, uncharacterized non-coding RNA gene (LOC1021365
  6458. 91), which overlaps the HBA-like gene (LOC102136192) on the opposite strand.
  6459. If counting is not performed in stranded mode (or if a non-strand-specific
  6460. sequencing protocol is used), many reads mapping to the globin gene will
  6461. be discarded as ambiguous due to their overlap with this ncRNA gene, resulting
  6462. in significant undercounting of globin reads.
  6463. Therefore, stranded sense counts were used for all further analysis in
  6464. the present study to insure that we accurately accounted for globin transcript
  6465. reduction.
  6466. However, we note that stranded reads are not necessary for RNA-seq using
  6467. our protocol in standard practice.
  6468. \end_layout
  6469. \begin_layout Subsection
  6470. Normalization and Exploratory Data Analysis
  6471. \end_layout
  6472. \begin_layout Standard
  6473. Libraries were normalized by computing scaling factors using the edgeR package’s
  6474. Trimmed Mean of M-values method
  6475. \begin_inset CommandInset citation
  6476. LatexCommand cite
  6477. key "Robinson2010"
  6478. literal "false"
  6479. \end_inset
  6480. .
  6481. Log2 counts per million values (logCPM) were calculated using the cpm function
  6482. in edgeR for individual samples and aveLogCPM function for averages across
  6483. groups of samples, using those functions’ default prior count values to
  6484. avoid taking the logarithm of 0.
  6485. Genes were considered “present” if their average normalized logCPM values
  6486. across all libraries were at least -1.
  6487. Normalizing for gene length was unnecessary because the sequencing protocol
  6488. is 3’-biased and hence the expected read count for each gene is related
  6489. to the transcript’s copy number but not its length.
  6490. \end_layout
  6491. \begin_layout Standard
  6492. In order to assess the effect of blocking on reproducibility, Pearson and
  6493. Spearman correlation coefficients were computed between the logCPM values
  6494. for every pair of libraries within the globin-blocked (GB) and unblocked
  6495. (non-GB) groups, and edgeR's “estimateDisp” function was used to compute
  6496. negative binomial dispersions separately for the two groups
  6497. \begin_inset CommandInset citation
  6498. LatexCommand cite
  6499. key "Chen2014"
  6500. literal "false"
  6501. \end_inset
  6502. .
  6503. \end_layout
  6504. \begin_layout Subsection
  6505. Differential Expression Analysis
  6506. \end_layout
  6507. \begin_layout Standard
  6508. All tests for differential gene expression were performed using edgeR, by
  6509. first fitting a negative binomial generalized linear model to the counts
  6510. and normalization factors and then performing a quasi-likelihood F-test
  6511. with robust estimation of outlier gene dispersions
  6512. \begin_inset CommandInset citation
  6513. LatexCommand cite
  6514. key "Lund2012,Phipson2016"
  6515. literal "false"
  6516. \end_inset
  6517. .
  6518. To investigate the effects of globin blocking on each gene, an additive
  6519. model was fit to the full data with coefficients for globin blocking and
  6520. SampleID.
  6521. To test the effect of globin blocking on detection of differentially expressed
  6522. genes, the GB samples and non-GB samples were each analyzed independently
  6523. as follows: for each animal with both a pre-transplant and a post-transplant
  6524. time point in the data set, the pre-transplant sample and the earliest
  6525. post-transplant sample were selected, and all others were excluded, yielding
  6526. a pre-/post-transplant pair of samples for each animal (N=7 animals with
  6527. paired samples).
  6528. These samples were analyzed for pre-transplant vs.
  6529. post-transplant differential gene expression while controlling for inter-animal
  6530. variation using an additive model with coefficients for transplant and
  6531. animal ID.
  6532. In all analyses, p-values were adjusted using the Benjamini-Hochberg procedure
  6533. for FDR control
  6534. \begin_inset CommandInset citation
  6535. LatexCommand cite
  6536. key "Benjamini1995"
  6537. literal "false"
  6538. \end_inset
  6539. .
  6540. \end_layout
  6541. \begin_layout Standard
  6542. \begin_inset Note Note
  6543. status open
  6544. \begin_layout Itemize
  6545. New blood RNA-seq protocol to block reverse transcription of globin genes
  6546. \end_layout
  6547. \begin_layout Itemize
  6548. Blood RNA-seq time course after transplants with/without MSC infusion
  6549. \end_layout
  6550. \end_inset
  6551. \end_layout
  6552. \begin_layout Section
  6553. Results
  6554. \end_layout
  6555. \begin_layout Subsection
  6556. Globin blocking yields a larger and more consistent fraction of useful reads
  6557. \end_layout
  6558. \begin_layout Standard
  6559. The objective of the present study was to validate a new protocol for deep
  6560. RNA-seq of whole blood drawn into PaxGene tubes from cynomolgus monkeys
  6561. undergoing islet transplantation, with particular focus on minimizing the
  6562. loss of useful sequencing space to uninformative globin reads.
  6563. The details of the analysis with respect to transplant outcomes and the
  6564. impact of mesenchymal stem cell treatment will be reported in a separate
  6565. manuscript (in preparation).
  6566. To focus on the efficacy of our globin blocking protocol, 37 blood samples,
  6567. 16 from pre-transplant and 21 from post-transplant time points, were each
  6568. prepped once with and once without globin blocking oligos, and were then
  6569. sequenced on an Illumina NextSeq500 instrument.
  6570. The number of reads aligning to each gene in the cynomolgus genome was
  6571. counted.
  6572. Table 1 summarizes the distribution of read fractions among the GB and
  6573. non-GB libraries.
  6574. In the libraries with no globin blocking, globin reads made up an average
  6575. of 44.6% of total input reads, while reads assigned to all other genes made
  6576. up an average of 26.3%.
  6577. The remaining reads either aligned to intergenic regions (that include
  6578. long non-coding RNAs) or did not align with any annotated transcripts in
  6579. the current build of the cynomolgus genome.
  6580. In the GB libraries, globin reads made up only 3.48% and reads assigned
  6581. to all other genes increased to 50.4%.
  6582. Thus, globin blocking resulted in a 92.2% reduction in globin reads and
  6583. a 91.6% increase in yield of useful non-globin reads.
  6584. \end_layout
  6585. \begin_layout Standard
  6586. This reduction is not quite as efficient as the previous analysis showed
  6587. for human samples by DeepSAGE (<0.4% globin reads after globin reduction)
  6588. \begin_inset CommandInset citation
  6589. LatexCommand cite
  6590. key "Mastrokolias2012"
  6591. literal "false"
  6592. \end_inset
  6593. .
  6594. Nonetheless, this degree of globin reduction is sufficient to nearly double
  6595. the yield of useful reads.
  6596. Thus, globin blocking cuts the required sequencing effort (and costs) to
  6597. achieve a target coverage depth by almost 50%.
  6598. Consistent with this near doubling of yield, the average difference in
  6599. un-normalized logCPM across all genes between the GB libraries and non-GB
  6600. libraries is approximately 1 (mean = 1.01, median = 1.08), an overall 2-fold
  6601. increase.
  6602. Un-normalized values are used here because the TMM normalization correctly
  6603. identifies this 2-fold difference as biologically irrelevant and removes
  6604. it.
  6605. \end_layout
  6606. \begin_layout Standard
  6607. \begin_inset Float figure
  6608. wide false
  6609. sideways false
  6610. status collapsed
  6611. \begin_layout Plain Layout
  6612. \align center
  6613. \begin_inset Graphics
  6614. filename graphics/Globin Paper/figure1 - globin-fractions.pdf
  6615. lyxscale 50
  6616. width 100col%
  6617. groupId colwidth
  6618. \end_inset
  6619. \end_layout
  6620. \begin_layout Plain Layout
  6621. \begin_inset Caption Standard
  6622. \begin_layout Plain Layout
  6623. \series bold
  6624. \begin_inset Argument 1
  6625. status collapsed
  6626. \begin_layout Plain Layout
  6627. Fraction of genic reads in each sample aligned to non-globin genes, with
  6628. and without globin blocking (GB).
  6629. \end_layout
  6630. \end_inset
  6631. \begin_inset CommandInset label
  6632. LatexCommand label
  6633. name "fig:Fraction-of-genic-reads"
  6634. \end_inset
  6635. Fraction of genic reads in each sample aligned to non-globin genes, with
  6636. and without globin blocking (GB).
  6637. \series default
  6638. All reads in each sequencing library were aligned to the cyno genome, and
  6639. the number of reads uniquely aligning to each gene was counted.
  6640. For each sample, counts were summed separately for all globin genes and
  6641. for the remainder of the genes (non-globin genes), and the fraction of
  6642. genic reads aligned to non-globin genes was computed.
  6643. Each point represents an individual sample.
  6644. Gray + signs indicate the means for globin-blocked libraries and unblocked
  6645. libraries.
  6646. The overall distribution for each group is represented as a notched box
  6647. plots.
  6648. Points are randomly spread vertically to avoid excessive overlapping.
  6649. \end_layout
  6650. \end_inset
  6651. \end_layout
  6652. \begin_layout Plain Layout
  6653. \end_layout
  6654. \end_inset
  6655. \end_layout
  6656. \begin_layout Standard
  6657. \begin_inset Float table
  6658. placement p
  6659. wide false
  6660. sideways true
  6661. status collapsed
  6662. \begin_layout Plain Layout
  6663. \align center
  6664. \begin_inset Tabular
  6665. <lyxtabular version="3" rows="4" columns="7">
  6666. <features tabularvalignment="middle">
  6667. <column alignment="center" valignment="top">
  6668. <column alignment="center" valignment="top">
  6669. <column alignment="center" valignment="top">
  6670. <column alignment="center" valignment="top">
  6671. <column alignment="center" valignment="top">
  6672. <column alignment="center" valignment="top">
  6673. <column alignment="center" valignment="top">
  6674. <row>
  6675. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  6676. \begin_inset Text
  6677. \begin_layout Plain Layout
  6678. \end_layout
  6679. \end_inset
  6680. </cell>
  6681. <cell multicolumn="1" alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  6682. \begin_inset Text
  6683. \begin_layout Plain Layout
  6684. \family roman
  6685. \series medium
  6686. \shape up
  6687. \size normal
  6688. \emph off
  6689. \bar no
  6690. \strikeout off
  6691. \xout off
  6692. \uuline off
  6693. \uwave off
  6694. \noun off
  6695. \color none
  6696. Percent of Total Reads
  6697. \end_layout
  6698. \end_inset
  6699. </cell>
  6700. <cell multicolumn="2" alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  6701. \begin_inset Text
  6702. \begin_layout Plain Layout
  6703. \end_layout
  6704. \end_inset
  6705. </cell>
  6706. <cell multicolumn="2" alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  6707. \begin_inset Text
  6708. \begin_layout Plain Layout
  6709. \end_layout
  6710. \end_inset
  6711. </cell>
  6712. <cell multicolumn="2" alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  6713. \begin_inset Text
  6714. \begin_layout Plain Layout
  6715. \end_layout
  6716. \end_inset
  6717. </cell>
  6718. <cell multicolumn="1" alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  6719. \begin_inset Text
  6720. \begin_layout Plain Layout
  6721. \family roman
  6722. \series medium
  6723. \shape up
  6724. \size normal
  6725. \emph off
  6726. \bar no
  6727. \strikeout off
  6728. \xout off
  6729. \uuline off
  6730. \uwave off
  6731. \noun off
  6732. \color none
  6733. Percent of Genic Reads
  6734. \end_layout
  6735. \end_inset
  6736. </cell>
  6737. <cell multicolumn="2" alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  6738. \begin_inset Text
  6739. \begin_layout Plain Layout
  6740. \end_layout
  6741. \end_inset
  6742. </cell>
  6743. </row>
  6744. <row>
  6745. <cell alignment="center" valignment="top" bottomline="true" leftline="true" usebox="none">
  6746. \begin_inset Text
  6747. \begin_layout Plain Layout
  6748. GB
  6749. \end_layout
  6750. \end_inset
  6751. </cell>
  6752. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  6753. \begin_inset Text
  6754. \begin_layout Plain Layout
  6755. \family roman
  6756. \series medium
  6757. \shape up
  6758. \size normal
  6759. \emph off
  6760. \bar no
  6761. \strikeout off
  6762. \xout off
  6763. \uuline off
  6764. \uwave off
  6765. \noun off
  6766. \color none
  6767. Non-globin Reads
  6768. \end_layout
  6769. \end_inset
  6770. </cell>
  6771. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  6772. \begin_inset Text
  6773. \begin_layout Plain Layout
  6774. \family roman
  6775. \series medium
  6776. \shape up
  6777. \size normal
  6778. \emph off
  6779. \bar no
  6780. \strikeout off
  6781. \xout off
  6782. \uuline off
  6783. \uwave off
  6784. \noun off
  6785. \color none
  6786. Globin Reads
  6787. \end_layout
  6788. \end_inset
  6789. </cell>
  6790. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  6791. \begin_inset Text
  6792. \begin_layout Plain Layout
  6793. \family roman
  6794. \series medium
  6795. \shape up
  6796. \size normal
  6797. \emph off
  6798. \bar no
  6799. \strikeout off
  6800. \xout off
  6801. \uuline off
  6802. \uwave off
  6803. \noun off
  6804. \color none
  6805. All Genic Reads
  6806. \end_layout
  6807. \end_inset
  6808. </cell>
  6809. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  6810. \begin_inset Text
  6811. \begin_layout Plain Layout
  6812. \family roman
  6813. \series medium
  6814. \shape up
  6815. \size normal
  6816. \emph off
  6817. \bar no
  6818. \strikeout off
  6819. \xout off
  6820. \uuline off
  6821. \uwave off
  6822. \noun off
  6823. \color none
  6824. All Aligned Reads
  6825. \end_layout
  6826. \end_inset
  6827. </cell>
  6828. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  6829. \begin_inset Text
  6830. \begin_layout Plain Layout
  6831. \family roman
  6832. \series medium
  6833. \shape up
  6834. \size normal
  6835. \emph off
  6836. \bar no
  6837. \strikeout off
  6838. \xout off
  6839. \uuline off
  6840. \uwave off
  6841. \noun off
  6842. \color none
  6843. Non-globin Reads
  6844. \end_layout
  6845. \end_inset
  6846. </cell>
  6847. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  6848. \begin_inset Text
  6849. \begin_layout Plain Layout
  6850. \family roman
  6851. \series medium
  6852. \shape up
  6853. \size normal
  6854. \emph off
  6855. \bar no
  6856. \strikeout off
  6857. \xout off
  6858. \uuline off
  6859. \uwave off
  6860. \noun off
  6861. \color none
  6862. Globin Reads
  6863. \end_layout
  6864. \end_inset
  6865. </cell>
  6866. </row>
  6867. <row>
  6868. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  6869. \begin_inset Text
  6870. \begin_layout Plain Layout
  6871. \family roman
  6872. \series medium
  6873. \shape up
  6874. \size normal
  6875. \emph off
  6876. \bar no
  6877. \strikeout off
  6878. \xout off
  6879. \uuline off
  6880. \uwave off
  6881. \noun off
  6882. \color none
  6883. Yes
  6884. \end_layout
  6885. \end_inset
  6886. </cell>
  6887. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  6888. \begin_inset Text
  6889. \begin_layout Plain Layout
  6890. \family roman
  6891. \series medium
  6892. \shape up
  6893. \size normal
  6894. \emph off
  6895. \bar no
  6896. \strikeout off
  6897. \xout off
  6898. \uuline off
  6899. \uwave off
  6900. \noun off
  6901. \color none
  6902. 50.4% ± 6.82
  6903. \end_layout
  6904. \end_inset
  6905. </cell>
  6906. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  6907. \begin_inset Text
  6908. \begin_layout Plain Layout
  6909. \family roman
  6910. \series medium
  6911. \shape up
  6912. \size normal
  6913. \emph off
  6914. \bar no
  6915. \strikeout off
  6916. \xout off
  6917. \uuline off
  6918. \uwave off
  6919. \noun off
  6920. \color none
  6921. 3.48% ± 2.94
  6922. \end_layout
  6923. \end_inset
  6924. </cell>
  6925. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  6926. \begin_inset Text
  6927. \begin_layout Plain Layout
  6928. \family roman
  6929. \series medium
  6930. \shape up
  6931. \size normal
  6932. \emph off
  6933. \bar no
  6934. \strikeout off
  6935. \xout off
  6936. \uuline off
  6937. \uwave off
  6938. \noun off
  6939. \color none
  6940. 53.9% ± 6.81
  6941. \end_layout
  6942. \end_inset
  6943. </cell>
  6944. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  6945. \begin_inset Text
  6946. \begin_layout Plain Layout
  6947. \family roman
  6948. \series medium
  6949. \shape up
  6950. \size normal
  6951. \emph off
  6952. \bar no
  6953. \strikeout off
  6954. \xout off
  6955. \uuline off
  6956. \uwave off
  6957. \noun off
  6958. \color none
  6959. 89.7% ± 2.40
  6960. \end_layout
  6961. \end_inset
  6962. </cell>
  6963. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  6964. \begin_inset Text
  6965. \begin_layout Plain Layout
  6966. \family roman
  6967. \series medium
  6968. \shape up
  6969. \size normal
  6970. \emph off
  6971. \bar no
  6972. \strikeout off
  6973. \xout off
  6974. \uuline off
  6975. \uwave off
  6976. \noun off
  6977. \color none
  6978. 93.5% ± 5.25
  6979. \end_layout
  6980. \end_inset
  6981. </cell>
  6982. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  6983. \begin_inset Text
  6984. \begin_layout Plain Layout
  6985. \family roman
  6986. \series medium
  6987. \shape up
  6988. \size normal
  6989. \emph off
  6990. \bar no
  6991. \strikeout off
  6992. \xout off
  6993. \uuline off
  6994. \uwave off
  6995. \noun off
  6996. \color none
  6997. 6.49% ± 5.25
  6998. \end_layout
  6999. \end_inset
  7000. </cell>
  7001. </row>
  7002. <row>
  7003. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  7004. \begin_inset Text
  7005. \begin_layout Plain Layout
  7006. \family roman
  7007. \series medium
  7008. \shape up
  7009. \size normal
  7010. \emph off
  7011. \bar no
  7012. \strikeout off
  7013. \xout off
  7014. \uuline off
  7015. \uwave off
  7016. \noun off
  7017. \color none
  7018. No
  7019. \end_layout
  7020. \end_inset
  7021. </cell>
  7022. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  7023. \begin_inset Text
  7024. \begin_layout Plain Layout
  7025. \family roman
  7026. \series medium
  7027. \shape up
  7028. \size normal
  7029. \emph off
  7030. \bar no
  7031. \strikeout off
  7032. \xout off
  7033. \uuline off
  7034. \uwave off
  7035. \noun off
  7036. \color none
  7037. 26.3% ± 8.95
  7038. \end_layout
  7039. \end_inset
  7040. </cell>
  7041. <cell alignment="center" valignment="top" topline="true" bottomline="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. 44.6% ± 16.6
  7057. \end_layout
  7058. \end_inset
  7059. </cell>
  7060. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  7061. \begin_inset Text
  7062. \begin_layout Plain Layout
  7063. \family roman
  7064. \series medium
  7065. \shape up
  7066. \size normal
  7067. \emph off
  7068. \bar no
  7069. \strikeout off
  7070. \xout off
  7071. \uuline off
  7072. \uwave off
  7073. \noun off
  7074. \color none
  7075. 70.1% ± 9.38
  7076. \end_layout
  7077. \end_inset
  7078. </cell>
  7079. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  7080. \begin_inset Text
  7081. \begin_layout Plain Layout
  7082. \family roman
  7083. \series medium
  7084. \shape up
  7085. \size normal
  7086. \emph off
  7087. \bar no
  7088. \strikeout off
  7089. \xout off
  7090. \uuline off
  7091. \uwave off
  7092. \noun off
  7093. \color none
  7094. 90.7% ± 5.16
  7095. \end_layout
  7096. \end_inset
  7097. </cell>
  7098. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  7099. \begin_inset Text
  7100. \begin_layout Plain Layout
  7101. \family roman
  7102. \series medium
  7103. \shape up
  7104. \size normal
  7105. \emph off
  7106. \bar no
  7107. \strikeout off
  7108. \xout off
  7109. \uuline off
  7110. \uwave off
  7111. \noun off
  7112. \color none
  7113. 38.8% ± 17.1
  7114. \end_layout
  7115. \end_inset
  7116. </cell>
  7117. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  7118. \begin_inset Text
  7119. \begin_layout Plain Layout
  7120. \family roman
  7121. \series medium
  7122. \shape up
  7123. \size normal
  7124. \emph off
  7125. \bar no
  7126. \strikeout off
  7127. \xout off
  7128. \uuline off
  7129. \uwave off
  7130. \noun off
  7131. \color none
  7132. 61.2% ± 17.1
  7133. \end_layout
  7134. \end_inset
  7135. </cell>
  7136. </row>
  7137. </lyxtabular>
  7138. \end_inset
  7139. \end_layout
  7140. \begin_layout Plain Layout
  7141. \begin_inset Caption Standard
  7142. \begin_layout Plain Layout
  7143. \series bold
  7144. \begin_inset Argument 1
  7145. status collapsed
  7146. \begin_layout Plain Layout
  7147. Fractions of reads mapping to genomic features in GB and non-GB samples.
  7148. \end_layout
  7149. \end_inset
  7150. \begin_inset CommandInset label
  7151. LatexCommand label
  7152. name "tab:Fractions-of-reads"
  7153. \end_inset
  7154. Fractions of reads mapping to genomic features in GB and non-GB samples.
  7155. \series default
  7156. All values are given as mean ± standard deviation.
  7157. \end_layout
  7158. \end_inset
  7159. \end_layout
  7160. \begin_layout Plain Layout
  7161. \end_layout
  7162. \end_inset
  7163. \end_layout
  7164. \begin_layout Standard
  7165. Another important aspect is that the standard deviations in Table
  7166. \begin_inset CommandInset ref
  7167. LatexCommand ref
  7168. reference "tab:Fractions-of-reads"
  7169. plural "false"
  7170. caps "false"
  7171. noprefix "false"
  7172. \end_inset
  7173. are uniformly smaller in the GB samples than the non-GB ones, indicating
  7174. much greater consistency of yield.
  7175. This is best seen in the percentage of non-globin reads as a fraction of
  7176. total reads aligned to annotated genes (genic reads).
  7177. For the non-GB samples, this measure ranges from 10.9% to 80.9%, while for
  7178. the GB samples it ranges from 81.9% to 99.9% (Figure
  7179. \begin_inset CommandInset ref
  7180. LatexCommand ref
  7181. reference "fig:Fraction-of-genic-reads"
  7182. plural "false"
  7183. caps "false"
  7184. noprefix "false"
  7185. \end_inset
  7186. ).
  7187. This means that for applications where it is critical that each sample
  7188. achieve a specified minimum coverage in order to provide useful information,
  7189. it would be necessary to budget up to 10 times the sequencing depth per
  7190. sample without globin blocking, even though the average yield improvement
  7191. for globin blocking is only 2-fold, because every sample has a chance of
  7192. being 90% globin and 10% useful reads.
  7193. Hence, the more consistent behavior of GB samples makes planning an experiment
  7194. easier and more efficient because it eliminates the need to over-sequence
  7195. every sample in order to guard against the worst case of a high-globin
  7196. fraction.
  7197. \end_layout
  7198. \begin_layout Subsection
  7199. Globin blocking lowers the noise floor and allows detection of about 2000
  7200. more genes
  7201. \end_layout
  7202. \begin_layout Standard
  7203. \begin_inset Flex TODO Note (inline)
  7204. status open
  7205. \begin_layout Plain Layout
  7206. Remove redundant titles from figures
  7207. \end_layout
  7208. \end_inset
  7209. \end_layout
  7210. \begin_layout Standard
  7211. \begin_inset Float figure
  7212. wide false
  7213. sideways false
  7214. status collapsed
  7215. \begin_layout Plain Layout
  7216. \align center
  7217. \begin_inset Graphics
  7218. filename graphics/Globin Paper/figure2 - aveLogCPM-colored.pdf
  7219. lyxscale 50
  7220. width 100col%
  7221. groupId colwidth
  7222. \end_inset
  7223. \end_layout
  7224. \begin_layout Plain Layout
  7225. \begin_inset Caption Standard
  7226. \begin_layout Plain Layout
  7227. \series bold
  7228. \begin_inset Argument 1
  7229. status collapsed
  7230. \begin_layout Plain Layout
  7231. Distributions of average group gene abundances when normalized separately
  7232. or together.
  7233. \end_layout
  7234. \end_inset
  7235. \begin_inset CommandInset label
  7236. LatexCommand label
  7237. name "fig:logcpm-dists"
  7238. \end_inset
  7239. Distributions of average group gene abundances when normalized separately
  7240. or together.
  7241. \series default
  7242. All reads in each sequencing library were aligned to the cyno genome, and
  7243. the number of reads uniquely aligning to each gene was counted.
  7244. Genes with zero counts in all libraries were discarded.
  7245. Libraries were normalized using the TMM method.
  7246. Libraries were split into globin-blocked (GB) and non-GB groups and the
  7247. average abundance for each gene in both groups, measured in log2 counts
  7248. per million reads counted, was computed using the aveLogCPM function.
  7249. The distribution of average gene logCPM values was plotted for both groups
  7250. using a kernel density plot to approximate a continuous distribution.
  7251. The logCPM GB distributions are marked in red, non-GB in blue.
  7252. The black vertical line denotes the chosen detection threshold of -1.
  7253. Top panel: Libraries were split into GB and non-GB groups first and normalized
  7254. separately.
  7255. Bottom panel: Libraries were all normalized together first and then split
  7256. into groups.
  7257. \end_layout
  7258. \end_inset
  7259. \end_layout
  7260. \begin_layout Plain Layout
  7261. \end_layout
  7262. \end_inset
  7263. \end_layout
  7264. \begin_layout Standard
  7265. Since globin blocking yields more usable sequencing depth, it should also
  7266. allow detection of more genes at any given threshold.
  7267. When we looked at the distribution of average normalized logCPM values
  7268. across all libraries for genes with at least one read assigned to them,
  7269. we observed the expected bimodal distribution, with a high-abundance "signal"
  7270. peak representing detected genes and a low-abundance "noise" peak representing
  7271. genes whose read count did not rise above the noise floor (Figure
  7272. \begin_inset CommandInset ref
  7273. LatexCommand ref
  7274. reference "fig:logcpm-dists"
  7275. plural "false"
  7276. caps "false"
  7277. noprefix "false"
  7278. \end_inset
  7279. ).
  7280. Consistent with the 2-fold increase in raw counts assigned to non-globin
  7281. genes, the signal peak for GB samples is shifted to the right relative
  7282. to the non-GB signal peak.
  7283. When all the samples are normalized together, this difference is normalized
  7284. out, lining up the signal peaks, and this reveals that, as expected, the
  7285. noise floor for the GB samples is about 2-fold lower.
  7286. This greater separation between signal and noise peaks in the GB samples
  7287. means that low-expression genes should be more easily detected and more
  7288. precisely quantified than in the non-GB samples.
  7289. \end_layout
  7290. \begin_layout Standard
  7291. \begin_inset Float figure
  7292. wide false
  7293. sideways false
  7294. status collapsed
  7295. \begin_layout Plain Layout
  7296. \align center
  7297. \begin_inset Graphics
  7298. filename graphics/Globin Paper/figure3 - detection.pdf
  7299. lyxscale 50
  7300. width 100col%
  7301. groupId colwidth
  7302. \end_inset
  7303. \end_layout
  7304. \begin_layout Plain Layout
  7305. \begin_inset Caption Standard
  7306. \begin_layout Plain Layout
  7307. \series bold
  7308. \begin_inset Argument 1
  7309. status collapsed
  7310. \begin_layout Plain Layout
  7311. Gene detections as a function of abundance thresholds in globin-blocked
  7312. (GB) and non-GB samples.
  7313. \end_layout
  7314. \end_inset
  7315. \begin_inset CommandInset label
  7316. LatexCommand label
  7317. name "fig:Gene-detections"
  7318. \end_inset
  7319. Gene detections as a function of abundance thresholds in globin-blocked
  7320. (GB) and non-GB samples.
  7321. \series default
  7322. Average abundance (logCPM,
  7323. \begin_inset Formula $\log_{2}$
  7324. \end_inset
  7325. counts per million reads counted) was computed by separate group normalization
  7326. as described in Figure
  7327. \begin_inset CommandInset ref
  7328. LatexCommand ref
  7329. reference "fig:logcpm-dists"
  7330. plural "false"
  7331. caps "false"
  7332. noprefix "false"
  7333. \end_inset
  7334. for both the GB and non-GB groups, as well as for all samples considered
  7335. as one large group.
  7336. For each every integer threshold from -2 to 3, the number of genes detected
  7337. at or above that logCPM threshold was plotted for each group.
  7338. \end_layout
  7339. \end_inset
  7340. \end_layout
  7341. \begin_layout Plain Layout
  7342. \end_layout
  7343. \end_inset
  7344. \end_layout
  7345. \begin_layout Standard
  7346. Based on these distributions, we selected a detection threshold of -1, which
  7347. is approximately the leftmost edge of the trough between the signal and
  7348. noise peaks.
  7349. This represents the most liberal possible detection threshold that doesn't
  7350. call substantial numbers of noise genes as detected.
  7351. Among the full dataset, 13429 genes were detected at this threshold, and
  7352. 22276 were not.
  7353. When considering the GB libraries and non-GB libraries separately and re-comput
  7354. ing normalization factors independently within each group, 14535 genes were
  7355. detected in the GB libraries while only 12460 were detected in the non-GB
  7356. libraries.
  7357. Thus, GB allowed the detection of 2000 extra genes that were buried under
  7358. the noise floor without GB.
  7359. This pattern of at least 2000 additional genes detected with GB was also
  7360. consistent across a wide range of possible detection thresholds, from -2
  7361. to 3 (see Figure
  7362. \begin_inset CommandInset ref
  7363. LatexCommand ref
  7364. reference "fig:Gene-detections"
  7365. plural "false"
  7366. caps "false"
  7367. noprefix "false"
  7368. \end_inset
  7369. ).
  7370. \end_layout
  7371. \begin_layout Subsection
  7372. Globin blocking does not add significant additional noise or decrease sample
  7373. quality
  7374. \end_layout
  7375. \begin_layout Standard
  7376. One potential worry is that the globin blocking protocol could perturb the
  7377. levels of non-globin genes.
  7378. There are two kinds of possible perturbations: systematic and random.
  7379. The former is not a major concern for detection of differential expression,
  7380. since a 2-fold change in every sample has no effect on the relative fold
  7381. change between samples.
  7382. In contrast, random perturbations would increase the noise and obscure
  7383. the signal in the dataset, reducing the capacity to detect differential
  7384. expression.
  7385. \end_layout
  7386. \begin_layout Standard
  7387. \begin_inset Float figure
  7388. wide false
  7389. sideways false
  7390. status collapsed
  7391. \begin_layout Plain Layout
  7392. \align center
  7393. \begin_inset Graphics
  7394. filename graphics/Globin Paper/figure4 - maplot-colored.pdf
  7395. lyxscale 50
  7396. width 100col%
  7397. groupId colwidth
  7398. \end_inset
  7399. \end_layout
  7400. \begin_layout Plain Layout
  7401. \begin_inset Caption Standard
  7402. \begin_layout Plain Layout
  7403. \begin_inset Argument 1
  7404. status collapsed
  7405. \begin_layout Plain Layout
  7406. MA plot showing effects of globin blocking on each gene's abundance.
  7407. \end_layout
  7408. \end_inset
  7409. \begin_inset CommandInset label
  7410. LatexCommand label
  7411. name "fig:MA-plot"
  7412. \end_inset
  7413. \series bold
  7414. MA plot showing effects of globin blocking on each gene's abundance.
  7415. \series default
  7416. All libraries were normalized together as described in Figure
  7417. \begin_inset CommandInset ref
  7418. LatexCommand ref
  7419. reference "fig:logcpm-dists"
  7420. plural "false"
  7421. caps "false"
  7422. noprefix "false"
  7423. \end_inset
  7424. , and genes with an average logCPM below -1 were filtered out.
  7425. Each remaining gene was tested for differential abundance with respect
  7426. to globin blocking (GB) using edgeR’s quasi-likelihod F-test, fitting a
  7427. negative binomial generalized linear model to table of read counts in each
  7428. library.
  7429. For each gene, edgeR reported average abundance (logCPM),
  7430. \begin_inset Formula $\log_{2}$
  7431. \end_inset
  7432. fold change (logFC), p-value, and Benjamini-Hochberg adjusted false discovery
  7433. rate (FDR).
  7434. Each gene's logFC was plotted against its logCPM, colored by FDR.
  7435. Red points are significant at ≤10% FDR, and blue are not significant at
  7436. that threshold.
  7437. The alpha and beta globin genes targeted for blocking are marked with large
  7438. triangles, while all other genes are represented as small points.
  7439. \end_layout
  7440. \end_inset
  7441. \end_layout
  7442. \begin_layout Plain Layout
  7443. \end_layout
  7444. \end_inset
  7445. \end_layout
  7446. \begin_layout Standard
  7447. \begin_inset Flex TODO Note (inline)
  7448. status open
  7449. \begin_layout Plain Layout
  7450. Standardize on
  7451. \begin_inset Quotes eld
  7452. \end_inset
  7453. log2
  7454. \begin_inset Quotes erd
  7455. \end_inset
  7456. notation
  7457. \end_layout
  7458. \end_inset
  7459. \end_layout
  7460. \begin_layout Standard
  7461. The data do indeed show small systematic perturbations in gene levels (Figure
  7462. \begin_inset CommandInset ref
  7463. LatexCommand ref
  7464. reference "fig:MA-plot"
  7465. plural "false"
  7466. caps "false"
  7467. noprefix "false"
  7468. \end_inset
  7469. ).
  7470. Other than the 3 designated alpha and beta globin genes, two other genes
  7471. stand out as having especially large negative log fold changes: HBD and
  7472. LOC1021365.
  7473. HBD, delta globin, is most likely targeted by the blocking oligos due to
  7474. high sequence homology with the other globin genes.
  7475. LOC1021365 is the aforementioned ncRNA that is reverse-complementary to
  7476. one of the alpha-like genes and that would be expected to be removed during
  7477. the globin blocking step.
  7478. All other genes appear in a cluster centered vertically at 0, and the vast
  7479. majority of genes in this cluster show an absolute log2(FC) of 0.5 or less.
  7480. Nevertheless, many of these small perturbations are still statistically
  7481. significant, indicating that the globin blocking oligos likely cause very
  7482. small but non-zero systematic perturbations in measured gene expression
  7483. levels.
  7484. \end_layout
  7485. \begin_layout Standard
  7486. \begin_inset Float figure
  7487. wide false
  7488. sideways false
  7489. status collapsed
  7490. \begin_layout Plain Layout
  7491. \align center
  7492. \begin_inset Graphics
  7493. filename graphics/Globin Paper/figure5 - corrplot.pdf
  7494. lyxscale 50
  7495. width 100col%
  7496. groupId colwidth
  7497. \end_inset
  7498. \end_layout
  7499. \begin_layout Plain Layout
  7500. \begin_inset Caption Standard
  7501. \begin_layout Plain Layout
  7502. \series bold
  7503. \begin_inset Argument 1
  7504. status collapsed
  7505. \begin_layout Plain Layout
  7506. Comparison of inter-sample gene abundance correlations with and without
  7507. globin blocking.
  7508. \end_layout
  7509. \end_inset
  7510. \begin_inset CommandInset label
  7511. LatexCommand label
  7512. name "fig:gene-abundance-correlations"
  7513. \end_inset
  7514. Comparison of inter-sample gene abundance correlations with and without
  7515. globin blocking (GB).
  7516. \series default
  7517. All libraries were normalized together as described in Figure 2, and genes
  7518. with an average abundance (logCPM, log2 counts per million reads counted)
  7519. less than -1 were filtered out.
  7520. Each gene’s logCPM was computed in each library using the edgeR cpm function.
  7521. For each pair of biological samples, the Pearson correlation between those
  7522. samples' GB libraries was plotted against the correlation between the same
  7523. samples’ non-GB libraries.
  7524. Each point represents an unique pair of samples.
  7525. The solid gray line shows a quantile-quantile plot of distribution of GB
  7526. correlations vs.
  7527. that of non-GB correlations.
  7528. The thin dashed line is the identity line, provided for reference.
  7529. \end_layout
  7530. \end_inset
  7531. \end_layout
  7532. \begin_layout Plain Layout
  7533. \end_layout
  7534. \end_inset
  7535. \end_layout
  7536. \begin_layout Standard
  7537. To evaluate the possibility of globin blocking causing random perturbations
  7538. and reducing sample quality, we computed the Pearson correlation between
  7539. logCPM values for every pair of samples with and without GB and plotted
  7540. them against each other (Figure
  7541. \begin_inset CommandInset ref
  7542. LatexCommand ref
  7543. reference "fig:gene-abundance-correlations"
  7544. plural "false"
  7545. caps "false"
  7546. noprefix "false"
  7547. \end_inset
  7548. ).
  7549. The plot indicated that the GB libraries have higher sample-to-sample correlati
  7550. ons than the non-GB libraries.
  7551. Parametric and nonparametric tests for differences between the correlations
  7552. with and without GB both confirmed that this difference was highly significant
  7553. (2-sided paired t-test: t = 37.2, df = 665, P ≪ 2.2e-16; 2-sided Wilcoxon
  7554. sign-rank test: V = 2195, P ≪ 2.2e-16).
  7555. Performing the same tests on the Spearman correlations gave the same conclusion
  7556. (t-test: t = 26.8, df = 665, P ≪ 2.2e-16; sign-rank test: V = 8781, P ≪ 2.2e-16).
  7557. The edgeR package was used to compute the overall biological coefficient
  7558. of variation (BCV) for GB and non-GB libraries, and found that globin blocking
  7559. resulted in a negligible increase in the BCV (0.417 with GB vs.
  7560. 0.400 without).
  7561. The near equality of the BCVs for both sets indicates that the higher correlati
  7562. ons in the GB libraries are most likely a result of the increased yield
  7563. of useful reads, which reduces the contribution of Poisson counting uncertainty
  7564. to the overall variance of the logCPM values
  7565. \begin_inset CommandInset citation
  7566. LatexCommand cite
  7567. key "McCarthy2012"
  7568. literal "false"
  7569. \end_inset
  7570. .
  7571. This improves the precision of expression measurements and more than offsets
  7572. the negligible increase in BCV.
  7573. \end_layout
  7574. \begin_layout Subsection
  7575. More differentially expressed genes are detected with globin blocking
  7576. \end_layout
  7577. \begin_layout Standard
  7578. \begin_inset Float table
  7579. wide false
  7580. sideways false
  7581. status collapsed
  7582. \begin_layout Plain Layout
  7583. \align center
  7584. \begin_inset Tabular
  7585. <lyxtabular version="3" rows="5" columns="5">
  7586. <features tabularvalignment="middle">
  7587. <column alignment="center" valignment="top">
  7588. <column alignment="center" valignment="top">
  7589. <column alignment="center" valignment="top">
  7590. <column alignment="center" valignment="top">
  7591. <column alignment="center" valignment="top">
  7592. <row>
  7593. <cell alignment="center" valignment="top" usebox="none">
  7594. \begin_inset Text
  7595. \begin_layout Plain Layout
  7596. \end_layout
  7597. \end_inset
  7598. </cell>
  7599. <cell alignment="center" valignment="top" usebox="none">
  7600. \begin_inset Text
  7601. \begin_layout Plain Layout
  7602. \end_layout
  7603. \end_inset
  7604. </cell>
  7605. <cell multicolumn="1" alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  7606. \begin_inset Text
  7607. \begin_layout Plain Layout
  7608. \series bold
  7609. No Globin Blocking
  7610. \end_layout
  7611. \end_inset
  7612. </cell>
  7613. <cell multicolumn="2" alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  7614. \begin_inset Text
  7615. \begin_layout Plain Layout
  7616. \end_layout
  7617. \end_inset
  7618. </cell>
  7619. <cell multicolumn="2" alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  7620. \begin_inset Text
  7621. \begin_layout Plain Layout
  7622. \end_layout
  7623. \end_inset
  7624. </cell>
  7625. </row>
  7626. <row>
  7627. <cell alignment="center" valignment="top" usebox="none">
  7628. \begin_inset Text
  7629. \begin_layout Plain Layout
  7630. \end_layout
  7631. \end_inset
  7632. </cell>
  7633. <cell alignment="center" valignment="top" usebox="none">
  7634. \begin_inset Text
  7635. \begin_layout Plain Layout
  7636. \end_layout
  7637. \end_inset
  7638. </cell>
  7639. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  7640. \begin_inset Text
  7641. \begin_layout Plain Layout
  7642. \series bold
  7643. Up
  7644. \end_layout
  7645. \end_inset
  7646. </cell>
  7647. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  7648. \begin_inset Text
  7649. \begin_layout Plain Layout
  7650. \series bold
  7651. NS
  7652. \end_layout
  7653. \end_inset
  7654. </cell>
  7655. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  7656. \begin_inset Text
  7657. \begin_layout Plain Layout
  7658. \series bold
  7659. Down
  7660. \end_layout
  7661. \end_inset
  7662. </cell>
  7663. </row>
  7664. <row>
  7665. <cell multirow="3" alignment="center" valignment="middle" topline="true" bottomline="true" leftline="true" usebox="none">
  7666. \begin_inset Text
  7667. \begin_layout Plain Layout
  7668. \series bold
  7669. Globin-Blocking
  7670. \end_layout
  7671. \end_inset
  7672. </cell>
  7673. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  7674. \begin_inset Text
  7675. \begin_layout Plain Layout
  7676. \series bold
  7677. Up
  7678. \end_layout
  7679. \end_inset
  7680. </cell>
  7681. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  7682. \begin_inset Text
  7683. \begin_layout Plain Layout
  7684. \family roman
  7685. \series medium
  7686. \shape up
  7687. \size normal
  7688. \emph off
  7689. \bar no
  7690. \strikeout off
  7691. \xout off
  7692. \uuline off
  7693. \uwave off
  7694. \noun off
  7695. \color none
  7696. 231
  7697. \end_layout
  7698. \end_inset
  7699. </cell>
  7700. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  7701. \begin_inset Text
  7702. \begin_layout Plain Layout
  7703. \family roman
  7704. \series medium
  7705. \shape up
  7706. \size normal
  7707. \emph off
  7708. \bar no
  7709. \strikeout off
  7710. \xout off
  7711. \uuline off
  7712. \uwave off
  7713. \noun off
  7714. \color none
  7715. 515
  7716. \end_layout
  7717. \end_inset
  7718. </cell>
  7719. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  7720. \begin_inset Text
  7721. \begin_layout Plain Layout
  7722. \family roman
  7723. \series medium
  7724. \shape up
  7725. \size normal
  7726. \emph off
  7727. \bar no
  7728. \strikeout off
  7729. \xout off
  7730. \uuline off
  7731. \uwave off
  7732. \noun off
  7733. \color none
  7734. 2
  7735. \end_layout
  7736. \end_inset
  7737. </cell>
  7738. </row>
  7739. <row>
  7740. <cell multirow="4" alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  7741. \begin_inset Text
  7742. \begin_layout Plain Layout
  7743. \end_layout
  7744. \end_inset
  7745. </cell>
  7746. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  7747. \begin_inset Text
  7748. \begin_layout Plain Layout
  7749. \series bold
  7750. NS
  7751. \end_layout
  7752. \end_inset
  7753. </cell>
  7754. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  7755. \begin_inset Text
  7756. \begin_layout Plain Layout
  7757. \family roman
  7758. \series medium
  7759. \shape up
  7760. \size normal
  7761. \emph off
  7762. \bar no
  7763. \strikeout off
  7764. \xout off
  7765. \uuline off
  7766. \uwave off
  7767. \noun off
  7768. \color none
  7769. 160
  7770. \end_layout
  7771. \end_inset
  7772. </cell>
  7773. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  7774. \begin_inset Text
  7775. \begin_layout Plain Layout
  7776. \family roman
  7777. \series medium
  7778. \shape up
  7779. \size normal
  7780. \emph off
  7781. \bar no
  7782. \strikeout off
  7783. \xout off
  7784. \uuline off
  7785. \uwave off
  7786. \noun off
  7787. \color none
  7788. 11235
  7789. \end_layout
  7790. \end_inset
  7791. </cell>
  7792. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  7793. \begin_inset Text
  7794. \begin_layout Plain Layout
  7795. \family roman
  7796. \series medium
  7797. \shape up
  7798. \size normal
  7799. \emph off
  7800. \bar no
  7801. \strikeout off
  7802. \xout off
  7803. \uuline off
  7804. \uwave off
  7805. \noun off
  7806. \color none
  7807. 136
  7808. \end_layout
  7809. \end_inset
  7810. </cell>
  7811. </row>
  7812. <row>
  7813. <cell multirow="4" alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  7814. \begin_inset Text
  7815. \begin_layout Plain Layout
  7816. \end_layout
  7817. \end_inset
  7818. </cell>
  7819. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  7820. \begin_inset Text
  7821. \begin_layout Plain Layout
  7822. \series bold
  7823. Down
  7824. \end_layout
  7825. \end_inset
  7826. </cell>
  7827. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  7828. \begin_inset Text
  7829. \begin_layout Plain Layout
  7830. \family roman
  7831. \series medium
  7832. \shape up
  7833. \size normal
  7834. \emph off
  7835. \bar no
  7836. \strikeout off
  7837. \xout off
  7838. \uuline off
  7839. \uwave off
  7840. \noun off
  7841. \color none
  7842. 0
  7843. \end_layout
  7844. \end_inset
  7845. </cell>
  7846. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  7847. \begin_inset Text
  7848. \begin_layout Plain Layout
  7849. \family roman
  7850. \series medium
  7851. \shape up
  7852. \size normal
  7853. \emph off
  7854. \bar no
  7855. \strikeout off
  7856. \xout off
  7857. \uuline off
  7858. \uwave off
  7859. \noun off
  7860. \color none
  7861. 548
  7862. \end_layout
  7863. \end_inset
  7864. </cell>
  7865. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  7866. \begin_inset Text
  7867. \begin_layout Plain Layout
  7868. \family roman
  7869. \series medium
  7870. \shape up
  7871. \size normal
  7872. \emph off
  7873. \bar no
  7874. \strikeout off
  7875. \xout off
  7876. \uuline off
  7877. \uwave off
  7878. \noun off
  7879. \color none
  7880. 127
  7881. \end_layout
  7882. \end_inset
  7883. </cell>
  7884. </row>
  7885. </lyxtabular>
  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 Argument 1
  7893. status open
  7894. \begin_layout Plain Layout
  7895. Comparison of significantly differentially expressed genes with and without
  7896. globin blocking.
  7897. \end_layout
  7898. \end_inset
  7899. \begin_inset CommandInset label
  7900. LatexCommand label
  7901. name "tab:Comparison-of-significant"
  7902. \end_inset
  7903. Comparison of significantly differentially expressed genes with and without
  7904. globin blocking.
  7905. \series default
  7906. Up, Down: Genes significantly up/down-regulated in post-transplant samples
  7907. relative to pre-transplant samples, with a false discovery rate of 10%
  7908. or less.
  7909. NS: Non-significant genes (false discovery rate greater than 10%).
  7910. \end_layout
  7911. \end_inset
  7912. \end_layout
  7913. \begin_layout Plain Layout
  7914. \end_layout
  7915. \end_inset
  7916. \end_layout
  7917. \begin_layout Standard
  7918. To compare performance on differential gene expression tests, we took subsets
  7919. of both the GB and non-GB libraries with exactly one pre-transplant and
  7920. one post-transplant sample for each animal that had paired samples available
  7921. for analysis (N=7 animals, N=14 samples in each subset).
  7922. The same test for pre- vs.
  7923. post-transplant differential gene expression was performed on the same
  7924. 7 pairs of samples from GB libraries and non-GB libraries, in each case
  7925. using an FDR of 10% as the threshold of significance.
  7926. Out of 12954 genes that passed the detection threshold in both subsets,
  7927. 358 were called significantly differentially expressed in the same direction
  7928. in both sets; 1063 were differentially expressed in the GB set only; 296
  7929. were differentially expressed in the non-GB set only; 2 genes were called
  7930. significantly up in the GB set but significantly down in the non-GB set;
  7931. and the remaining 11235 were not called differentially expressed in either
  7932. set.
  7933. These data are summarized in Table
  7934. \begin_inset CommandInset ref
  7935. LatexCommand ref
  7936. reference "tab:Comparison-of-significant"
  7937. plural "false"
  7938. caps "false"
  7939. noprefix "false"
  7940. \end_inset
  7941. .
  7942. The differences in BCV calculated by EdgeR for these subsets of samples
  7943. were negligible (BCV = 0.302 for GB and 0.297 for non-GB).
  7944. \end_layout
  7945. \begin_layout Standard
  7946. The key point is that the GB data results in substantially more differentially
  7947. expressed calls than the non-GB data.
  7948. Since there is no gold standard for this dataset, it is impossible to be
  7949. certain whether this is due to under-calling of differential expression
  7950. in the non-GB samples or over-calling in the GB samples.
  7951. However, given that both datasets are derived from the same biological
  7952. samples and have nearly equal BCVs, it is more likely that the larger number
  7953. of DE calls in the GB samples are genuine detections that were enabled
  7954. by the higher sequencing depth and measurement precision of the GB samples.
  7955. Note that the same set of genes was considered in both subsets, so the
  7956. larger number of differentially expressed gene calls in the GB data set
  7957. reflects a greater sensitivity to detect significant differential gene
  7958. expression and not simply the larger total number of detected genes in
  7959. GB samples described earlier.
  7960. \end_layout
  7961. \begin_layout Section
  7962. Discussion
  7963. \end_layout
  7964. \begin_layout Standard
  7965. The original experience with whole blood gene expression profiling on DNA
  7966. microarrays demonstrated that the high concentration of globin transcripts
  7967. reduced the sensitivity to detect genes with relatively low expression
  7968. levels, in effect, significantly reducing the sensitivity.
  7969. To address this limitation, commercial protocols for globin reduction were
  7970. developed based on strategies to block globin transcript amplification
  7971. during labeling or physically removing globin transcripts by affinity bead
  7972. methods
  7973. \begin_inset CommandInset citation
  7974. LatexCommand cite
  7975. key "Winn2010"
  7976. literal "false"
  7977. \end_inset
  7978. .
  7979. More recently, using the latest generation of labeling protocols and arrays,
  7980. it was determined that globin reduction was no longer necessary to obtain
  7981. sufficient sensitivity to detect differential transcript expression
  7982. \begin_inset CommandInset citation
  7983. LatexCommand cite
  7984. key "NuGEN2010"
  7985. literal "false"
  7986. \end_inset
  7987. .
  7988. However, we are not aware of any publications using these currently available
  7989. protocols the with latest generation of microarrays that actually compare
  7990. the detection sensitivity with and without globin reduction.
  7991. However, in practice this has now been adopted generally primarily driven
  7992. by concerns for cost control.
  7993. The main objective of our work was to directly test the impact of globin
  7994. gene transcripts and a new globin blocking protocol for application to
  7995. the newest generation of differential gene expression profiling determined
  7996. using next generation sequencing.
  7997. \end_layout
  7998. \begin_layout Standard
  7999. The challenge of doing global gene expression profiling in cynomolgus monkeys
  8000. is that the current available arrays were never designed to comprehensively
  8001. cover this genome and have not been updated since the first assemblies
  8002. of the cynomolgus genome were published.
  8003. Therefore, we determined that the best strategy for peripheral blood profiling
  8004. was to do deep RNA-seq and inform the workflow using the latest available
  8005. genome assembly and annotation
  8006. \begin_inset CommandInset citation
  8007. LatexCommand cite
  8008. key "Wilson2013"
  8009. literal "false"
  8010. \end_inset
  8011. .
  8012. However, it was not immediately clear whether globin reduction was necessary
  8013. for RNA-seq or how much improvement in efficiency or sensitivity to detect
  8014. differential gene expression would be achieved for the added cost and work.
  8015. \end_layout
  8016. \begin_layout Standard
  8017. We only found one report that demonstrated that globin reduction significantly
  8018. improved the effective read yields for sequencing of human peripheral blood
  8019. cell RNA using a DeepSAGE protocol
  8020. \begin_inset CommandInset citation
  8021. LatexCommand cite
  8022. key "Mastrokolias2012"
  8023. literal "false"
  8024. \end_inset
  8025. .
  8026. The approach to DeepSAGE involves two different restriction enzymes that
  8027. purify and then tag small fragments of transcripts at specific locations
  8028. and thus, significantly reduces the complexity of the transcriptome.
  8029. Therefore, we could not determine how DeepSAGE results would translate
  8030. to the common strategy in the field for assaying the entire transcript
  8031. population by whole-transcriptome 3’-end RNA-seq.
  8032. Furthermore, if globin reduction is necessary, we also needed a globin
  8033. reduction method specific to cynomolgus globin sequences that would work
  8034. an organism for which no kit is available off the shelf.
  8035. \end_layout
  8036. \begin_layout Standard
  8037. As mentioned above, the addition of globin blocking oligos has a very small
  8038. impact on measured expression levels of gene expression.
  8039. However, this is a non-issue for the purposes of differential expression
  8040. testing, since a systematic change in a gene in all samples does not affect
  8041. relative expression levels between samples.
  8042. However, we must acknowledge that simple comparisons of gene expression
  8043. data obtained by GB and non-GB protocols are not possible without additional
  8044. normalization.
  8045. \end_layout
  8046. \begin_layout Standard
  8047. More importantly, globin blocking not only nearly doubles the yield of usable
  8048. reads, it also increases inter-sample correlation and sensitivity to detect
  8049. differential gene expression relative to the same set of samples profiled
  8050. without blocking.
  8051. In addition, globin blocking does not add a significant amount of random
  8052. noise to the data.
  8053. Globin blocking thus represents a cost-effective way to squeeze more data
  8054. and statistical power out of the same blood samples and the same amount
  8055. of sequencing.
  8056. In conclusion, globin reduction greatly increases the yield of useful RNA-seq
  8057. reads mapping to the rest of the genome, with minimal perturbations in
  8058. the relative levels of non-globin genes.
  8059. Based on these results, globin transcript reduction using sequence-specific,
  8060. complementary blocking oligonucleotides is recommended for all deep RNA-seq
  8061. of cynomolgus and other nonhuman primate blood samples.
  8062. \end_layout
  8063. \begin_layout Chapter
  8064. Future Directions
  8065. \end_layout
  8066. \begin_layout Standard
  8067. \begin_inset Flex TODO Note (inline)
  8068. status open
  8069. \begin_layout Plain Layout
  8070. Consider per-chapter future directions.
  8071. Check instructions.
  8072. \end_layout
  8073. \end_inset
  8074. \end_layout
  8075. \begin_layout Itemize
  8076. Study other epigenetic marks in more contexts
  8077. \end_layout
  8078. \begin_deeper
  8079. \begin_layout Itemize
  8080. DNA methylation, histone marks, chromatin accessibility & conformation in
  8081. CD4 T-cells
  8082. \end_layout
  8083. \begin_layout Itemize
  8084. Also look at other types of lymphocytes: CD8 T-cells, B-cells, NK cells
  8085. \end_layout
  8086. \end_deeper
  8087. \begin_layout Itemize
  8088. Use CV or bootstrap to better evaluate classifiers
  8089. \end_layout
  8090. \begin_layout Itemize
  8091. fRMAtools could be adapted to not require equal-sized groups
  8092. \end_layout
  8093. \begin_layout Standard
  8094. \begin_inset ERT
  8095. status open
  8096. \begin_layout Plain Layout
  8097. % Call it "References" instead of "Bibliography"
  8098. \end_layout
  8099. \begin_layout Plain Layout
  8100. \backslash
  8101. renewcommand{
  8102. \backslash
  8103. bibname}{References}
  8104. \end_layout
  8105. \end_inset
  8106. \end_layout
  8107. \begin_layout Standard
  8108. \begin_inset Flex TODO Note (inline)
  8109. status open
  8110. \begin_layout Plain Layout
  8111. Check bib entry formatting & sort order
  8112. \end_layout
  8113. \end_inset
  8114. \end_layout
  8115. \begin_layout Standard
  8116. \begin_inset Flex TODO Note (inline)
  8117. status open
  8118. \begin_layout Plain Layout
  8119. Check in-text citation format.
  8120. Probably don't just want [1], [2], etc.
  8121. \end_layout
  8122. \end_inset
  8123. \end_layout
  8124. \begin_layout Standard
  8125. \begin_inset CommandInset bibtex
  8126. LatexCommand bibtex
  8127. btprint "btPrintCited"
  8128. bibfiles "refs,code-refs"
  8129. options "bibtotoc,unsrt"
  8130. \end_inset
  8131. \end_layout
  8132. \end_body
  8133. \end_document