thesis.lyx 297 KB

1234567891011121314151617181920212223242526272829303132333435363738394041424344454647484950515253545556575859606162636465666768697071727374757677787980818283848586878889909192939495969798991001011021031041051061071081091101111121131141151161171181191201211221231241251261271281291301311321331341351361371381391401411421431441451461471481491501511521531541551561571581591601611621631641651661671681691701711721731741751761771781791801811821831841851861871881891901911921931941951961971981992002012022032042052062072082092102112122132142152162172182192202212222232242252262272282292302312322332342352362372382392402412422432442452462472482492502512522532542552562572582592602612622632642652662672682692702712722732742752762772782792802812822832842852862872882892902912922932942952962972982993003013023033043053063073083093103113123133143153163173183193203213223233243253263273283293303313323333343353363373383393403413423433443453463473483493503513523533543553563573583593603613623633643653663673683693703713723733743753763773783793803813823833843853863873883893903913923933943953963973983994004014024034044054064074084094104114124134144154164174184194204214224234244254264274284294304314324334344354364374384394404414424434444454464474484494504514524534544554564574584594604614624634644654664674684694704714724734744754764774784794804814824834844854864874884894904914924934944954964974984995005015025035045055065075085095105115125135145155165175185195205215225235245255265275285295305315325335345355365375385395405415425435445455465475485495505515525535545555565575585595605615625635645655665675685695705715725735745755765775785795805815825835845855865875885895905915925935945955965975985996006016026036046056066076086096106116126136146156166176186196206216226236246256266276286296306316326336346356366376386396406416426436446456466476486496506516526536546556566576586596606616626636646656666676686696706716726736746756766776786796806816826836846856866876886896906916926936946956966976986997007017027037047057067077087097107117127137147157167177187197207217227237247257267277287297307317327337347357367377387397407417427437447457467477487497507517527537547557567577587597607617627637647657667677687697707717727737747757767777787797807817827837847857867877887897907917927937947957967977987998008018028038048058068078088098108118128138148158168178188198208218228238248258268278288298308318328338348358368378388398408418428438448458468478488498508518528538548558568578588598608618628638648658668678688698708718728738748758768778788798808818828838848858868878888898908918928938948958968978988999009019029039049059069079089099109119129139149159169179189199209219229239249259269279289299309319329339349359369379389399409419429439449459469479489499509519529539549559569579589599609619629639649659669679689699709719729739749759769779789799809819829839849859869879889899909919929939949959969979989991000100110021003100410051006100710081009101010111012101310141015101610171018101910201021102210231024102510261027102810291030103110321033103410351036103710381039104010411042104310441045104610471048104910501051105210531054105510561057105810591060106110621063106410651066106710681069107010711072107310741075107610771078107910801081108210831084108510861087108810891090109110921093109410951096109710981099110011011102110311041105110611071108110911101111111211131114111511161117111811191120112111221123112411251126112711281129113011311132113311341135113611371138113911401141114211431144114511461147114811491150115111521153115411551156115711581159116011611162116311641165116611671168116911701171117211731174117511761177117811791180118111821183118411851186118711881189119011911192119311941195119611971198119912001201120212031204120512061207120812091210121112121213121412151216121712181219122012211222122312241225122612271228122912301231123212331234123512361237123812391240124112421243124412451246124712481249125012511252125312541255125612571258125912601261126212631264126512661267126812691270127112721273127412751276127712781279128012811282128312841285128612871288128912901291129212931294129512961297129812991300130113021303130413051306130713081309131013111312131313141315131613171318131913201321132213231324132513261327132813291330133113321333133413351336133713381339134013411342134313441345134613471348134913501351135213531354135513561357135813591360136113621363136413651366136713681369137013711372137313741375137613771378137913801381138213831384138513861387138813891390139113921393139413951396139713981399140014011402140314041405140614071408140914101411141214131414141514161417141814191420142114221423142414251426142714281429143014311432143314341435143614371438143914401441144214431444144514461447144814491450145114521453145414551456145714581459146014611462146314641465146614671468146914701471147214731474147514761477147814791480148114821483148414851486148714881489149014911492149314941495149614971498149915001501150215031504150515061507150815091510151115121513151415151516151715181519152015211522152315241525152615271528152915301531153215331534153515361537153815391540154115421543154415451546154715481549155015511552155315541555155615571558155915601561156215631564156515661567156815691570157115721573157415751576157715781579158015811582158315841585158615871588158915901591159215931594159515961597159815991600160116021603160416051606160716081609161016111612161316141615161616171618161916201621162216231624162516261627162816291630163116321633163416351636163716381639164016411642164316441645164616471648164916501651165216531654165516561657165816591660166116621663166416651666166716681669167016711672167316741675167616771678167916801681168216831684168516861687168816891690169116921693169416951696169716981699170017011702170317041705170617071708170917101711171217131714171517161717171817191720172117221723172417251726172717281729173017311732173317341735173617371738173917401741174217431744174517461747174817491750175117521753175417551756175717581759176017611762176317641765176617671768176917701771177217731774177517761777177817791780178117821783178417851786178717881789179017911792179317941795179617971798179918001801180218031804180518061807180818091810181118121813181418151816181718181819182018211822182318241825182618271828182918301831183218331834183518361837183818391840184118421843184418451846184718481849185018511852185318541855185618571858185918601861186218631864186518661867186818691870187118721873187418751876187718781879188018811882188318841885188618871888188918901891189218931894189518961897189818991900190119021903190419051906190719081909191019111912191319141915191619171918191919201921192219231924192519261927192819291930193119321933193419351936193719381939194019411942194319441945194619471948194919501951195219531954195519561957195819591960196119621963196419651966196719681969197019711972197319741975197619771978197919801981198219831984198519861987198819891990199119921993199419951996199719981999200020012002200320042005200620072008200920102011201220132014201520162017201820192020202120222023202420252026202720282029203020312032203320342035203620372038203920402041204220432044204520462047204820492050205120522053205420552056205720582059206020612062206320642065206620672068206920702071207220732074207520762077207820792080208120822083208420852086208720882089209020912092209320942095209620972098209921002101210221032104210521062107210821092110211121122113211421152116211721182119212021212122212321242125212621272128212921302131213221332134213521362137213821392140214121422143214421452146214721482149215021512152215321542155215621572158215921602161216221632164216521662167216821692170217121722173217421752176217721782179218021812182218321842185218621872188218921902191219221932194219521962197219821992200220122022203220422052206220722082209221022112212221322142215221622172218221922202221222222232224222522262227222822292230223122322233223422352236223722382239224022412242224322442245224622472248224922502251225222532254225522562257225822592260226122622263226422652266226722682269227022712272227322742275227622772278227922802281228222832284228522862287228822892290229122922293229422952296229722982299230023012302230323042305230623072308230923102311231223132314231523162317231823192320232123222323232423252326232723282329233023312332233323342335233623372338233923402341234223432344234523462347234823492350235123522353235423552356235723582359236023612362236323642365236623672368236923702371237223732374237523762377237823792380238123822383238423852386238723882389239023912392239323942395239623972398239924002401240224032404240524062407240824092410241124122413241424152416241724182419242024212422242324242425242624272428242924302431243224332434243524362437243824392440244124422443244424452446244724482449245024512452245324542455245624572458245924602461246224632464246524662467246824692470247124722473247424752476247724782479248024812482248324842485248624872488248924902491249224932494249524962497249824992500250125022503250425052506250725082509251025112512251325142515251625172518251925202521252225232524252525262527252825292530253125322533253425352536253725382539254025412542254325442545254625472548254925502551255225532554255525562557255825592560256125622563256425652566256725682569257025712572257325742575257625772578257925802581258225832584258525862587258825892590259125922593259425952596259725982599260026012602260326042605260626072608260926102611261226132614261526162617261826192620262126222623262426252626262726282629263026312632263326342635263626372638263926402641264226432644264526462647264826492650265126522653265426552656265726582659266026612662266326642665266626672668266926702671267226732674267526762677267826792680268126822683268426852686268726882689269026912692269326942695269626972698269927002701270227032704270527062707270827092710271127122713271427152716271727182719272027212722272327242725272627272728272927302731273227332734273527362737273827392740274127422743274427452746274727482749275027512752275327542755275627572758275927602761276227632764276527662767276827692770277127722773277427752776277727782779278027812782278327842785278627872788278927902791279227932794279527962797279827992800280128022803280428052806280728082809281028112812281328142815281628172818281928202821282228232824282528262827282828292830283128322833283428352836283728382839284028412842284328442845284628472848284928502851285228532854285528562857285828592860286128622863286428652866286728682869287028712872287328742875287628772878287928802881288228832884288528862887288828892890289128922893289428952896289728982899290029012902290329042905290629072908290929102911291229132914291529162917291829192920292129222923292429252926292729282929293029312932293329342935293629372938293929402941294229432944294529462947294829492950295129522953295429552956295729582959296029612962296329642965296629672968296929702971297229732974297529762977297829792980298129822983298429852986298729882989299029912992299329942995299629972998299930003001300230033004300530063007300830093010301130123013301430153016301730183019302030213022302330243025302630273028302930303031303230333034303530363037303830393040304130423043304430453046304730483049305030513052305330543055305630573058305930603061306230633064306530663067306830693070307130723073307430753076307730783079308030813082308330843085308630873088308930903091309230933094309530963097309830993100310131023103310431053106310731083109311031113112311331143115311631173118311931203121312231233124312531263127312831293130313131323133313431353136313731383139314031413142314331443145314631473148314931503151315231533154315531563157315831593160316131623163316431653166316731683169317031713172317331743175317631773178317931803181318231833184318531863187318831893190319131923193319431953196319731983199320032013202320332043205320632073208320932103211321232133214321532163217321832193220322132223223322432253226322732283229323032313232323332343235323632373238323932403241324232433244324532463247324832493250325132523253325432553256325732583259326032613262326332643265326632673268326932703271327232733274327532763277327832793280328132823283328432853286328732883289329032913292329332943295329632973298329933003301330233033304330533063307330833093310331133123313331433153316331733183319332033213322332333243325332633273328332933303331333233333334333533363337333833393340334133423343334433453346334733483349335033513352335333543355335633573358335933603361336233633364336533663367336833693370337133723373337433753376337733783379338033813382338333843385338633873388338933903391339233933394339533963397339833993400340134023403340434053406340734083409341034113412341334143415341634173418341934203421342234233424342534263427342834293430343134323433343434353436343734383439344034413442344334443445344634473448344934503451345234533454345534563457345834593460346134623463346434653466346734683469347034713472347334743475347634773478347934803481348234833484348534863487348834893490349134923493349434953496349734983499350035013502350335043505350635073508350935103511351235133514351535163517351835193520352135223523352435253526352735283529353035313532353335343535353635373538353935403541354235433544354535463547354835493550355135523553355435553556355735583559356035613562356335643565356635673568356935703571357235733574357535763577357835793580358135823583358435853586358735883589359035913592359335943595359635973598359936003601360236033604360536063607360836093610361136123613361436153616361736183619362036213622362336243625362636273628362936303631363236333634363536363637363836393640364136423643364436453646364736483649365036513652365336543655365636573658365936603661366236633664366536663667366836693670367136723673367436753676367736783679368036813682368336843685368636873688368936903691369236933694369536963697369836993700370137023703370437053706370737083709371037113712371337143715371637173718371937203721372237233724372537263727372837293730373137323733373437353736373737383739374037413742374337443745374637473748374937503751375237533754375537563757375837593760376137623763376437653766376737683769377037713772377337743775377637773778377937803781378237833784378537863787378837893790379137923793379437953796379737983799380038013802380338043805380638073808380938103811381238133814381538163817381838193820382138223823382438253826382738283829383038313832383338343835383638373838383938403841384238433844384538463847384838493850385138523853385438553856385738583859386038613862386338643865386638673868386938703871387238733874387538763877387838793880388138823883388438853886388738883889389038913892389338943895389638973898389939003901390239033904390539063907390839093910391139123913391439153916391739183919392039213922392339243925392639273928392939303931393239333934393539363937393839393940394139423943394439453946394739483949395039513952395339543955395639573958395939603961396239633964396539663967396839693970397139723973397439753976397739783979398039813982398339843985398639873988398939903991399239933994399539963997399839994000400140024003400440054006400740084009401040114012401340144015401640174018401940204021402240234024402540264027402840294030403140324033403440354036403740384039404040414042404340444045404640474048404940504051405240534054405540564057405840594060406140624063406440654066406740684069407040714072407340744075407640774078407940804081408240834084408540864087408840894090409140924093409440954096409740984099410041014102410341044105410641074108410941104111411241134114411541164117411841194120412141224123412441254126412741284129413041314132413341344135413641374138413941404141414241434144414541464147414841494150415141524153415441554156415741584159416041614162416341644165416641674168416941704171417241734174417541764177417841794180418141824183418441854186418741884189419041914192419341944195419641974198419942004201420242034204420542064207420842094210421142124213421442154216421742184219422042214222422342244225422642274228422942304231423242334234423542364237423842394240424142424243424442454246424742484249425042514252425342544255425642574258425942604261426242634264426542664267426842694270427142724273427442754276427742784279428042814282428342844285428642874288428942904291429242934294429542964297429842994300430143024303430443054306430743084309431043114312431343144315431643174318431943204321432243234324432543264327432843294330433143324333433443354336433743384339434043414342434343444345434643474348434943504351435243534354435543564357435843594360436143624363436443654366436743684369437043714372437343744375437643774378437943804381438243834384438543864387438843894390439143924393439443954396439743984399440044014402440344044405440644074408440944104411441244134414441544164417441844194420442144224423442444254426442744284429443044314432443344344435443644374438443944404441444244434444444544464447444844494450445144524453445444554456445744584459446044614462446344644465446644674468446944704471447244734474447544764477447844794480448144824483448444854486448744884489449044914492449344944495449644974498449945004501450245034504450545064507450845094510451145124513451445154516451745184519452045214522452345244525452645274528452945304531453245334534453545364537453845394540454145424543454445454546454745484549455045514552455345544555455645574558455945604561456245634564456545664567456845694570457145724573457445754576457745784579458045814582458345844585458645874588458945904591459245934594459545964597459845994600460146024603460446054606460746084609461046114612461346144615461646174618461946204621462246234624462546264627462846294630463146324633463446354636463746384639464046414642464346444645464646474648464946504651465246534654465546564657465846594660466146624663466446654666466746684669467046714672467346744675467646774678467946804681468246834684468546864687468846894690469146924693469446954696469746984699470047014702470347044705470647074708470947104711471247134714471547164717471847194720472147224723472447254726472747284729473047314732473347344735473647374738473947404741474247434744474547464747474847494750475147524753475447554756475747584759476047614762476347644765476647674768476947704771477247734774477547764777477847794780478147824783478447854786478747884789479047914792479347944795479647974798479948004801480248034804480548064807480848094810481148124813481448154816481748184819482048214822482348244825482648274828482948304831483248334834483548364837483848394840484148424843484448454846484748484849485048514852485348544855485648574858485948604861486248634864486548664867486848694870487148724873487448754876487748784879488048814882488348844885488648874888488948904891489248934894489548964897489848994900490149024903490449054906490749084909491049114912491349144915491649174918491949204921492249234924492549264927492849294930493149324933493449354936493749384939494049414942494349444945494649474948494949504951495249534954495549564957495849594960496149624963496449654966496749684969497049714972497349744975497649774978497949804981498249834984498549864987498849894990499149924993499449954996499749984999500050015002500350045005500650075008500950105011501250135014501550165017501850195020502150225023502450255026502750285029503050315032503350345035503650375038503950405041504250435044504550465047504850495050505150525053505450555056505750585059506050615062506350645065506650675068506950705071507250735074507550765077507850795080508150825083508450855086508750885089509050915092509350945095509650975098509951005101510251035104510551065107510851095110511151125113511451155116511751185119512051215122512351245125512651275128512951305131513251335134513551365137513851395140514151425143514451455146514751485149515051515152515351545155515651575158515951605161516251635164516551665167516851695170517151725173517451755176517751785179518051815182518351845185518651875188518951905191519251935194519551965197519851995200520152025203520452055206520752085209521052115212521352145215521652175218521952205221522252235224522552265227522852295230523152325233523452355236523752385239524052415242524352445245524652475248524952505251525252535254525552565257525852595260526152625263526452655266526752685269527052715272527352745275527652775278527952805281528252835284528552865287528852895290529152925293529452955296529752985299530053015302530353045305530653075308530953105311531253135314531553165317531853195320532153225323532453255326532753285329533053315332533353345335533653375338533953405341534253435344534553465347534853495350535153525353535453555356535753585359536053615362536353645365536653675368536953705371537253735374537553765377537853795380538153825383538453855386538753885389539053915392539353945395539653975398539954005401540254035404540554065407540854095410541154125413541454155416541754185419542054215422542354245425542654275428542954305431543254335434543554365437543854395440544154425443544454455446544754485449545054515452545354545455545654575458545954605461546254635464546554665467546854695470547154725473547454755476547754785479548054815482548354845485548654875488548954905491549254935494549554965497549854995500550155025503550455055506550755085509551055115512551355145515551655175518551955205521552255235524552555265527552855295530553155325533553455355536553755385539554055415542554355445545554655475548554955505551555255535554555555565557555855595560556155625563556455655566556755685569557055715572557355745575557655775578557955805581558255835584558555865587558855895590559155925593559455955596559755985599560056015602560356045605560656075608560956105611561256135614561556165617561856195620562156225623562456255626562756285629563056315632563356345635563656375638563956405641564256435644564556465647564856495650565156525653565456555656565756585659566056615662566356645665566656675668566956705671567256735674567556765677567856795680568156825683568456855686568756885689569056915692569356945695569656975698569957005701570257035704570557065707570857095710571157125713571457155716571757185719572057215722572357245725572657275728572957305731573257335734573557365737573857395740574157425743574457455746574757485749575057515752575357545755575657575758575957605761576257635764576557665767576857695770577157725773577457755776577757785779578057815782578357845785578657875788578957905791579257935794579557965797579857995800580158025803580458055806580758085809581058115812581358145815581658175818581958205821582258235824582558265827582858295830583158325833583458355836583758385839584058415842584358445845584658475848584958505851585258535854585558565857585858595860586158625863586458655866586758685869587058715872587358745875587658775878587958805881588258835884588558865887588858895890589158925893589458955896589758985899590059015902590359045905590659075908590959105911591259135914591559165917591859195920592159225923592459255926592759285929593059315932593359345935593659375938593959405941594259435944594559465947594859495950595159525953595459555956595759585959596059615962596359645965596659675968596959705971597259735974597559765977597859795980598159825983598459855986598759885989599059915992599359945995599659975998599960006001600260036004600560066007600860096010601160126013601460156016601760186019602060216022602360246025602660276028602960306031603260336034603560366037603860396040604160426043604460456046604760486049605060516052605360546055605660576058605960606061606260636064606560666067606860696070607160726073607460756076607760786079608060816082608360846085608660876088608960906091609260936094609560966097609860996100610161026103610461056106610761086109611061116112611361146115611661176118611961206121612261236124612561266127612861296130613161326133613461356136613761386139614061416142614361446145614661476148614961506151615261536154615561566157615861596160616161626163616461656166616761686169617061716172617361746175617661776178617961806181618261836184618561866187618861896190619161926193619461956196619761986199620062016202620362046205620662076208620962106211621262136214621562166217621862196220622162226223622462256226622762286229623062316232623362346235623662376238623962406241624262436244624562466247624862496250625162526253625462556256625762586259626062616262626362646265626662676268626962706271627262736274627562766277627862796280628162826283628462856286628762886289629062916292629362946295629662976298629963006301630263036304630563066307630863096310631163126313631463156316631763186319632063216322632363246325632663276328632963306331633263336334633563366337633863396340634163426343634463456346634763486349635063516352635363546355635663576358635963606361636263636364636563666367636863696370637163726373637463756376637763786379638063816382638363846385638663876388638963906391639263936394639563966397639863996400640164026403640464056406640764086409641064116412641364146415641664176418641964206421642264236424642564266427642864296430643164326433643464356436643764386439644064416442644364446445644664476448644964506451645264536454645564566457645864596460646164626463646464656466646764686469647064716472647364746475647664776478647964806481648264836484648564866487648864896490649164926493649464956496649764986499650065016502650365046505650665076508650965106511651265136514651565166517651865196520652165226523652465256526652765286529653065316532653365346535653665376538653965406541654265436544654565466547654865496550655165526553655465556556655765586559656065616562656365646565656665676568656965706571657265736574657565766577657865796580658165826583658465856586658765886589659065916592659365946595659665976598659966006601660266036604660566066607660866096610661166126613661466156616661766186619662066216622662366246625662666276628662966306631663266336634663566366637663866396640664166426643664466456646664766486649665066516652665366546655665666576658665966606661666266636664666566666667666866696670667166726673667466756676667766786679668066816682668366846685668666876688668966906691669266936694669566966697669866996700670167026703670467056706670767086709671067116712671367146715671667176718671967206721672267236724672567266727672867296730673167326733673467356736673767386739674067416742674367446745674667476748674967506751675267536754675567566757675867596760676167626763676467656766676767686769677067716772677367746775677667776778677967806781678267836784678567866787678867896790679167926793679467956796679767986799680068016802680368046805680668076808680968106811681268136814681568166817681868196820682168226823682468256826682768286829683068316832683368346835683668376838683968406841684268436844684568466847684868496850685168526853685468556856685768586859686068616862686368646865686668676868686968706871687268736874687568766877687868796880688168826883688468856886688768886889689068916892689368946895689668976898689969006901690269036904690569066907690869096910691169126913691469156916691769186919692069216922692369246925692669276928692969306931693269336934693569366937693869396940694169426943694469456946694769486949695069516952695369546955695669576958695969606961696269636964696569666967696869696970697169726973697469756976697769786979698069816982698369846985698669876988698969906991699269936994699569966997699869997000700170027003700470057006700770087009701070117012701370147015701670177018701970207021702270237024702570267027702870297030703170327033703470357036703770387039704070417042704370447045704670477048704970507051705270537054705570567057705870597060706170627063706470657066706770687069707070717072707370747075707670777078707970807081708270837084708570867087708870897090709170927093709470957096709770987099710071017102710371047105710671077108710971107111711271137114711571167117711871197120712171227123712471257126712771287129713071317132713371347135713671377138713971407141714271437144714571467147714871497150715171527153715471557156715771587159716071617162716371647165716671677168716971707171717271737174717571767177717871797180718171827183718471857186718771887189719071917192719371947195719671977198719972007201720272037204720572067207720872097210721172127213721472157216721772187219722072217222722372247225722672277228722972307231723272337234723572367237723872397240724172427243724472457246724772487249725072517252725372547255725672577258725972607261726272637264726572667267726872697270727172727273727472757276727772787279728072817282728372847285728672877288728972907291729272937294729572967297729872997300730173027303730473057306730773087309731073117312731373147315731673177318731973207321732273237324732573267327732873297330733173327333733473357336733773387339734073417342734373447345734673477348734973507351735273537354735573567357735873597360736173627363736473657366736773687369737073717372737373747375737673777378737973807381738273837384738573867387738873897390739173927393739473957396739773987399740074017402740374047405740674077408740974107411741274137414741574167417741874197420742174227423742474257426742774287429743074317432743374347435743674377438743974407441744274437444744574467447744874497450745174527453745474557456745774587459746074617462746374647465746674677468746974707471747274737474747574767477747874797480748174827483748474857486748774887489749074917492749374947495749674977498749975007501750275037504750575067507750875097510751175127513751475157516751775187519752075217522752375247525752675277528752975307531753275337534753575367537753875397540754175427543754475457546754775487549755075517552755375547555755675577558755975607561756275637564756575667567756875697570757175727573757475757576757775787579758075817582758375847585758675877588758975907591759275937594759575967597759875997600760176027603760476057606760776087609761076117612761376147615761676177618761976207621762276237624762576267627762876297630763176327633763476357636763776387639764076417642764376447645764676477648764976507651765276537654765576567657765876597660766176627663766476657666766776687669767076717672767376747675767676777678767976807681768276837684768576867687768876897690769176927693769476957696769776987699770077017702770377047705770677077708770977107711771277137714771577167717771877197720772177227723772477257726772777287729773077317732773377347735773677377738773977407741774277437744774577467747774877497750775177527753775477557756775777587759776077617762776377647765776677677768776977707771777277737774777577767777777877797780778177827783778477857786778777887789779077917792779377947795779677977798779978007801780278037804780578067807780878097810781178127813781478157816781778187819782078217822782378247825782678277828782978307831783278337834783578367837783878397840784178427843784478457846784778487849785078517852785378547855785678577858785978607861786278637864786578667867786878697870787178727873787478757876787778787879788078817882788378847885788678877888788978907891789278937894789578967897789878997900790179027903790479057906790779087909791079117912791379147915791679177918791979207921792279237924792579267927792879297930793179327933793479357936793779387939794079417942794379447945794679477948794979507951795279537954795579567957795879597960796179627963796479657966796779687969797079717972797379747975797679777978797979807981798279837984798579867987798879897990799179927993799479957996799779987999800080018002800380048005800680078008800980108011801280138014801580168017801880198020802180228023802480258026802780288029803080318032803380348035803680378038803980408041804280438044804580468047804880498050805180528053805480558056805780588059806080618062806380648065806680678068806980708071807280738074807580768077807880798080808180828083808480858086808780888089809080918092809380948095809680978098809981008101810281038104810581068107810881098110811181128113811481158116811781188119812081218122812381248125812681278128812981308131813281338134813581368137813881398140814181428143814481458146814781488149815081518152815381548155815681578158815981608161816281638164816581668167816881698170817181728173817481758176817781788179818081818182818381848185818681878188818981908191819281938194819581968197819881998200820182028203820482058206820782088209821082118212821382148215821682178218821982208221822282238224822582268227822882298230823182328233823482358236823782388239824082418242824382448245824682478248824982508251825282538254825582568257825882598260826182628263826482658266826782688269827082718272827382748275827682778278827982808281828282838284828582868287828882898290829182928293829482958296829782988299830083018302830383048305830683078308830983108311831283138314831583168317831883198320832183228323832483258326832783288329833083318332833383348335833683378338833983408341834283438344834583468347834883498350835183528353835483558356835783588359836083618362836383648365836683678368836983708371837283738374837583768377837883798380838183828383838483858386838783888389839083918392839383948395839683978398839984008401840284038404840584068407840884098410841184128413841484158416841784188419842084218422842384248425842684278428842984308431843284338434843584368437843884398440844184428443844484458446844784488449845084518452845384548455845684578458845984608461846284638464846584668467846884698470847184728473847484758476847784788479848084818482848384848485848684878488848984908491849284938494849584968497849884998500850185028503850485058506850785088509851085118512851385148515851685178518851985208521852285238524852585268527852885298530853185328533853485358536853785388539854085418542854385448545854685478548854985508551855285538554855585568557855885598560856185628563856485658566856785688569857085718572857385748575857685778578857985808581858285838584858585868587858885898590859185928593859485958596859785988599860086018602860386048605860686078608860986108611861286138614861586168617861886198620862186228623862486258626862786288629863086318632863386348635863686378638863986408641864286438644864586468647864886498650865186528653865486558656865786588659866086618662866386648665866686678668866986708671867286738674867586768677867886798680868186828683868486858686868786888689869086918692869386948695869686978698869987008701870287038704870587068707870887098710871187128713871487158716871787188719872087218722872387248725872687278728872987308731873287338734873587368737873887398740874187428743874487458746874787488749875087518752875387548755875687578758875987608761876287638764876587668767876887698770877187728773877487758776877787788779878087818782878387848785878687878788878987908791879287938794879587968797879887998800880188028803880488058806880788088809881088118812881388148815881688178818881988208821882288238824882588268827882888298830883188328833883488358836883788388839884088418842884388448845884688478848884988508851885288538854885588568857885888598860886188628863886488658866886788688869887088718872887388748875887688778878887988808881888288838884888588868887888888898890889188928893889488958896889788988899890089018902890389048905890689078908890989108911891289138914891589168917891889198920892189228923892489258926892789288929893089318932893389348935893689378938893989408941894289438944894589468947894889498950895189528953895489558956895789588959896089618962896389648965896689678968896989708971897289738974897589768977897889798980898189828983898489858986898789888989899089918992899389948995899689978998899990009001900290039004900590069007900890099010901190129013901490159016901790189019902090219022902390249025902690279028902990309031903290339034903590369037903890399040904190429043904490459046904790489049905090519052905390549055905690579058905990609061906290639064906590669067906890699070907190729073907490759076907790789079908090819082908390849085908690879088908990909091909290939094909590969097909890999100910191029103910491059106910791089109911091119112911391149115911691179118911991209121912291239124912591269127912891299130913191329133913491359136913791389139914091419142914391449145914691479148914991509151915291539154915591569157915891599160916191629163916491659166916791689169917091719172917391749175917691779178917991809181918291839184918591869187918891899190919191929193919491959196919791989199920092019202920392049205920692079208920992109211921292139214921592169217921892199220922192229223922492259226922792289229923092319232923392349235923692379238923992409241924292439244924592469247924892499250925192529253925492559256925792589259926092619262926392649265926692679268926992709271927292739274927592769277927892799280928192829283928492859286928792889289929092919292929392949295929692979298929993009301930293039304930593069307930893099310931193129313931493159316931793189319932093219322932393249325932693279328932993309331933293339334933593369337933893399340934193429343934493459346934793489349935093519352935393549355935693579358935993609361936293639364936593669367936893699370937193729373937493759376937793789379938093819382938393849385938693879388938993909391939293939394939593969397939893999400940194029403940494059406940794089409941094119412941394149415941694179418941994209421942294239424942594269427942894299430943194329433943494359436943794389439944094419442944394449445944694479448944994509451945294539454945594569457945894599460946194629463946494659466946794689469947094719472947394749475947694779478947994809481948294839484948594869487948894899490949194929493949494959496949794989499950095019502950395049505950695079508950995109511951295139514951595169517951895199520952195229523952495259526952795289529953095319532953395349535953695379538953995409541954295439544954595469547954895499550955195529553955495559556955795589559956095619562956395649565956695679568956995709571957295739574957595769577957895799580958195829583958495859586958795889589959095919592959395949595959695979598959996009601960296039604960596069607960896099610961196129613961496159616961796189619962096219622962396249625962696279628962996309631963296339634963596369637963896399640964196429643964496459646964796489649965096519652965396549655965696579658965996609661966296639664966596669667966896699670967196729673967496759676967796789679968096819682968396849685968696879688968996909691969296939694969596969697969896999700970197029703970497059706970797089709971097119712971397149715971697179718971997209721972297239724972597269727972897299730973197329733973497359736973797389739974097419742974397449745974697479748974997509751975297539754975597569757975897599760976197629763976497659766976797689769977097719772977397749775977697779778977997809781978297839784978597869787978897899790979197929793979497959796979797989799980098019802980398049805980698079808980998109811981298139814981598169817981898199820982198229823982498259826982798289829983098319832983398349835983698379838983998409841984298439844984598469847984898499850985198529853985498559856985798589859986098619862986398649865986698679868986998709871987298739874987598769877987898799880988198829883988498859886988798889889989098919892989398949895989698979898989999009901990299039904990599069907990899099910991199129913991499159916991799189919992099219922992399249925992699279928992999309931993299339934993599369937993899399940994199429943994499459946994799489949995099519952995399549955995699579958995999609961996299639964996599669967996899699970997199729973997499759976997799789979998099819982998399849985998699879988998999909991999299939994999599969997999899991000010001100021000310004100051000610007100081000910010100111001210013100141001510016100171001810019100201002110022100231002410025100261002710028100291003010031100321003310034100351003610037100381003910040100411004210043100441004510046100471004810049100501005110052100531005410055100561005710058100591006010061100621006310064100651006610067100681006910070100711007210073100741007510076100771007810079100801008110082100831008410085100861008710088100891009010091100921009310094100951009610097100981009910100101011010210103101041010510106101071010810109101101011110112101131011410115101161011710118101191012010121101221012310124101251012610127101281012910130101311013210133101341013510136101371013810139101401014110142101431014410145101461014710148101491015010151101521015310154101551015610157101581015910160101611016210163101641016510166101671016810169101701017110172101731017410175101761017710178101791018010181101821018310184101851018610187101881018910190101911019210193101941019510196101971019810199102001020110202102031020410205102061020710208102091021010211102121021310214102151021610217102181021910220102211022210223102241022510226102271022810229102301023110232102331023410235102361023710238102391024010241102421024310244102451024610247102481024910250102511025210253102541025510256102571025810259102601026110262102631026410265102661026710268102691027010271102721027310274102751027610277102781027910280102811028210283102841028510286102871028810289102901029110292102931029410295102961029710298102991030010301103021030310304103051030610307103081030910310103111031210313103141031510316103171031810319103201032110322103231032410325103261032710328103291033010331103321033310334103351033610337103381033910340103411034210343103441034510346103471034810349103501035110352103531035410355103561035710358103591036010361103621036310364103651036610367103681036910370103711037210373103741037510376103771037810379103801038110382103831038410385103861038710388103891039010391103921039310394103951039610397103981039910400104011040210403104041040510406104071040810409104101041110412104131041410415104161041710418104191042010421104221042310424104251042610427104281042910430104311043210433104341043510436104371043810439104401044110442104431044410445104461044710448104491045010451104521045310454104551045610457104581045910460104611046210463104641046510466104671046810469104701047110472104731047410475104761047710478104791048010481104821048310484104851048610487104881048910490104911049210493104941049510496104971049810499105001050110502105031050410505105061050710508105091051010511105121051310514105151051610517105181051910520105211052210523105241052510526105271052810529105301053110532105331053410535105361053710538105391054010541105421054310544105451054610547105481054910550105511055210553105541055510556105571055810559105601056110562105631056410565105661056710568105691057010571105721057310574105751057610577105781057910580105811058210583105841058510586105871058810589105901059110592105931059410595105961059710598105991060010601106021060310604106051060610607106081060910610106111061210613106141061510616106171061810619106201062110622106231062410625106261062710628106291063010631106321063310634106351063610637106381063910640106411064210643106441064510646106471064810649106501065110652106531065410655106561065710658106591066010661106621066310664106651066610667106681066910670106711067210673106741067510676106771067810679106801068110682106831068410685106861068710688106891069010691106921069310694106951069610697106981069910700107011070210703107041070510706107071070810709107101071110712107131071410715107161071710718107191072010721107221072310724107251072610727107281072910730107311073210733107341073510736107371073810739107401074110742107431074410745107461074710748107491075010751107521075310754107551075610757107581075910760107611076210763107641076510766107671076810769107701077110772107731077410775107761077710778107791078010781107821078310784107851078610787107881078910790107911079210793107941079510796107971079810799108001080110802108031080410805108061080710808108091081010811108121081310814108151081610817108181081910820108211082210823108241082510826108271082810829108301083110832108331083410835108361083710838108391084010841108421084310844108451084610847108481084910850108511085210853108541085510856108571085810859108601086110862108631086410865108661086710868108691087010871108721087310874108751087610877108781087910880108811088210883108841088510886108871088810889108901089110892108931089410895108961089710898108991090010901109021090310904109051090610907109081090910910109111091210913109141091510916109171091810919109201092110922109231092410925109261092710928109291093010931109321093310934109351093610937109381093910940109411094210943109441094510946109471094810949109501095110952109531095410955109561095710958109591096010961109621096310964109651096610967109681096910970109711097210973109741097510976109771097810979109801098110982109831098410985109861098710988109891099010991109921099310994109951099610997109981099911000110011100211003110041100511006110071100811009110101101111012110131101411015110161101711018110191102011021110221102311024110251102611027110281102911030110311103211033110341103511036110371103811039110401104111042110431104411045110461104711048110491105011051110521105311054110551105611057110581105911060110611106211063110641106511066110671106811069110701107111072110731107411075110761107711078110791108011081110821108311084110851108611087110881108911090110911109211093110941109511096110971109811099111001110111102111031110411105111061110711108111091111011111111121111311114111151111611117111181111911120111211112211123111241112511126111271112811129111301113111132111331113411135111361113711138111391114011141111421114311144111451114611147111481114911150111511115211153111541115511156111571115811159111601116111162111631116411165111661116711168111691117011171111721117311174111751117611177111781117911180111811118211183111841118511186111871118811189111901119111192111931119411195111961119711198111991120011201112021120311204112051120611207112081120911210112111121211213112141121511216112171121811219112201122111222112231122411225112261122711228112291123011231112321123311234112351123611237112381123911240112411124211243112441124511246112471124811249112501125111252112531125411255112561125711258112591126011261112621126311264112651126611267112681126911270112711127211273112741127511276112771127811279112801128111282112831128411285112861128711288112891129011291112921129311294112951129611297112981129911300113011130211303113041130511306113071130811309113101131111312113131131411315113161131711318113191132011321113221132311324113251132611327113281132911330113311133211333113341133511336113371133811339113401134111342113431134411345113461134711348113491135011351113521135311354113551135611357113581135911360113611136211363113641136511366113671136811369113701137111372113731137411375113761137711378113791138011381113821138311384113851138611387113881138911390113911139211393113941139511396113971139811399114001140111402114031140411405114061140711408114091141011411114121141311414114151141611417114181141911420114211142211423114241142511426114271142811429114301143111432114331143411435114361143711438114391144011441114421144311444114451144611447114481144911450114511145211453114541145511456114571145811459114601146111462114631146411465114661146711468114691147011471114721147311474114751147611477114781147911480114811148211483114841148511486114871148811489114901149111492114931149411495114961149711498114991150011501115021150311504115051150611507115081150911510115111151211513115141151511516115171151811519115201152111522115231152411525115261152711528115291153011531115321153311534115351153611537115381153911540115411154211543115441154511546115471154811549115501155111552115531155411555115561155711558115591156011561115621156311564115651156611567115681156911570115711157211573115741157511576115771157811579115801158111582115831158411585115861158711588115891159011591115921159311594115951159611597115981159911600116011160211603116041160511606116071160811609116101161111612116131161411615116161161711618116191162011621116221162311624116251162611627116281162911630116311163211633116341163511636116371163811639116401164111642116431164411645116461164711648116491165011651116521165311654116551165611657116581165911660116611166211663116641166511666116671166811669116701167111672116731167411675116761167711678116791168011681116821168311684116851168611687116881168911690116911169211693116941169511696116971169811699117001170111702117031170411705117061170711708117091171011711117121171311714117151171611717117181171911720117211172211723117241172511726117271172811729117301173111732117331173411735117361173711738117391174011741117421174311744117451174611747117481174911750117511175211753117541175511756117571175811759117601176111762117631176411765117661176711768117691177011771117721177311774117751177611777117781177911780117811178211783117841178511786117871178811789117901179111792117931179411795117961179711798117991180011801118021180311804118051180611807118081180911810118111181211813118141181511816118171181811819118201182111822118231182411825118261182711828118291183011831118321183311834118351183611837118381183911840118411184211843118441184511846118471184811849118501185111852118531185411855118561185711858118591186011861118621186311864118651186611867118681186911870118711187211873118741187511876118771187811879118801188111882118831188411885118861188711888118891189011891118921189311894118951189611897118981189911900119011190211903119041190511906119071190811909119101191111912119131191411915119161191711918119191192011921119221192311924119251192611927119281192911930119311193211933119341193511936119371193811939119401194111942119431194411945119461194711948119491195011951119521195311954119551195611957119581195911960119611196211963119641196511966119671196811969119701197111972119731197411975119761197711978119791198011981119821198311984119851198611987119881198911990119911199211993119941199511996119971199811999120001200112002120031200412005120061200712008120091201012011120121201312014120151201612017120181201912020120211202212023120241202512026120271202812029120301203112032120331203412035120361203712038120391204012041120421204312044120451204612047120481204912050120511205212053120541205512056120571205812059120601206112062120631206412065120661206712068120691207012071120721207312074120751207612077120781207912080120811208212083120841208512086120871208812089120901209112092120931209412095120961209712098120991210012101121021210312104121051210612107121081210912110121111211212113121141211512116121171211812119121201212112122121231212412125121261212712128121291213012131121321213312134121351213612137121381213912140121411214212143121441214512146121471214812149121501215112152121531215412155121561215712158121591216012161121621216312164121651216612167121681216912170121711217212173121741217512176121771217812179121801218112182121831218412185121861218712188121891219012191121921219312194121951219612197121981219912200122011220212203122041220512206122071220812209122101221112212122131221412215122161221712218122191222012221122221222312224122251222612227122281222912230122311223212233122341223512236122371223812239122401224112242122431224412245122461224712248122491225012251122521225312254122551225612257122581225912260122611226212263122641226512266122671226812269122701227112272122731227412275122761227712278122791228012281122821228312284122851228612287122881228912290122911229212293122941229512296122971229812299123001230112302123031230412305123061230712308123091231012311123121231312314123151231612317123181231912320123211232212323123241232512326123271232812329123301233112332123331233412335123361233712338123391234012341123421234312344123451234612347123481234912350123511235212353123541235512356123571235812359123601236112362123631236412365123661236712368123691237012371123721237312374123751237612377123781237912380123811238212383123841238512386123871238812389123901239112392123931239412395123961239712398123991240012401124021240312404124051240612407124081240912410124111241212413124141241512416124171241812419124201242112422124231242412425124261242712428124291243012431124321243312434124351243612437124381243912440124411244212443124441244512446124471244812449124501245112452124531245412455124561245712458124591246012461124621246312464124651246612467124681246912470124711247212473124741247512476124771247812479124801248112482124831248412485124861248712488124891249012491124921249312494124951249612497124981249912500125011250212503125041250512506125071250812509125101251112512125131251412515125161251712518125191252012521125221252312524125251252612527125281252912530125311253212533125341253512536125371253812539125401254112542125431254412545125461254712548125491255012551125521255312554125551255612557125581255912560125611256212563125641256512566125671256812569125701257112572125731257412575125761257712578125791258012581125821258312584125851258612587125881258912590125911259212593125941259512596125971259812599126001260112602126031260412605126061260712608126091261012611126121261312614126151261612617126181261912620126211262212623126241262512626126271262812629126301263112632126331263412635126361263712638126391264012641126421264312644126451264612647126481264912650126511265212653126541265512656126571265812659126601266112662126631266412665126661266712668126691267012671126721267312674126751267612677126781267912680126811268212683126841268512686126871268812689126901269112692126931269412695126961269712698126991270012701127021270312704127051270612707127081270912710127111271212713127141271512716127171271812719127201272112722127231272412725127261272712728127291273012731127321273312734127351273612737127381273912740127411274212743127441274512746127471274812749127501275112752127531275412755127561275712758127591276012761127621276312764127651276612767127681276912770127711277212773127741277512776127771277812779127801278112782127831278412785127861278712788127891279012791127921279312794127951279612797127981279912800128011280212803128041280512806128071280812809128101281112812128131281412815128161281712818128191282012821128221282312824128251282612827128281282912830128311283212833128341283512836128371283812839128401284112842128431284412845128461284712848128491285012851128521285312854128551285612857128581285912860128611286212863128641286512866128671286812869128701287112872128731287412875128761287712878128791288012881128821288312884128851288612887128881288912890128911289212893128941289512896128971289812899129001290112902129031290412905129061290712908129091291012911129121291312914129151291612917129181291912920129211292212923129241292512926129271292812929129301293112932129331293412935129361293712938129391294012941129421294312944129451294612947129481294912950129511295212953129541295512956129571295812959129601296112962129631296412965129661296712968129691297012971129721297312974129751297612977129781297912980129811298212983129841298512986129871298812989129901299112992129931299412995129961299712998129991300013001130021300313004130051300613007130081300913010130111301213013130141301513016130171301813019130201302113022130231302413025130261302713028130291303013031130321303313034130351303613037130381303913040130411304213043130441304513046130471304813049130501305113052130531305413055130561305713058130591306013061130621306313064130651306613067130681306913070130711307213073130741307513076130771307813079130801308113082130831308413085130861308713088130891309013091130921309313094130951309613097130981309913100131011310213103131041310513106131071310813109131101311113112131131311413115131161311713118131191312013121131221312313124131251312613127131281312913130131311313213133131341313513136131371313813139131401314113142131431314413145131461314713148131491315013151131521315313154131551315613157131581315913160131611316213163131641316513166131671316813169131701317113172131731317413175131761317713178131791318013181131821318313184131851318613187131881318913190131911319213193131941319513196131971319813199132001320113202132031320413205132061320713208132091321013211132121321313214132151321613217132181321913220132211322213223132241322513226132271322813229132301323113232132331323413235132361323713238132391324013241132421324313244132451324613247132481324913250132511325213253132541325513256132571325813259132601326113262132631326413265132661326713268132691327013271132721327313274132751327613277132781327913280132811328213283132841328513286132871328813289132901329113292132931329413295132961329713298132991330013301133021330313304133051330613307133081330913310133111331213313133141331513316133171331813319133201332113322133231332413325133261332713328133291333013331133321333313334133351333613337133381333913340133411334213343133441334513346133471334813349133501335113352133531335413355133561335713358133591336013361133621336313364133651336613367133681336913370133711337213373133741337513376133771337813379133801338113382133831338413385133861338713388133891339013391133921339313394133951339613397133981339913400134011340213403134041340513406134071340813409134101341113412134131341413415134161341713418134191342013421134221342313424134251342613427134281342913430134311343213433134341343513436134371343813439134401344113442134431344413445134461344713448134491345013451134521345313454134551345613457134581345913460134611346213463134641346513466134671346813469134701347113472134731347413475134761347713478134791348013481134821348313484134851348613487134881348913490134911349213493134941349513496134971349813499135001350113502135031350413505135061350713508135091351013511135121351313514135151351613517135181351913520135211352213523135241352513526135271352813529135301353113532135331353413535135361353713538135391354013541135421354313544135451354613547135481354913550135511355213553135541355513556135571355813559135601356113562135631356413565135661356713568135691357013571135721357313574135751357613577135781357913580135811358213583135841358513586135871358813589135901359113592135931359413595135961359713598135991360013601136021360313604136051360613607136081360913610136111361213613136141361513616136171361813619136201362113622136231362413625136261362713628136291363013631136321363313634136351363613637136381363913640136411364213643136441364513646136471364813649136501365113652136531365413655136561365713658136591366013661136621366313664136651366613667136681366913670136711367213673136741367513676136771367813679136801368113682136831368413685136861368713688136891369013691136921369313694136951369613697136981369913700137011370213703137041370513706137071370813709137101371113712137131371413715137161371713718137191372013721137221372313724137251372613727137281372913730137311373213733137341373513736137371373813739137401374113742137431374413745137461374713748137491375013751137521375313754137551375613757137581375913760137611376213763137641376513766137671376813769137701377113772137731377413775137761377713778137791378013781137821378313784137851378613787137881378913790137911379213793137941379513796137971379813799138001380113802138031380413805138061380713808138091381013811138121381313814138151381613817138181381913820138211382213823138241382513826138271382813829138301383113832138331383413835138361383713838138391384013841138421384313844138451384613847138481384913850138511385213853138541385513856138571385813859138601386113862138631386413865138661386713868138691387013871138721387313874138751387613877138781387913880138811388213883138841388513886138871388813889138901389113892138931389413895138961389713898138991390013901139021390313904139051390613907139081390913910139111391213913139141391513916139171391813919139201392113922139231392413925139261392713928139291393013931139321393313934139351393613937139381393913940139411394213943139441394513946139471394813949139501395113952139531395413955139561395713958139591396013961139621396313964139651396613967139681396913970139711397213973139741397513976139771397813979139801398113982139831398413985139861398713988139891399013991139921399313994139951399613997139981399914000140011400214003140041400514006140071400814009140101401114012140131401414015140161401714018140191402014021140221402314024140251402614027140281402914030140311403214033140341403514036140371403814039140401404114042140431404414045140461404714048140491405014051140521405314054140551405614057140581405914060140611406214063140641406514066140671406814069140701407114072140731407414075140761407714078140791408014081140821408314084140851408614087140881408914090140911409214093140941409514096140971409814099141001410114102141031410414105141061410714108141091411014111141121411314114141151411614117141181411914120141211412214123141241412514126141271412814129141301413114132141331413414135141361413714138141391414014141141421414314144141451414614147141481414914150141511415214153141541415514156
  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. % Allow landscape pages
  27. \usepackage{pdflscape}
  28. % Allow doing things after the end of the current page
  29. % (to avoid landscape figures breaking up text)
  30. \usepackage{afterpage}
  31. % This one breaks subfigs so it's disabled
  32. % https://tex.stackexchange.com/questions/65680/automatically-bold-first-sentence-of-a-floats-caption
  33. \end_preamble
  34. \use_default_options true
  35. \begin_modules
  36. todonotes
  37. \end_modules
  38. \maintain_unincluded_children false
  39. \language english
  40. \language_package default
  41. \inputencoding utf8
  42. \fontencoding default
  43. \font_roman "default" "default"
  44. \font_sans "default" "default"
  45. \font_typewriter "default" "default"
  46. \font_math "auto" "auto"
  47. \font_default_family default
  48. \use_non_tex_fonts false
  49. \font_sc false
  50. \font_osf false
  51. \font_sf_scale 100 100
  52. \font_tt_scale 100 100
  53. \use_microtype false
  54. \use_dash_ligatures true
  55. \graphics default
  56. \default_output_format pdf4
  57. \output_sync 0
  58. \bibtex_command biber
  59. \index_command default
  60. \paperfontsize 12
  61. \spacing double
  62. \use_hyperref true
  63. \pdf_bookmarks true
  64. \pdf_bookmarksnumbered false
  65. \pdf_bookmarksopen false
  66. \pdf_bookmarksopenlevel 1
  67. \pdf_breaklinks false
  68. \pdf_pdfborder false
  69. \pdf_colorlinks false
  70. \pdf_backref false
  71. \pdf_pdfusetitle true
  72. \papersize letterpaper
  73. \use_geometry true
  74. \use_package amsmath 1
  75. \use_package amssymb 1
  76. \use_package cancel 1
  77. \use_package esint 1
  78. \use_package mathdots 1
  79. \use_package mathtools 1
  80. \use_package mhchem 1
  81. \use_package stackrel 1
  82. \use_package stmaryrd 1
  83. \use_package undertilde 1
  84. \cite_engine biblatex
  85. \cite_engine_type authoryear
  86. \biblio_style plain
  87. \biblatex_bibstyle authoryear
  88. \biblatex_citestyle numeric
  89. \use_bibtopic false
  90. \use_indices false
  91. \paperorientation portrait
  92. \suppress_date false
  93. \justification true
  94. \use_refstyle 1
  95. \use_minted 0
  96. \index Index
  97. \shortcut idx
  98. \color #008000
  99. \end_index
  100. \leftmargin 1.5in
  101. \topmargin 1in
  102. \rightmargin 1in
  103. \bottommargin 1in
  104. \secnumdepth 3
  105. \tocdepth 3
  106. \paragraph_separation indent
  107. \paragraph_indentation default
  108. \is_math_indent 0
  109. \math_numbering_side default
  110. \quotes_style english
  111. \dynamic_quotes 0
  112. \papercolumns 1
  113. \papersides 1
  114. \paperpagestyle default
  115. \tracking_changes false
  116. \output_changes false
  117. \html_math_output 0
  118. \html_css_as_file 0
  119. \html_be_strict false
  120. \end_header
  121. \begin_body
  122. \begin_layout Title
  123. Bioinformatic analysis of complex, high-throughput genomic and epigenomic
  124. data in the context of immunology and transplant rejection
  125. \end_layout
  126. \begin_layout Author
  127. A thesis presented
  128. \begin_inset Newline newline
  129. \end_inset
  130. by
  131. \begin_inset Newline newline
  132. \end_inset
  133. Ryan C.
  134. Thompson
  135. \begin_inset Newline newline
  136. \end_inset
  137. to
  138. \begin_inset Newline newline
  139. \end_inset
  140. The Scripps Research Institute Graduate Program
  141. \begin_inset Newline newline
  142. \end_inset
  143. in partial fulfillment of the requirements for the degree of
  144. \begin_inset Newline newline
  145. \end_inset
  146. Doctor of Philosophy in the subject of Biology
  147. \begin_inset Newline newline
  148. \end_inset
  149. for
  150. \begin_inset Newline newline
  151. \end_inset
  152. The Scripps Research Institute
  153. \begin_inset Newline newline
  154. \end_inset
  155. La Jolla, California
  156. \end_layout
  157. \begin_layout Date
  158. October 2019
  159. \end_layout
  160. \begin_layout Standard
  161. [Copyright notice]
  162. \end_layout
  163. \begin_layout Standard
  164. [Thesis acceptance form]
  165. \end_layout
  166. \begin_layout Standard
  167. [Dedication]
  168. \end_layout
  169. \begin_layout Standard
  170. [Acknowledgements]
  171. \end_layout
  172. \begin_layout Standard
  173. \begin_inset CommandInset toc
  174. LatexCommand tableofcontents
  175. \end_inset
  176. \end_layout
  177. \begin_layout Standard
  178. \begin_inset FloatList table
  179. \end_inset
  180. \end_layout
  181. \begin_layout Standard
  182. \begin_inset FloatList figure
  183. \end_inset
  184. \end_layout
  185. \begin_layout Standard
  186. [List of Abbreviations]
  187. \end_layout
  188. \begin_layout List of TODOs
  189. \end_layout
  190. \begin_layout Standard
  191. \begin_inset Flex TODO Note (inline)
  192. status open
  193. \begin_layout Plain Layout
  194. Check all figures to make sure they fit on the page with their legends.
  195. \end_layout
  196. \end_inset
  197. \end_layout
  198. \begin_layout Standard
  199. \begin_inset Flex TODO Note (inline)
  200. status open
  201. \begin_layout Plain Layout
  202. Search and replace: naive -> naïve
  203. \end_layout
  204. \end_inset
  205. \end_layout
  206. \begin_layout Standard
  207. \begin_inset Flex TODO Note (inline)
  208. status open
  209. \begin_layout Plain Layout
  210. Look into auto-generated nomenclature list: https://wiki.lyx.org/Tips/Nomenclature.
  211. Otherwise, do a manual pass for all abbreviations at the end.
  212. Do nomenclature/abbreviations independently for each chapter.
  213. \end_layout
  214. \end_inset
  215. \end_layout
  216. \begin_layout Standard
  217. \begin_inset Flex TODO Note (inline)
  218. status open
  219. \begin_layout Plain Layout
  220. Make all descriptions consistent in terms of
  221. \begin_inset Quotes eld
  222. \end_inset
  223. we did X
  224. \begin_inset Quotes erd
  225. \end_inset
  226. vs
  227. \begin_inset Quotes eld
  228. \end_inset
  229. I did X
  230. \begin_inset Quotes erd
  231. \end_inset
  232. vs
  233. \begin_inset Quotes eld
  234. \end_inset
  235. X was done
  236. \begin_inset Quotes erd
  237. \end_inset
  238. .
  239. \end_layout
  240. \end_inset
  241. \end_layout
  242. \begin_layout Chapter*
  243. Abstract
  244. \end_layout
  245. \begin_layout Standard
  246. \begin_inset Note Note
  247. status open
  248. \begin_layout Plain Layout
  249. It is included as an integral part of the thesis and should immediately
  250. precede the introduction.
  251. \end_layout
  252. \begin_layout Plain Layout
  253. Preparing your Abstract.
  254. Your abstract (a succinct description of your work) is limited to 350 words.
  255. UMI will shorten it if they must; please do not exceed the limit.
  256. \end_layout
  257. \begin_layout Itemize
  258. Include pertinent place names, names of persons (in full), and other proper
  259. nouns.
  260. These are useful in automated retrieval.
  261. \end_layout
  262. \begin_layout Itemize
  263. Display symbols, as well as foreign words and phrases, clearly and accurately.
  264. Include transliterations for characters other than Roman and Greek letters
  265. and Arabic numerals.
  266. Include accents and diacritical marks.
  267. \end_layout
  268. \begin_layout Itemize
  269. Do not include graphs, charts, tables, or illustrations in your abstract.
  270. \end_layout
  271. \end_inset
  272. \end_layout
  273. \begin_layout Standard
  274. \begin_inset Flex TODO Note (inline)
  275. status open
  276. \begin_layout Plain Layout
  277. Obviously the abstract gets written last.
  278. \end_layout
  279. \end_inset
  280. \end_layout
  281. \begin_layout Chapter
  282. Introduction
  283. \end_layout
  284. \begin_layout Section
  285. Background & Significance
  286. \end_layout
  287. \begin_layout Subsection
  288. Biological motivation
  289. \end_layout
  290. \begin_layout Itemize
  291. Rejection is the major long-term threat to organ and tissue grafts
  292. \end_layout
  293. \begin_deeper
  294. \begin_layout Itemize
  295. Common mechanisms of rejection
  296. \end_layout
  297. \begin_layout Itemize
  298. Effective immune suppression requires monitoring for rejection and tuning
  299. \end_layout
  300. \begin_layout Itemize
  301. Current tests for rejection (tissue biopsy) are invasive and biased
  302. \end_layout
  303. \begin_layout Itemize
  304. A blood test based on microarrays would be less biased and invasive
  305. \end_layout
  306. \end_deeper
  307. \begin_layout Itemize
  308. Memory cells are resistant to immune suppression
  309. \end_layout
  310. \begin_deeper
  311. \begin_layout Itemize
  312. Mechanisms of resistance in memory cells are poorly understood
  313. \end_layout
  314. \begin_layout Itemize
  315. A better understanding of immune memory formation is needed
  316. \end_layout
  317. \end_deeper
  318. \begin_layout Itemize
  319. Mesenchymal stem cell infusion is a promising new treatment to prevent/delay
  320. rejection
  321. \end_layout
  322. \begin_deeper
  323. \begin_layout Itemize
  324. Demonstrated in mice, but not yet in primates
  325. \end_layout
  326. \begin_layout Itemize
  327. Mechanism currently unknown, but MSC are known to be immune modulatory
  328. \end_layout
  329. \end_deeper
  330. \begin_layout Subsection
  331. Overview of bioinformatic analysis methods
  332. \end_layout
  333. \begin_layout Standard
  334. An overview of all the methods used, including what problem they solve,
  335. what assumptions they make, and a basic description of how they work.
  336. \end_layout
  337. \begin_layout Itemize
  338. ChIP-seq Peak calling
  339. \end_layout
  340. \begin_deeper
  341. \begin_layout Itemize
  342. Cross-correlation analysis to determine fragment size
  343. \end_layout
  344. \begin_layout Itemize
  345. Broad vs narrow peaks
  346. \end_layout
  347. \begin_layout Itemize
  348. SICER for broad peaks
  349. \end_layout
  350. \begin_layout Itemize
  351. IDR for biologically reproducible peaks
  352. \end_layout
  353. \begin_layout Itemize
  354. csaw peak filtering guidelines for unbiased downstream analysis
  355. \end_layout
  356. \end_deeper
  357. \begin_layout Itemize
  358. Normalization is non-trivial and application-dependant
  359. \end_layout
  360. \begin_deeper
  361. \begin_layout Itemize
  362. Expression arrays: RMA & fRMA; why fRMA is needed
  363. \end_layout
  364. \begin_layout Itemize
  365. Methylation arrays: M-value transformation approximates normal data but
  366. induces heteroskedasticity
  367. \end_layout
  368. \begin_layout Itemize
  369. RNA-seq: normalize based on assumption that the average gene is not changing
  370. \end_layout
  371. \begin_layout Itemize
  372. ChIP-seq: complex with many considerations, dependent on experimental methods,
  373. biological system, and analysis goals
  374. \end_layout
  375. \end_deeper
  376. \begin_layout Itemize
  377. Limma: The standard linear modeling framework for genomics
  378. \end_layout
  379. \begin_deeper
  380. \begin_layout Itemize
  381. empirical Bayes variance modeling: limma's core feature
  382. \end_layout
  383. \begin_layout Itemize
  384. edgeR & DESeq2: Extend with negative bonomial GLM for RNA-seq and other
  385. count data
  386. \end_layout
  387. \begin_layout Itemize
  388. voom: Extend with precision weights to model mean-variance trend
  389. \end_layout
  390. \begin_layout Itemize
  391. arrayWeights and duplicateCorrelation to handle complex variance structures
  392. \end_layout
  393. \end_deeper
  394. \begin_layout Itemize
  395. sva and ComBat for batch correction
  396. \end_layout
  397. \begin_layout Itemize
  398. Factor analysis: PCA, MDS, MOFA
  399. \end_layout
  400. \begin_deeper
  401. \begin_layout Itemize
  402. Batch-corrected PCA is informative, but careful application is required
  403. to avoid bias
  404. \end_layout
  405. \end_deeper
  406. \begin_layout Itemize
  407. Gene set analysis: camera and SPIA
  408. \end_layout
  409. \begin_layout Section
  410. Innovation
  411. \end_layout
  412. \begin_layout Itemize
  413. MSC infusion to improve transplant outcomes (prevent/delay rejection)
  414. \end_layout
  415. \begin_deeper
  416. \begin_layout Itemize
  417. Characterize MSC response to interferon gamma
  418. \end_layout
  419. \begin_layout Itemize
  420. IFN-g is thought to stimulate their function
  421. \end_layout
  422. \begin_layout Itemize
  423. Test IFN-g treated MSC infusion as a therapy to delay graft rejection in
  424. cynomolgus monkeys
  425. \end_layout
  426. \begin_layout Itemize
  427. Monitor animals post-transplant using blood RNA-seq at serial time points
  428. \end_layout
  429. \end_deeper
  430. \begin_layout Itemize
  431. Investigate dynamics of histone marks in CD4 T-cell activation and memory
  432. \end_layout
  433. \begin_deeper
  434. \begin_layout Itemize
  435. Previous studies have looked at single snapshots of histone marks
  436. \end_layout
  437. \begin_layout Itemize
  438. Instead, look at changes in histone marks across activation and memory
  439. \end_layout
  440. \end_deeper
  441. \begin_layout Itemize
  442. High-throughput sequencing and microarray technologies
  443. \end_layout
  444. \begin_deeper
  445. \begin_layout Itemize
  446. Powerful methods for assaying gene expression and epigenetics across entire
  447. genomes
  448. \end_layout
  449. \begin_layout Itemize
  450. Proper analysis requires finding and exploiting systematic genome-wide trends
  451. \end_layout
  452. \end_deeper
  453. \begin_layout Chapter
  454. Reproducible genome-wide epigenetic analysis of H3K4 and H3K27 methylation
  455. in naive and memory CD4 T-cell activation
  456. \end_layout
  457. \begin_layout Standard
  458. \begin_inset Flex TODO Note (inline)
  459. status open
  460. \begin_layout Plain Layout
  461. Chapter author list: Me, Sarah, Dan
  462. \end_layout
  463. \end_inset
  464. \end_layout
  465. \begin_layout Standard
  466. \begin_inset Flex TODO Note (inline)
  467. status open
  468. \begin_layout Plain Layout
  469. Need better section titles throughout the entire chapter
  470. \end_layout
  471. \end_inset
  472. \end_layout
  473. \begin_layout Section
  474. Approach
  475. \end_layout
  476. \begin_layout Standard
  477. \begin_inset Flex TODO Note (inline)
  478. status open
  479. \begin_layout Plain Layout
  480. Check on the exact correct way to write
  481. \begin_inset Quotes eld
  482. \end_inset
  483. CD4 T-cell
  484. \begin_inset Quotes erd
  485. \end_inset
  486. .
  487. I think there might be a plus sign somwehere in there now? Also, maybe
  488. figure out a reasonable way to abbreviate
  489. \begin_inset Quotes eld
  490. \end_inset
  491. naive CD4 T-cells
  492. \begin_inset Quotes erd
  493. \end_inset
  494. and
  495. \begin_inset Quotes eld
  496. \end_inset
  497. memory CD4 T-cells
  498. \begin_inset Quotes erd
  499. \end_inset
  500. .
  501. \end_layout
  502. \end_inset
  503. \end_layout
  504. \begin_layout Standard
  505. \begin_inset Flex TODO Note (inline)
  506. status open
  507. \begin_layout Plain Layout
  508. Is it ok to just copy a bunch of citations from the intros to Sarah's papers?
  509. That feels like cheating somehow.
  510. \end_layout
  511. \end_inset
  512. \end_layout
  513. \begin_layout Standard
  514. \begin_inset Flex TODO Note (inline)
  515. status open
  516. \begin_layout Plain Layout
  517. How much of this goes in Chapter 1?
  518. \end_layout
  519. \end_inset
  520. \end_layout
  521. \begin_layout Standard
  522. CD4 T-cells are central to all adaptive immune responses, as well as immune
  523. memory [CITE?].
  524. After an infection is cleared, a subset of the naive CD4 T-cells that responded
  525. to that infection differentiate into memory CD4 T-cells, which are responsible
  526. for responding to the same pathogen in the future.
  527. Memory CD4 T-cells are functionally distinct, able to respond to an infection
  528. more quickly and without the co-stimulation requried by naive CD4 T-cells.
  529. However, the molecular mechanisms underlying this functional distinction
  530. are not well-understood.
  531. Epigenetic regulation is thought to be
  532. \end_layout
  533. \begin_layout Standard
  534. H3K4me2, H3K4me3 and H3K27me3 are three histone marks thought to be major
  535. epigenetic regulators of gene expression.
  536. The goal of the present study is to investigate the role of these histone
  537. marks in CD4 T-cell activation kinetics and memory differentiation.
  538. \end_layout
  539. \begin_layout Standard
  540. \begin_inset Note Note
  541. status open
  542. \begin_layout Plain Layout
  543. Probably goes in CH1:
  544. \end_layout
  545. \begin_layout Plain Layout
  546. Generally, H3K4me2 and H3K4me3 are often observed in the promoters of highly
  547. transcribed genes, while H3K27me3 is more often observed in promoters of
  548. inactive genes with little to no transcription occurring.
  549. The causal relationship between these histone modifications and gene transcript
  550. ion is complex, and likely involves positive and negative feedback loops
  551. between the two.
  552. \end_layout
  553. \end_inset
  554. \end_layout
  555. \begin_layout Itemize
  556. Looking at these marks during CD4 activation and memory should reveal new
  557. mechanistic details
  558. \end_layout
  559. \begin_layout Itemize
  560. Test
  561. \begin_inset Quotes eld
  562. \end_inset
  563. poised promoter
  564. \begin_inset Quotes erd
  565. \end_inset
  566. hypothesis in which H3K4 and H3K27 are both methylated
  567. \end_layout
  568. \begin_layout Itemize
  569. Expand scope of analysis beyond simple promoter counts
  570. \end_layout
  571. \begin_deeper
  572. \begin_layout Itemize
  573. Analyze peaks genome-wide, including in intergenic regions
  574. \end_layout
  575. \begin_layout Itemize
  576. Analysis of coverage distribution shape within promoters, e.g.
  577. upstream vs downstream coverage
  578. \end_layout
  579. \end_deeper
  580. \begin_layout Section
  581. Methods
  582. \end_layout
  583. \begin_layout Standard
  584. \begin_inset Flex TODO Note (inline)
  585. status open
  586. \begin_layout Plain Layout
  587. Look up some more details from the papers (e.g.
  588. activation method).
  589. \end_layout
  590. \end_inset
  591. \end_layout
  592. \begin_layout Standard
  593. A reproducible workflow was written to analyze the raw ChIP-seq and RNA-seq
  594. data from previous studies
  595. \begin_inset CommandInset citation
  596. LatexCommand cite
  597. key "gh-cd4-csaw,LaMere2016,LaMere2017"
  598. literal "true"
  599. \end_inset
  600. .
  601. Briefly, this data consists of RNA-seq and ChIP-seq from CD4 T-cells cultured
  602. from 4 donors.
  603. From each donor, naive and memory CD4 T-cells were isolated separately.
  604. Then cultures of both cells were activated [how?], and samples were taken
  605. at 4 time points: Day 0 (pre-activation), Day 1 (early activation), Day
  606. 5 (peak activation), and Day 14 (post-activation).
  607. For each combination of cell type and time point, RNA was isolated and
  608. sequenced, and ChIP-seq was performed for each of 3 histone marks: H3K4me2,
  609. H3K4me3, and H3K27me3.
  610. The ChIP-seq input DNA was also sequenced for each sample.
  611. The result was 32 samples for each assay.
  612. \end_layout
  613. \begin_layout Subsection
  614. RNA-seq differential expression analysis
  615. \end_layout
  616. \begin_layout Standard
  617. \begin_inset Note Note
  618. status collapsed
  619. \begin_layout Plain Layout
  620. \begin_inset Float figure
  621. wide false
  622. sideways false
  623. status open
  624. \begin_layout Plain Layout
  625. \align center
  626. \begin_inset Float figure
  627. wide false
  628. sideways false
  629. status collapsed
  630. \begin_layout Plain Layout
  631. \align center
  632. \begin_inset Graphics
  633. filename graphics/CD4-csaw/rnaseq-compare/ensmebl-vs-entrez-star-CROP.png
  634. lyxscale 25
  635. width 35col%
  636. groupId rna-comp-subfig
  637. \end_inset
  638. \end_layout
  639. \begin_layout Plain Layout
  640. \begin_inset Caption Standard
  641. \begin_layout Plain Layout
  642. STAR quantification, Entrez vs Ensembl gene annotation
  643. \end_layout
  644. \end_inset
  645. \end_layout
  646. \end_inset
  647. \begin_inset space \qquad{}
  648. \end_inset
  649. \begin_inset Float figure
  650. wide false
  651. sideways false
  652. status collapsed
  653. \begin_layout Plain Layout
  654. \align center
  655. \begin_inset Graphics
  656. filename graphics/CD4-csaw/rnaseq-compare/ensmebl-vs-entrez-shoal-CROP.png
  657. lyxscale 25
  658. width 35col%
  659. groupId rna-comp-subfig
  660. \end_inset
  661. \end_layout
  662. \begin_layout Plain Layout
  663. \begin_inset Caption Standard
  664. \begin_layout Plain Layout
  665. Salmon+Shoal quantification, Entrez vs Ensembl gene annotation
  666. \end_layout
  667. \end_inset
  668. \end_layout
  669. \end_inset
  670. \end_layout
  671. \begin_layout Plain Layout
  672. \align center
  673. \begin_inset Float figure
  674. wide false
  675. sideways false
  676. status collapsed
  677. \begin_layout Plain Layout
  678. \align center
  679. \begin_inset Graphics
  680. filename graphics/CD4-csaw/rnaseq-compare/star-vs-hisat2-CROP.png
  681. lyxscale 25
  682. width 35col%
  683. groupId rna-comp-subfig
  684. \end_inset
  685. \end_layout
  686. \begin_layout Plain Layout
  687. \begin_inset Caption Standard
  688. \begin_layout Plain Layout
  689. STAR vs HISAT2 quantification, Ensembl gene annotation
  690. \end_layout
  691. \end_inset
  692. \end_layout
  693. \end_inset
  694. \begin_inset space \qquad{}
  695. \end_inset
  696. \begin_inset Float figure
  697. wide false
  698. sideways false
  699. status collapsed
  700. \begin_layout Plain Layout
  701. \align center
  702. \begin_inset Graphics
  703. filename graphics/CD4-csaw/rnaseq-compare/star-vs-salmon-CROP.png
  704. lyxscale 25
  705. width 35col%
  706. groupId rna-comp-subfig
  707. \end_inset
  708. \end_layout
  709. \begin_layout Plain Layout
  710. \begin_inset Caption Standard
  711. \begin_layout Plain Layout
  712. Salomn vs STAR quantification, Ensembl gene annotation
  713. \end_layout
  714. \end_inset
  715. \end_layout
  716. \end_inset
  717. \end_layout
  718. \begin_layout Plain Layout
  719. \align center
  720. \begin_inset Float figure
  721. wide false
  722. sideways false
  723. status collapsed
  724. \begin_layout Plain Layout
  725. \align center
  726. \begin_inset Graphics
  727. filename graphics/CD4-csaw/rnaseq-compare/salmon-vs-kallisto-CROP.png
  728. lyxscale 25
  729. width 35col%
  730. groupId rna-comp-subfig
  731. \end_inset
  732. \end_layout
  733. \begin_layout Plain Layout
  734. \begin_inset Caption Standard
  735. \begin_layout Plain Layout
  736. Salmon vs Kallisto quantification, Ensembl gene annotation
  737. \end_layout
  738. \end_inset
  739. \end_layout
  740. \end_inset
  741. \begin_inset space \qquad{}
  742. \end_inset
  743. \begin_inset Float figure
  744. wide false
  745. sideways false
  746. status collapsed
  747. \begin_layout Plain Layout
  748. \align center
  749. \begin_inset Graphics
  750. filename graphics/CD4-csaw/rnaseq-compare/salmon-vs-shoal-CROP.png
  751. lyxscale 25
  752. width 35col%
  753. groupId rna-comp-subfig
  754. \end_inset
  755. \end_layout
  756. \begin_layout Plain Layout
  757. \begin_inset Caption Standard
  758. \begin_layout Plain Layout
  759. Salmon+Shoal vs Salmon alone, Ensembl gene annotation
  760. \end_layout
  761. \end_inset
  762. \end_layout
  763. \end_inset
  764. \end_layout
  765. \begin_layout Plain Layout
  766. \begin_inset Caption Standard
  767. \begin_layout Plain Layout
  768. \begin_inset CommandInset label
  769. LatexCommand label
  770. name "fig:RNA-norm-comp"
  771. \end_inset
  772. RNA-seq comparisons
  773. \end_layout
  774. \end_inset
  775. \end_layout
  776. \end_inset
  777. \end_layout
  778. \end_inset
  779. \end_layout
  780. \begin_layout Standard
  781. Sequence reads were retrieved from the Sequence Read Archive (SRA)
  782. \begin_inset CommandInset citation
  783. LatexCommand cite
  784. key "Leinonen2011"
  785. literal "false"
  786. \end_inset
  787. .
  788. Five different alignment and quantification methods were tested for the
  789. RNA-seq data
  790. \begin_inset CommandInset citation
  791. LatexCommand cite
  792. key "Dobin2012,Kim2019,Liao2014,Pimentel2016,Patro2017,gh-shoal,gh-hg38-ref"
  793. literal "false"
  794. \end_inset
  795. .
  796. Each quantification was tested with both Ensembl transcripts and UCSC known
  797. gene annotations [CITE? Also which versions of each?].
  798. Comparisons of downstream results from each combination of quantification
  799. method and reference revealed that all quantifications gave broadly similar
  800. results for most genes, so shoal with the Ensembl annotation was chosen
  801. as the method theoretically most likely to partially mitigate some of the
  802. batch effect in the data.
  803. \end_layout
  804. \begin_layout Standard
  805. \begin_inset Float figure
  806. wide false
  807. sideways false
  808. status collapsed
  809. \begin_layout Plain Layout
  810. \align center
  811. \begin_inset Float figure
  812. wide false
  813. sideways false
  814. status open
  815. \begin_layout Plain Layout
  816. \align center
  817. \begin_inset Graphics
  818. filename graphics/CD4-csaw/RNA-seq/PCA-no-batchsub-CROP.png
  819. lyxscale 25
  820. width 75col%
  821. groupId rna-pca-subfig
  822. \end_inset
  823. \end_layout
  824. \begin_layout Plain Layout
  825. \begin_inset Caption Standard
  826. \begin_layout Plain Layout
  827. \series bold
  828. \begin_inset CommandInset label
  829. LatexCommand label
  830. name "fig:RNA-PCA-no-batchsub"
  831. \end_inset
  832. Before batch correction
  833. \end_layout
  834. \end_inset
  835. \end_layout
  836. \end_inset
  837. \end_layout
  838. \begin_layout Plain Layout
  839. \align center
  840. \begin_inset Float figure
  841. wide false
  842. sideways false
  843. status open
  844. \begin_layout Plain Layout
  845. \align center
  846. \begin_inset Graphics
  847. filename graphics/CD4-csaw/RNA-seq/PCA-combat-batchsub-CROP.png
  848. lyxscale 25
  849. width 75col%
  850. groupId rna-pca-subfig
  851. \end_inset
  852. \end_layout
  853. \begin_layout Plain Layout
  854. \begin_inset Caption Standard
  855. \begin_layout Plain Layout
  856. \series bold
  857. \begin_inset CommandInset label
  858. LatexCommand label
  859. name "fig:RNA-PCA-ComBat-batchsub"
  860. \end_inset
  861. After batch correction with ComBat
  862. \end_layout
  863. \end_inset
  864. \end_layout
  865. \end_inset
  866. \end_layout
  867. \begin_layout Plain Layout
  868. \begin_inset Caption Standard
  869. \begin_layout Plain Layout
  870. \series bold
  871. \begin_inset CommandInset label
  872. LatexCommand label
  873. name "fig:RNA-PCA"
  874. \end_inset
  875. PCoA plots of RNA-seq data showing effect of batch correction.
  876. \end_layout
  877. \end_inset
  878. \end_layout
  879. \end_inset
  880. \end_layout
  881. \begin_layout Standard
  882. Due to an error in sample preparation, the RNA from the samples for days
  883. 0 and 5 were sequenced using a different kit than those for days 1 and
  884. 14.
  885. This induced a substantial batch effect in the data due to differences
  886. in sequencing biases between the two kits, and this batch effect is unfortunate
  887. ly confounded with the time point variable (Figure
  888. \begin_inset CommandInset ref
  889. LatexCommand ref
  890. reference "fig:RNA-PCA-no-batchsub"
  891. plural "false"
  892. caps "false"
  893. noprefix "false"
  894. \end_inset
  895. ).
  896. To do the best possible analysis with this data, this batch effect was
  897. subtracted out from the data using ComBat
  898. \begin_inset CommandInset citation
  899. LatexCommand cite
  900. key "Johnson2007"
  901. literal "false"
  902. \end_inset
  903. , ignoring the time point variable due to the confounding with the batch
  904. variable.
  905. The result is a marked improvement, but the unavoidable counfounding with
  906. time point means that certain real patterns of gene expression will be
  907. indistinguishable from the batch effect and subtracted out as a result.
  908. Specifically, any
  909. \begin_inset Quotes eld
  910. \end_inset
  911. zig-zag
  912. \begin_inset Quotes erd
  913. \end_inset
  914. pattern, such as a gene whose expression goes up on day 1, down on day
  915. 5, and back up again on day 14, will be attenuated or eliminated entirely.
  916. In the context of a T-cell activation time course, it is unlikely that
  917. many genes of interest will follow such an expression pattern, so this
  918. loss was deemed an acceptable cost for correcting the batch effect.
  919. \end_layout
  920. \begin_layout Standard
  921. \begin_inset Float figure
  922. wide false
  923. sideways false
  924. status collapsed
  925. \begin_layout Plain Layout
  926. \begin_inset Flex TODO Note (inline)
  927. status open
  928. \begin_layout Plain Layout
  929. Just take the top row
  930. \end_layout
  931. \end_inset
  932. \end_layout
  933. \begin_layout Plain Layout
  934. \align center
  935. \begin_inset Graphics
  936. filename graphics/CD4-csaw/RNA-seq/weights-vs-covars-CROP.png
  937. lyxscale 25
  938. width 100col%
  939. groupId colwidth-raster
  940. \end_inset
  941. \end_layout
  942. \begin_layout Plain Layout
  943. \begin_inset Caption Standard
  944. \begin_layout Plain Layout
  945. \series bold
  946. \begin_inset CommandInset label
  947. LatexCommand label
  948. name "fig:RNA-seq-weights-vs-covars"
  949. \end_inset
  950. RNA-seq sample weights, grouped by experimental and technical covariates.
  951. \end_layout
  952. \end_inset
  953. \end_layout
  954. \end_inset
  955. \end_layout
  956. \begin_layout Standard
  957. However, removing the systematic component of the batch effect still leaves
  958. the noise component.
  959. The gene quantifications from the first batch are substantially noisier
  960. than those in the second batch.
  961. This analysis corrected for this by using limma's sample weighting method
  962. to assign lower weights to the noisy samples of batch 1
  963. \begin_inset CommandInset citation
  964. LatexCommand cite
  965. key "Ritchie2006,Liu2015"
  966. literal "false"
  967. \end_inset
  968. .
  969. The resulting analysis gives an accurate assessment of statistical significance
  970. for all comparisons, which unfortuantely means a loss of statistical power
  971. for comparisons involving samples in batch 1.
  972. \end_layout
  973. \begin_layout Standard
  974. In any case, the RNA-seq counts were first normalized using trimmed mean
  975. of M-values
  976. \begin_inset CommandInset citation
  977. LatexCommand cite
  978. key "Robinson2010"
  979. literal "false"
  980. \end_inset
  981. , converted to normalized logCPM with quality weights using voomWithQualityWeigh
  982. ts
  983. \begin_inset CommandInset citation
  984. LatexCommand cite
  985. key "Law2013,Liu2015"
  986. literal "false"
  987. \end_inset
  988. , and batch-corrected at this point using ComBat.
  989. A linear model was fit to the batch-corrected, quality-weighted data for
  990. each gene using limma, and each gene was tested for differential expression
  991. using limma's empirical Bayes moderated
  992. \begin_inset Formula $t$
  993. \end_inset
  994. -test
  995. \begin_inset CommandInset citation
  996. LatexCommand cite
  997. key "Smyth2005,Law2013,Phipson2013"
  998. literal "false"
  999. \end_inset
  1000. .
  1001. \end_layout
  1002. \begin_layout Subsection
  1003. ChIP-seq differential modification analysis
  1004. \end_layout
  1005. \begin_layout Standard
  1006. \begin_inset Float figure
  1007. wide false
  1008. sideways false
  1009. status collapsed
  1010. \begin_layout Plain Layout
  1011. \align center
  1012. \begin_inset Float figure
  1013. wide false
  1014. sideways false
  1015. status open
  1016. \begin_layout Plain Layout
  1017. \align center
  1018. \begin_inset Graphics
  1019. filename graphics/CD4-csaw/csaw/CCF-plots-noBL-PAGE2-CROP.pdf
  1020. lyxscale 50
  1021. height 40theight%
  1022. groupId ccf-subfig
  1023. \end_inset
  1024. \end_layout
  1025. \begin_layout Plain Layout
  1026. \begin_inset Caption Standard
  1027. \begin_layout Plain Layout
  1028. \series bold
  1029. \begin_inset CommandInset label
  1030. LatexCommand label
  1031. name "fig:CCF-without-blacklist"
  1032. \end_inset
  1033. Cross-correlation plots without removing blacklisted reads.
  1034. \series default
  1035. Without blacklisting, many artifactual peaks are visible in the cross-correlatio
  1036. ns of the ChIP-seq samples, and the peak at the true fragment size (147
  1037. \begin_inset space ~
  1038. \end_inset
  1039. bp) is frequently overshadowed by the artifactual peak at the read length
  1040. (100
  1041. \begin_inset space ~
  1042. \end_inset
  1043. bp).
  1044. \end_layout
  1045. \end_inset
  1046. \end_layout
  1047. \end_inset
  1048. \end_layout
  1049. \begin_layout Plain Layout
  1050. \align center
  1051. \begin_inset Float figure
  1052. wide false
  1053. sideways false
  1054. status open
  1055. \begin_layout Plain Layout
  1056. \align center
  1057. \begin_inset Graphics
  1058. filename graphics/CD4-csaw/csaw/CCF-plots-PAGE2-CROP.pdf
  1059. lyxscale 50
  1060. height 40theight%
  1061. groupId ccf-subfig
  1062. \end_inset
  1063. \end_layout
  1064. \begin_layout Plain Layout
  1065. \begin_inset Caption Standard
  1066. \begin_layout Plain Layout
  1067. \series bold
  1068. \begin_inset CommandInset label
  1069. LatexCommand label
  1070. name "fig:CCF-with-blacklist"
  1071. \end_inset
  1072. Cross-correlation plots with blacklisted reads removed.
  1073. \series default
  1074. After blacklisting, most ChIP-seq samples have clean-looking periodic cross-cor
  1075. relation plots, with the largest peak around 147
  1076. \begin_inset space ~
  1077. \end_inset
  1078. bp, the expected size for a fragment of DNA from a single nucleosome, and
  1079. little to no peak at the read length, 100
  1080. \begin_inset space ~
  1081. \end_inset
  1082. bp.
  1083. \end_layout
  1084. \end_inset
  1085. \end_layout
  1086. \end_inset
  1087. \end_layout
  1088. \begin_layout Plain Layout
  1089. \begin_inset Caption Standard
  1090. \begin_layout Plain Layout
  1091. \series bold
  1092. \begin_inset CommandInset label
  1093. LatexCommand label
  1094. name "fig:CCF-master"
  1095. \end_inset
  1096. Strand cross-correlation plots for ChIP-seq data, before and after blacklisting.
  1097. \end_layout
  1098. \end_inset
  1099. \end_layout
  1100. \end_inset
  1101. \end_layout
  1102. \begin_layout Standard
  1103. \begin_inset Note Note
  1104. status open
  1105. \begin_layout Plain Layout
  1106. \begin_inset Float figure
  1107. wide false
  1108. sideways false
  1109. status collapsed
  1110. \begin_layout Plain Layout
  1111. \align center
  1112. \begin_inset Graphics
  1113. filename graphics/CD4-csaw/ChIP-seq/H3K4me2-sample-MAplot-bins-CROP.png
  1114. lyxscale 25
  1115. width 100col%
  1116. groupId colwidth-raster
  1117. \end_inset
  1118. \end_layout
  1119. \begin_layout Plain Layout
  1120. \begin_inset Caption Standard
  1121. \begin_layout Plain Layout
  1122. \series bold
  1123. \begin_inset CommandInset label
  1124. LatexCommand label
  1125. name "fig:MA-plot-bigbins"
  1126. \end_inset
  1127. MA plot of H3K4me2 read counts in 10kb bins for two arbitrary samples.
  1128. \end_layout
  1129. \end_inset
  1130. \end_layout
  1131. \end_inset
  1132. \end_layout
  1133. \end_inset
  1134. \end_layout
  1135. \begin_layout Standard
  1136. \begin_inset Flex TODO Note (inline)
  1137. status open
  1138. \begin_layout Plain Layout
  1139. Be consistent about use of
  1140. \begin_inset Quotes eld
  1141. \end_inset
  1142. differential binding
  1143. \begin_inset Quotes erd
  1144. \end_inset
  1145. vs
  1146. \begin_inset Quotes eld
  1147. \end_inset
  1148. differential modification
  1149. \begin_inset Quotes erd
  1150. \end_inset
  1151. throughout this chapter.
  1152. The latter is usually preferred.
  1153. \end_layout
  1154. \end_inset
  1155. \end_layout
  1156. \begin_layout Standard
  1157. Sequence reads were retrieved from SRA
  1158. \begin_inset CommandInset citation
  1159. LatexCommand cite
  1160. key "Leinonen2011"
  1161. literal "false"
  1162. \end_inset
  1163. .
  1164. ChIP-seq (and input) reads were aligned to GRCh38 genome assembly using
  1165. Bowtie 2
  1166. \begin_inset CommandInset citation
  1167. LatexCommand cite
  1168. key "Langmead2012,Schneider2017,gh-hg38-ref"
  1169. literal "false"
  1170. \end_inset
  1171. .
  1172. Artifact regions were annotated using a custom implementation of the GreyListCh
  1173. IP algorithm, and these
  1174. \begin_inset Quotes eld
  1175. \end_inset
  1176. greylists
  1177. \begin_inset Quotes erd
  1178. \end_inset
  1179. were merged with the published ENCODE blacklists
  1180. \begin_inset CommandInset citation
  1181. LatexCommand cite
  1182. key "greylistchip,Amemiya2019,Dunham2012,gh-cd4-csaw"
  1183. literal "false"
  1184. \end_inset
  1185. .
  1186. Any read or called peak overlapping one of these regions was regarded as
  1187. artifactual and excluded from downstream analyses.
  1188. Figure
  1189. \begin_inset CommandInset ref
  1190. LatexCommand ref
  1191. reference "fig:CCF-master"
  1192. plural "false"
  1193. caps "false"
  1194. noprefix "false"
  1195. \end_inset
  1196. shows the improvement after blacklisting in the strand cross-correlation
  1197. plots, a common quality control plot for ChIP-seq data.
  1198. Peaks were called using epic, an implementation of the SICER algorithm
  1199. \begin_inset CommandInset citation
  1200. LatexCommand cite
  1201. key "Zang2009,gh-epic"
  1202. literal "false"
  1203. \end_inset
  1204. .
  1205. Peaks were also called separately using MACS, but MACS was determined to
  1206. be a poor fit for the data, and these peak calls are not used in any further
  1207. analyses
  1208. \begin_inset CommandInset citation
  1209. LatexCommand cite
  1210. key "Zhang2008"
  1211. literal "false"
  1212. \end_inset
  1213. .
  1214. Consensus peaks were determined by applying the irreproducible discovery
  1215. rate (IDR) framework
  1216. \begin_inset CommandInset citation
  1217. LatexCommand cite
  1218. key "Li2006,gh-idr"
  1219. literal "false"
  1220. \end_inset
  1221. to find peaks consistently called in the same locations across all 4 donors.
  1222. \end_layout
  1223. \begin_layout Standard
  1224. Promoters were defined by computing the distance from each annotated TSS
  1225. to the nearest called peak and examining the distribution of distances,
  1226. observing that peaks for each histone mark were enriched within a certain
  1227. distance of the TSS.
  1228. For H3K4me2 and H3K4me3, this distance was about 1
  1229. \begin_inset space ~
  1230. \end_inset
  1231. kb, while for H3K27me3 it was 2.5
  1232. \begin_inset space ~
  1233. \end_inset
  1234. kb.
  1235. These distances were used as an
  1236. \begin_inset Quotes eld
  1237. \end_inset
  1238. effective promoter radius
  1239. \begin_inset Quotes erd
  1240. \end_inset
  1241. for each mark.
  1242. The promoter region for each gene was defined as the region of the genome
  1243. within this distance upstream or downstream of the gene's annotated TSS.
  1244. For genes with multiple annotated TSSs, a promoter region was defined for
  1245. each TSS individually, and any promoters that overlapped (due to multiple
  1246. TSSs being closer than 2 times the radius) were merged into one large promoter.
  1247. Thus, some genes had multiple promoters defined, which were each analyzed
  1248. separately for differential modification.
  1249. \end_layout
  1250. \begin_layout Standard
  1251. \begin_inset Float figure
  1252. wide false
  1253. sideways false
  1254. status collapsed
  1255. \begin_layout Plain Layout
  1256. \begin_inset Float figure
  1257. wide false
  1258. sideways false
  1259. status collapsed
  1260. \begin_layout Plain Layout
  1261. \align center
  1262. \begin_inset Graphics
  1263. filename graphics/CD4-csaw/ChIP-seq/H3K4me2-PCA-raw-CROP.png
  1264. lyxscale 25
  1265. width 45col%
  1266. groupId pcoa-subfig
  1267. \end_inset
  1268. \end_layout
  1269. \begin_layout Plain Layout
  1270. \begin_inset Caption Standard
  1271. \begin_layout Plain Layout
  1272. \series bold
  1273. \begin_inset CommandInset label
  1274. LatexCommand label
  1275. name "fig:PCoA-H3K4me2-bad"
  1276. \end_inset
  1277. H3K4me2, no correction
  1278. \end_layout
  1279. \end_inset
  1280. \end_layout
  1281. \end_inset
  1282. \begin_inset space \hfill{}
  1283. \end_inset
  1284. \begin_inset Float figure
  1285. wide false
  1286. sideways false
  1287. status collapsed
  1288. \begin_layout Plain Layout
  1289. \align center
  1290. \begin_inset Graphics
  1291. filename graphics/CD4-csaw/ChIP-seq/H3K4me2-PCA-SVsub-CROP.png
  1292. lyxscale 25
  1293. width 45col%
  1294. groupId pcoa-subfig
  1295. \end_inset
  1296. \end_layout
  1297. \begin_layout Plain Layout
  1298. \begin_inset Caption Standard
  1299. \begin_layout Plain Layout
  1300. \series bold
  1301. \begin_inset CommandInset label
  1302. LatexCommand label
  1303. name "fig:PCoA-H3K4me2-good"
  1304. \end_inset
  1305. H3K4me2, SVs subtracted
  1306. \end_layout
  1307. \end_inset
  1308. \end_layout
  1309. \end_inset
  1310. \end_layout
  1311. \begin_layout Plain Layout
  1312. \begin_inset Float figure
  1313. wide false
  1314. sideways false
  1315. status collapsed
  1316. \begin_layout Plain Layout
  1317. \align center
  1318. \begin_inset Graphics
  1319. filename graphics/CD4-csaw/ChIP-seq/H3K4me3-PCA-raw-CROP.png
  1320. lyxscale 25
  1321. width 45col%
  1322. groupId pcoa-subfig
  1323. \end_inset
  1324. \end_layout
  1325. \begin_layout Plain Layout
  1326. \begin_inset Caption Standard
  1327. \begin_layout Plain Layout
  1328. \series bold
  1329. \begin_inset CommandInset label
  1330. LatexCommand label
  1331. name "fig:PCoA-H3K4me3-bad"
  1332. \end_inset
  1333. H3K4me3, no correction
  1334. \end_layout
  1335. \end_inset
  1336. \end_layout
  1337. \end_inset
  1338. \begin_inset space \hfill{}
  1339. \end_inset
  1340. \begin_inset Float figure
  1341. wide false
  1342. sideways false
  1343. status collapsed
  1344. \begin_layout Plain Layout
  1345. \align center
  1346. \begin_inset Graphics
  1347. filename graphics/CD4-csaw/ChIP-seq/H3K4me3-PCA-SVsub-CROP.png
  1348. lyxscale 25
  1349. width 45col%
  1350. groupId pcoa-subfig
  1351. \end_inset
  1352. \end_layout
  1353. \begin_layout Plain Layout
  1354. \begin_inset Caption Standard
  1355. \begin_layout Plain Layout
  1356. \series bold
  1357. \begin_inset CommandInset label
  1358. LatexCommand label
  1359. name "fig:PCoA-H3K4me3-good"
  1360. \end_inset
  1361. H3K4me3, SVs subtracted
  1362. \end_layout
  1363. \end_inset
  1364. \end_layout
  1365. \end_inset
  1366. \end_layout
  1367. \begin_layout Plain Layout
  1368. \begin_inset Float figure
  1369. wide false
  1370. sideways false
  1371. status collapsed
  1372. \begin_layout Plain Layout
  1373. \align center
  1374. \begin_inset Graphics
  1375. filename graphics/CD4-csaw/ChIP-seq/H3K27me3-PCA-raw-CROP.png
  1376. lyxscale 25
  1377. width 45col%
  1378. groupId pcoa-subfig
  1379. \end_inset
  1380. \end_layout
  1381. \begin_layout Plain Layout
  1382. \begin_inset Caption Standard
  1383. \begin_layout Plain Layout
  1384. \series bold
  1385. \begin_inset CommandInset label
  1386. LatexCommand label
  1387. name "fig:PCoA-H3K27me3-bad"
  1388. \end_inset
  1389. H3K27me3, no correction
  1390. \end_layout
  1391. \end_inset
  1392. \end_layout
  1393. \end_inset
  1394. \begin_inset space \hfill{}
  1395. \end_inset
  1396. \begin_inset Float figure
  1397. wide false
  1398. sideways false
  1399. status collapsed
  1400. \begin_layout Plain Layout
  1401. \align center
  1402. \begin_inset Graphics
  1403. filename graphics/CD4-csaw/ChIP-seq/H3K27me3-PCA-SVsub-CROP.png
  1404. lyxscale 25
  1405. width 45col%
  1406. groupId pcoa-subfig
  1407. \end_inset
  1408. \end_layout
  1409. \begin_layout Plain Layout
  1410. \begin_inset Caption Standard
  1411. \begin_layout Plain Layout
  1412. \series bold
  1413. \begin_inset CommandInset label
  1414. LatexCommand label
  1415. name "fig:PCoA-H3K27me3-good"
  1416. \end_inset
  1417. H3K27me3, SVs subtracted
  1418. \end_layout
  1419. \end_inset
  1420. \end_layout
  1421. \end_inset
  1422. \end_layout
  1423. \begin_layout Plain Layout
  1424. \begin_inset Caption Standard
  1425. \begin_layout Plain Layout
  1426. \series bold
  1427. \begin_inset CommandInset label
  1428. LatexCommand label
  1429. name "fig:PCoA-ChIP"
  1430. \end_inset
  1431. PCoA plots of ChIP-seq sliding window data, before and after subtracting
  1432. surrogate variables (SVs).
  1433. \end_layout
  1434. \end_inset
  1435. \end_layout
  1436. \end_inset
  1437. \end_layout
  1438. \begin_layout Standard
  1439. Reads in promoters, peaks, and sliding windows across the genome were counted
  1440. and normalized using csaw and analyzed for differential modification using
  1441. edgeR
  1442. \begin_inset CommandInset citation
  1443. LatexCommand cite
  1444. key "Lun2014,Lun2015a,Lund2012,Phipson2016"
  1445. literal "false"
  1446. \end_inset
  1447. .
  1448. Unobserved confounding factors in the ChIP-seq data were corrected using
  1449. SVA
  1450. \begin_inset CommandInset citation
  1451. LatexCommand cite
  1452. key "Leek2007,Leek2014"
  1453. literal "false"
  1454. \end_inset
  1455. .
  1456. Principal coordinate plots of the promoter count data for each histone
  1457. mark before and after subtracting surrogate variable effects are shown
  1458. in Figure
  1459. \begin_inset CommandInset ref
  1460. LatexCommand ref
  1461. reference "fig:PCoA-ChIP"
  1462. plural "false"
  1463. caps "false"
  1464. noprefix "false"
  1465. \end_inset
  1466. .
  1467. \end_layout
  1468. \begin_layout Standard
  1469. To investigate whether the location of a peak within the promoter region
  1470. was important,
  1471. \begin_inset Quotes eld
  1472. \end_inset
  1473. relative coverage profiles
  1474. \begin_inset Quotes erd
  1475. \end_inset
  1476. were generated.
  1477. First, 500-bp sliding windows were tiled around each annotated TSS: one
  1478. window centered on the TSS itself, and 10 windows each upstream and downstream,
  1479. thus covering a 10.5-kb region centered on the TSS with 21 windows.
  1480. Reads in each window for each TSS were counted in each sample, and the
  1481. counts were normalized and converted to log CPM as in the differential
  1482. modification analysis.
  1483. Then, the logCPM values within each promoter were normalized to an average
  1484. of zero, such that each window's normalized abundance now represents the
  1485. relative read depth of that window compared to all other windows in the
  1486. same promoter.
  1487. The normalized abundance values for each window in a promoter are collectively
  1488. referred to as that promoter's
  1489. \begin_inset Quotes eld
  1490. \end_inset
  1491. relative coverage profile
  1492. \begin_inset Quotes erd
  1493. \end_inset
  1494. .
  1495. \end_layout
  1496. \begin_layout Subsection
  1497. MOFA recovers biologically relevant variation from blind analysis by correlating
  1498. across datasets
  1499. \end_layout
  1500. \begin_layout Standard
  1501. \begin_inset ERT
  1502. status open
  1503. \begin_layout Plain Layout
  1504. \backslash
  1505. afterpage{
  1506. \end_layout
  1507. \begin_layout Plain Layout
  1508. \backslash
  1509. begin{landscape}
  1510. \end_layout
  1511. \end_inset
  1512. \end_layout
  1513. \begin_layout Standard
  1514. \begin_inset Float figure
  1515. wide false
  1516. sideways false
  1517. status open
  1518. \begin_layout Plain Layout
  1519. \begin_inset Float figure
  1520. wide false
  1521. sideways false
  1522. status open
  1523. \begin_layout Plain Layout
  1524. \align center
  1525. \begin_inset Graphics
  1526. filename graphics/CD4-csaw/MOFA-varExplaiend-matrix-CROP.png
  1527. lyxscale 25
  1528. width 45col%
  1529. groupId mofa-subfig
  1530. \end_inset
  1531. \end_layout
  1532. \begin_layout Plain Layout
  1533. \begin_inset Caption Standard
  1534. \begin_layout Plain Layout
  1535. \series bold
  1536. \begin_inset CommandInset label
  1537. LatexCommand label
  1538. name "fig:mofa-varexplained"
  1539. \end_inset
  1540. Variance explained in each data set by each latent factor estimated by MOFA.
  1541. \series default
  1542. For each latent factor (LF) learned by MOFA, the variance explained by
  1543. that factor in each data set (
  1544. \begin_inset Quotes eld
  1545. \end_inset
  1546. view
  1547. \begin_inset Quotes erd
  1548. \end_inset
  1549. ) is shown by the shading of the cells in the lower section.
  1550. The upper section shows the total fraction of each data set's variance
  1551. that is explained by all LFs combined.
  1552. \end_layout
  1553. \end_inset
  1554. \end_layout
  1555. \end_inset
  1556. \begin_inset space \hfill{}
  1557. \end_inset
  1558. \begin_inset Float figure
  1559. wide false
  1560. sideways false
  1561. status open
  1562. \begin_layout Plain Layout
  1563. \align center
  1564. \begin_inset Graphics
  1565. filename graphics/CD4-csaw/MOFA-LF-scatter-CROP.png
  1566. lyxscale 25
  1567. width 45col%
  1568. groupId mofa-subfig
  1569. \end_inset
  1570. \end_layout
  1571. \begin_layout Plain Layout
  1572. \begin_inset Caption Standard
  1573. \begin_layout Plain Layout
  1574. \series bold
  1575. \begin_inset CommandInset label
  1576. LatexCommand label
  1577. name "fig:mofa-lf-scatter"
  1578. \end_inset
  1579. Scatter plots of specific pairs of MOFA latent factors.
  1580. \series default
  1581. LFs 1, 4, and 5 explain substantial variation in all data sets, so they
  1582. are plotted against each other in order to reveal patterns of variation
  1583. that are shared across all data sets.
  1584. \end_layout
  1585. \end_inset
  1586. \end_layout
  1587. \end_inset
  1588. \end_layout
  1589. \begin_layout Plain Layout
  1590. \begin_inset Caption Standard
  1591. \begin_layout Plain Layout
  1592. \series bold
  1593. \begin_inset CommandInset label
  1594. LatexCommand label
  1595. name "fig:MOFA-master"
  1596. \end_inset
  1597. MOFA latent factors separate technical confounders from
  1598. \end_layout
  1599. \end_inset
  1600. \end_layout
  1601. \end_inset
  1602. \end_layout
  1603. \begin_layout Standard
  1604. \begin_inset ERT
  1605. status open
  1606. \begin_layout Plain Layout
  1607. \backslash
  1608. end{landscape}
  1609. \end_layout
  1610. \begin_layout Plain Layout
  1611. }
  1612. \end_layout
  1613. \end_inset
  1614. \end_layout
  1615. \begin_layout Standard
  1616. MOFA was run on all the ChIP-seq windows overlapping consensus peaks for
  1617. each histone mark, as well as the RNA-seq data, in order to identify patterns
  1618. of coordinated variation across all data sets
  1619. \begin_inset CommandInset citation
  1620. LatexCommand cite
  1621. key "Argelaguet2018"
  1622. literal "false"
  1623. \end_inset
  1624. .
  1625. The results are summarized in Figure
  1626. \begin_inset CommandInset ref
  1627. LatexCommand ref
  1628. reference "fig:MOFA-master"
  1629. plural "false"
  1630. caps "false"
  1631. noprefix "false"
  1632. \end_inset
  1633. .
  1634. Latent factors 1, 4, and 5 were determined to explain the most variation
  1635. consistently across all data sets (Fgure
  1636. \begin_inset CommandInset ref
  1637. LatexCommand ref
  1638. reference "fig:mofa-varexplained"
  1639. plural "false"
  1640. caps "false"
  1641. noprefix "false"
  1642. \end_inset
  1643. ), and scatter plots of these factors show that they also correlate best
  1644. with the experimental factors (Figure
  1645. \begin_inset CommandInset ref
  1646. LatexCommand ref
  1647. reference "fig:mofa-lf-scatter"
  1648. plural "false"
  1649. caps "false"
  1650. noprefix "false"
  1651. \end_inset
  1652. ).
  1653. Latent factor 2 captures the batch effect in the RNA-seq data.
  1654. Removing the effect of LF2 using MOFA theoretically yields a batch correction
  1655. that does not depend on knowing the experimental factors.
  1656. When this was attempted, the resulting batch correction was comparable
  1657. to ComBat (see Figure
  1658. \begin_inset CommandInset ref
  1659. LatexCommand ref
  1660. reference "fig:RNA-PCA-ComBat-batchsub"
  1661. plural "false"
  1662. caps "false"
  1663. noprefix "false"
  1664. \end_inset
  1665. ), indicating that the ComBat-based batch correction has little room for
  1666. improvement given the problems with the data set.
  1667. \end_layout
  1668. \begin_layout Standard
  1669. \begin_inset Note Note
  1670. status collapsed
  1671. \begin_layout Plain Layout
  1672. \begin_inset Float figure
  1673. wide false
  1674. sideways false
  1675. status open
  1676. \begin_layout Plain Layout
  1677. \align center
  1678. \begin_inset Graphics
  1679. filename graphics/CD4-csaw/MOFA-batch-correct-CROP.png
  1680. lyxscale 25
  1681. width 100col%
  1682. groupId colwidth-raster
  1683. \end_inset
  1684. \end_layout
  1685. \begin_layout Plain Layout
  1686. \begin_inset Caption Standard
  1687. \begin_layout Plain Layout
  1688. \series bold
  1689. \begin_inset CommandInset label
  1690. LatexCommand label
  1691. name "fig:mofa-batchsub"
  1692. \end_inset
  1693. Result of RNA-seq batch-correction using MOFA latent factors
  1694. \end_layout
  1695. \end_inset
  1696. \end_layout
  1697. \end_inset
  1698. \end_layout
  1699. \end_inset
  1700. \end_layout
  1701. \begin_layout Section
  1702. Results
  1703. \end_layout
  1704. \begin_layout Standard
  1705. \begin_inset Flex TODO Note (inline)
  1706. status open
  1707. \begin_layout Plain Layout
  1708. Focus on what hypotheses were tested, then select figures that show how
  1709. those hypotheses were tested, even if the result is a negative.
  1710. Not every interesting result needs to be in here.
  1711. Chapter should tell a story.
  1712. \end_layout
  1713. \end_inset
  1714. \end_layout
  1715. \begin_layout Standard
  1716. \begin_inset Flex TODO Note (inline)
  1717. status open
  1718. \begin_layout Plain Layout
  1719. Maybe reorder these sections to do RNA-seq, then ChIP-seq, then combined
  1720. analyses?
  1721. \end_layout
  1722. \end_inset
  1723. \end_layout
  1724. \begin_layout Subsection
  1725. Interpretation of RNA-seq analysis is limited by a major confounding factor
  1726. \end_layout
  1727. \begin_layout Standard
  1728. \begin_inset Float table
  1729. wide false
  1730. sideways false
  1731. status collapsed
  1732. \begin_layout Plain Layout
  1733. \align center
  1734. \begin_inset Tabular
  1735. <lyxtabular version="3" rows="11" columns="3">
  1736. <features tabularvalignment="middle">
  1737. <column alignment="center" valignment="top">
  1738. <column alignment="center" valignment="top">
  1739. <column alignment="center" valignment="top">
  1740. <row>
  1741. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  1742. \begin_inset Text
  1743. \begin_layout Plain Layout
  1744. Test
  1745. \end_layout
  1746. \end_inset
  1747. </cell>
  1748. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  1749. \begin_inset Text
  1750. \begin_layout Plain Layout
  1751. Est.
  1752. non-null
  1753. \end_layout
  1754. \end_inset
  1755. </cell>
  1756. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  1757. \begin_inset Text
  1758. \begin_layout Plain Layout
  1759. \begin_inset Formula $\mathrm{FDR}\le10\%$
  1760. \end_inset
  1761. \end_layout
  1762. \end_inset
  1763. </cell>
  1764. </row>
  1765. <row>
  1766. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  1767. \begin_inset Text
  1768. \begin_layout Plain Layout
  1769. Naive Day 0 vs Day 1
  1770. \end_layout
  1771. \end_inset
  1772. </cell>
  1773. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  1774. \begin_inset Text
  1775. \begin_layout Plain Layout
  1776. 5992
  1777. \end_layout
  1778. \end_inset
  1779. </cell>
  1780. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  1781. \begin_inset Text
  1782. \begin_layout Plain Layout
  1783. 1613
  1784. \end_layout
  1785. \end_inset
  1786. </cell>
  1787. </row>
  1788. <row>
  1789. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  1790. \begin_inset Text
  1791. \begin_layout Plain Layout
  1792. Naive Day 0 vs Day 5
  1793. \end_layout
  1794. \end_inset
  1795. </cell>
  1796. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  1797. \begin_inset Text
  1798. \begin_layout Plain Layout
  1799. 3038
  1800. \end_layout
  1801. \end_inset
  1802. </cell>
  1803. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  1804. \begin_inset Text
  1805. \begin_layout Plain Layout
  1806. 32
  1807. \end_layout
  1808. \end_inset
  1809. </cell>
  1810. </row>
  1811. <row>
  1812. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  1813. \begin_inset Text
  1814. \begin_layout Plain Layout
  1815. Naive Day 0 vs Day 14
  1816. \end_layout
  1817. \end_inset
  1818. </cell>
  1819. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  1820. \begin_inset Text
  1821. \begin_layout Plain Layout
  1822. 1870
  1823. \end_layout
  1824. \end_inset
  1825. </cell>
  1826. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  1827. \begin_inset Text
  1828. \begin_layout Plain Layout
  1829. 190
  1830. \end_layout
  1831. \end_inset
  1832. </cell>
  1833. </row>
  1834. <row>
  1835. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  1836. \begin_inset Text
  1837. \begin_layout Plain Layout
  1838. Memory Day 0 vs Day 1
  1839. \end_layout
  1840. \end_inset
  1841. </cell>
  1842. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  1843. \begin_inset Text
  1844. \begin_layout Plain Layout
  1845. 3195
  1846. \end_layout
  1847. \end_inset
  1848. </cell>
  1849. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  1850. \begin_inset Text
  1851. \begin_layout Plain Layout
  1852. 411
  1853. \end_layout
  1854. \end_inset
  1855. </cell>
  1856. </row>
  1857. <row>
  1858. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  1859. \begin_inset Text
  1860. \begin_layout Plain Layout
  1861. Memory Day 0 vs Day 5
  1862. \end_layout
  1863. \end_inset
  1864. </cell>
  1865. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  1866. \begin_inset Text
  1867. \begin_layout Plain Layout
  1868. 2688
  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. 18
  1876. \end_layout
  1877. \end_inset
  1878. </cell>
  1879. </row>
  1880. <row>
  1881. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  1882. \begin_inset Text
  1883. \begin_layout Plain Layout
  1884. Memory Day 0 vs Day 14
  1885. \end_layout
  1886. \end_inset
  1887. </cell>
  1888. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  1889. \begin_inset Text
  1890. \begin_layout Plain Layout
  1891. 1911
  1892. \end_layout
  1893. \end_inset
  1894. </cell>
  1895. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  1896. \begin_inset Text
  1897. \begin_layout Plain Layout
  1898. 227
  1899. \end_layout
  1900. \end_inset
  1901. </cell>
  1902. </row>
  1903. <row>
  1904. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  1905. \begin_inset Text
  1906. \begin_layout Plain Layout
  1907. Day 0 Naive vs Memory
  1908. \end_layout
  1909. \end_inset
  1910. </cell>
  1911. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  1912. \begin_inset Text
  1913. \begin_layout Plain Layout
  1914. 0
  1915. \end_layout
  1916. \end_inset
  1917. </cell>
  1918. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  1919. \begin_inset Text
  1920. \begin_layout Plain Layout
  1921. 2
  1922. \end_layout
  1923. \end_inset
  1924. </cell>
  1925. </row>
  1926. <row>
  1927. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  1928. \begin_inset Text
  1929. \begin_layout Plain Layout
  1930. Day 1 Naive vs Memory
  1931. \end_layout
  1932. \end_inset
  1933. </cell>
  1934. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  1935. \begin_inset Text
  1936. \begin_layout Plain Layout
  1937. 9167
  1938. \end_layout
  1939. \end_inset
  1940. </cell>
  1941. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  1942. \begin_inset Text
  1943. \begin_layout Plain Layout
  1944. 5532
  1945. \end_layout
  1946. \end_inset
  1947. </cell>
  1948. </row>
  1949. <row>
  1950. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  1951. \begin_inset Text
  1952. \begin_layout Plain Layout
  1953. Day 5 Naive vs Memory
  1954. \end_layout
  1955. \end_inset
  1956. </cell>
  1957. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  1958. \begin_inset Text
  1959. \begin_layout Plain Layout
  1960. 0
  1961. \end_layout
  1962. \end_inset
  1963. </cell>
  1964. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  1965. \begin_inset Text
  1966. \begin_layout Plain Layout
  1967. 0
  1968. \end_layout
  1969. \end_inset
  1970. </cell>
  1971. </row>
  1972. <row>
  1973. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  1974. \begin_inset Text
  1975. \begin_layout Plain Layout
  1976. Day 14 Naive vs Memory
  1977. \end_layout
  1978. \end_inset
  1979. </cell>
  1980. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  1981. \begin_inset Text
  1982. \begin_layout Plain Layout
  1983. 6446
  1984. \end_layout
  1985. \end_inset
  1986. </cell>
  1987. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  1988. \begin_inset Text
  1989. \begin_layout Plain Layout
  1990. 2319
  1991. \end_layout
  1992. \end_inset
  1993. </cell>
  1994. </row>
  1995. </lyxtabular>
  1996. \end_inset
  1997. \end_layout
  1998. \begin_layout Plain Layout
  1999. \begin_inset Caption Standard
  2000. \begin_layout Plain Layout
  2001. \series bold
  2002. \begin_inset CommandInset label
  2003. LatexCommand label
  2004. name "tab:Estimated-and-detected-rnaseq"
  2005. \end_inset
  2006. Estimated and detected differentially expressed genes.
  2007. \series default
  2008. \begin_inset Quotes eld
  2009. \end_inset
  2010. Test
  2011. \begin_inset Quotes erd
  2012. \end_inset
  2013. : Which sample groups were compared;
  2014. \begin_inset Quotes eld
  2015. \end_inset
  2016. Est non-null
  2017. \begin_inset Quotes erd
  2018. \end_inset
  2019. : Estimated number of differentially expressed genes, using the method of
  2020. averaging local FDR values
  2021. \begin_inset CommandInset citation
  2022. LatexCommand cite
  2023. key "Phipson2013Thesis"
  2024. literal "false"
  2025. \end_inset
  2026. ;
  2027. \begin_inset Quotes eld
  2028. \end_inset
  2029. \begin_inset Formula $\mathrm{FDR}\le10\%$
  2030. \end_inset
  2031. \begin_inset Quotes erd
  2032. \end_inset
  2033. : Number of significantly differentially expressed genes at an FDR threshold
  2034. of 10%.
  2035. The total number of genes tested was 16707.
  2036. \end_layout
  2037. \end_inset
  2038. \end_layout
  2039. \end_inset
  2040. \end_layout
  2041. \begin_layout Standard
  2042. \begin_inset Float figure
  2043. wide false
  2044. sideways false
  2045. status collapsed
  2046. \begin_layout Plain Layout
  2047. \align center
  2048. \begin_inset Graphics
  2049. filename graphics/CD4-csaw/RNA-seq/PCA-final-12-CROP.png
  2050. lyxscale 25
  2051. width 100col%
  2052. groupId colwidth-raster
  2053. \end_inset
  2054. \end_layout
  2055. \begin_layout Plain Layout
  2056. \begin_inset Caption Standard
  2057. \begin_layout Plain Layout
  2058. \series bold
  2059. \begin_inset CommandInset label
  2060. LatexCommand label
  2061. name "fig:rna-pca-final"
  2062. \end_inset
  2063. PCoA plot of RNA-seq samples after ComBat batch correction.
  2064. \series default
  2065. Each point represents an individual sample.
  2066. Samples with the same combination of cell type and time point are encircled
  2067. with a shaded region to aid in visual identification of the sample groups.
  2068. Samples with of same cell type from the same donor are connected by lines
  2069. to indicate the
  2070. \begin_inset Quotes eld
  2071. \end_inset
  2072. trajectory
  2073. \begin_inset Quotes erd
  2074. \end_inset
  2075. of each donor's cells over time in PCoA space.
  2076. \end_layout
  2077. \end_inset
  2078. \end_layout
  2079. \begin_layout Plain Layout
  2080. \end_layout
  2081. \end_inset
  2082. \end_layout
  2083. \begin_layout Standard
  2084. Genes called present in the RNA-seq data were tested for differential expression
  2085. between all time points and cell types.
  2086. The counts of differentially expressed genes are shown in Table
  2087. \begin_inset CommandInset ref
  2088. LatexCommand ref
  2089. reference "tab:Estimated-and-detected-rnaseq"
  2090. plural "false"
  2091. caps "false"
  2092. noprefix "false"
  2093. \end_inset
  2094. .
  2095. Notably, all the results for Day 0 and Day 5 have substantially fewer genes
  2096. called differentially expressed than any of the results for other time
  2097. points.
  2098. This is an unfortunate result of the difference in sample quality between
  2099. the two batches of RNA-seq data.
  2100. All the samples in Batch 1, which includes all the samples from Days 0
  2101. and 5, have substantially more variability than the samples in Batch 2,
  2102. which includes the other time points.
  2103. This is reflected in the substantially higher weights assigned to Batch
  2104. 2 (Figure
  2105. \begin_inset CommandInset ref
  2106. LatexCommand ref
  2107. reference "fig:RNA-seq-weights-vs-covars"
  2108. plural "false"
  2109. caps "false"
  2110. noprefix "false"
  2111. \end_inset
  2112. ).
  2113. The batch effect has both a systematic component and a random noise component.
  2114. While the systematic component was subtracted out using ComBat (Figure
  2115. \begin_inset CommandInset ref
  2116. LatexCommand ref
  2117. reference "fig:RNA-PCA"
  2118. plural "false"
  2119. caps "false"
  2120. noprefix "false"
  2121. \end_inset
  2122. ), no such correction is possible for the noise component: Batch 1 simply
  2123. has substantially more random noise in it, which reduces the statistical
  2124. power for any differential expression tests involving samples in that batch.
  2125. \end_layout
  2126. \begin_layout Standard
  2127. Despite the difficulty in detecting specific differentially expressed genes,
  2128. there is still evidence that differential expression is present for these
  2129. time points.
  2130. In Figure
  2131. \begin_inset CommandInset ref
  2132. LatexCommand ref
  2133. reference "fig:rna-pca-final"
  2134. plural "false"
  2135. caps "false"
  2136. noprefix "false"
  2137. \end_inset
  2138. , there is a clear separation between naive and memory samples at Day 0,
  2139. despite the fact that only 2 genes were significantly differentially expressed
  2140. for this comparison.
  2141. Similarly, the small numbers of genes detected for the Day 0 vs Day 5 compariso
  2142. ns do not reflect the large separation between these time points in Figure
  2143. \begin_inset CommandInset ref
  2144. LatexCommand ref
  2145. reference "fig:rna-pca-final"
  2146. plural "false"
  2147. caps "false"
  2148. noprefix "false"
  2149. \end_inset
  2150. .
  2151. In addition, the MOFA latent factor plots in Figure
  2152. \begin_inset CommandInset ref
  2153. LatexCommand ref
  2154. reference "fig:mofa-lf-scatter"
  2155. plural "false"
  2156. caps "false"
  2157. noprefix "false"
  2158. \end_inset
  2159. .
  2160. This suggests that there is indeed a differential expression signal present
  2161. in the data for these comparisons, but the large variability in the Batch
  2162. 1 samples obfuscates this signal at the individual gene level.
  2163. As a result, it is impossible to make any meaningful statements about the
  2164. \begin_inset Quotes eld
  2165. \end_inset
  2166. size
  2167. \begin_inset Quotes erd
  2168. \end_inset
  2169. of the gene signature for any time point, since the number of significant
  2170. genes as well as the estimated number of differentially expressed genes
  2171. depends so strongly on the variations in sample quality in addition to
  2172. the size of the differential expression signal in the data.
  2173. Gene-set enrichment analyses are similarly impractical.
  2174. However, analyses looking at genome-wide patterns of expression are still
  2175. practical.
  2176. \end_layout
  2177. \begin_layout Subsection
  2178. H3K4 and H3K27 methylation occur in broad regions and are enriched near
  2179. promoters
  2180. \end_layout
  2181. \begin_layout Standard
  2182. \begin_inset Float table
  2183. wide false
  2184. sideways false
  2185. status collapsed
  2186. \begin_layout Plain Layout
  2187. \align center
  2188. \begin_inset Flex TODO Note (inline)
  2189. status open
  2190. \begin_layout Plain Layout
  2191. Also get
  2192. \emph on
  2193. median
  2194. \emph default
  2195. peak width and maybe other quantiles (25%, 75%)
  2196. \end_layout
  2197. \end_inset
  2198. \end_layout
  2199. \begin_layout Plain Layout
  2200. \align center
  2201. \begin_inset Tabular
  2202. <lyxtabular version="3" rows="4" columns="5">
  2203. <features tabularvalignment="middle">
  2204. <column alignment="center" valignment="top">
  2205. <column alignment="center" valignment="top">
  2206. <column alignment="center" valignment="top">
  2207. <column alignment="center" valignment="top">
  2208. <column alignment="center" valignment="top">
  2209. <row>
  2210. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  2211. \begin_inset Text
  2212. \begin_layout Plain Layout
  2213. Histone Mark
  2214. \end_layout
  2215. \end_inset
  2216. </cell>
  2217. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  2218. \begin_inset Text
  2219. \begin_layout Plain Layout
  2220. # Peaks
  2221. \end_layout
  2222. \end_inset
  2223. </cell>
  2224. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  2225. \begin_inset Text
  2226. \begin_layout Plain Layout
  2227. Mean peak width
  2228. \end_layout
  2229. \end_inset
  2230. </cell>
  2231. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  2232. \begin_inset Text
  2233. \begin_layout Plain Layout
  2234. genome coverage
  2235. \end_layout
  2236. \end_inset
  2237. </cell>
  2238. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  2239. \begin_inset Text
  2240. \begin_layout Plain Layout
  2241. FRiP
  2242. \end_layout
  2243. \end_inset
  2244. </cell>
  2245. </row>
  2246. <row>
  2247. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  2248. \begin_inset Text
  2249. \begin_layout Plain Layout
  2250. H3K4me2
  2251. \end_layout
  2252. \end_inset
  2253. </cell>
  2254. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  2255. \begin_inset Text
  2256. \begin_layout Plain Layout
  2257. 14965
  2258. \end_layout
  2259. \end_inset
  2260. </cell>
  2261. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  2262. \begin_inset Text
  2263. \begin_layout Plain Layout
  2264. 3970
  2265. \end_layout
  2266. \end_inset
  2267. </cell>
  2268. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  2269. \begin_inset Text
  2270. \begin_layout Plain Layout
  2271. 1.92%
  2272. \end_layout
  2273. \end_inset
  2274. </cell>
  2275. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  2276. \begin_inset Text
  2277. \begin_layout Plain Layout
  2278. 14.2%
  2279. \end_layout
  2280. \end_inset
  2281. </cell>
  2282. </row>
  2283. <row>
  2284. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  2285. \begin_inset Text
  2286. \begin_layout Plain Layout
  2287. H3K4me3
  2288. \end_layout
  2289. \end_inset
  2290. </cell>
  2291. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  2292. \begin_inset Text
  2293. \begin_layout Plain Layout
  2294. 6163
  2295. \end_layout
  2296. \end_inset
  2297. </cell>
  2298. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  2299. \begin_inset Text
  2300. \begin_layout Plain Layout
  2301. 2946
  2302. \end_layout
  2303. \end_inset
  2304. </cell>
  2305. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  2306. \begin_inset Text
  2307. \begin_layout Plain Layout
  2308. 0.588%
  2309. \end_layout
  2310. \end_inset
  2311. </cell>
  2312. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  2313. \begin_inset Text
  2314. \begin_layout Plain Layout
  2315. 6.57%
  2316. \end_layout
  2317. \end_inset
  2318. </cell>
  2319. </row>
  2320. <row>
  2321. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  2322. \begin_inset Text
  2323. \begin_layout Plain Layout
  2324. H3K27me3
  2325. \end_layout
  2326. \end_inset
  2327. </cell>
  2328. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  2329. \begin_inset Text
  2330. \begin_layout Plain Layout
  2331. 18139
  2332. \end_layout
  2333. \end_inset
  2334. </cell>
  2335. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  2336. \begin_inset Text
  2337. \begin_layout Plain Layout
  2338. 18967
  2339. \end_layout
  2340. \end_inset
  2341. </cell>
  2342. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  2343. \begin_inset Text
  2344. \begin_layout Plain Layout
  2345. 11.1%
  2346. \end_layout
  2347. \end_inset
  2348. </cell>
  2349. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  2350. \begin_inset Text
  2351. \begin_layout Plain Layout
  2352. 22.5%
  2353. \end_layout
  2354. \end_inset
  2355. </cell>
  2356. </row>
  2357. </lyxtabular>
  2358. \end_inset
  2359. \end_layout
  2360. \begin_layout Plain Layout
  2361. \begin_inset Caption Standard
  2362. \begin_layout Plain Layout
  2363. \series bold
  2364. \begin_inset CommandInset label
  2365. LatexCommand label
  2366. name "tab:peak-calling-summary"
  2367. \end_inset
  2368. Peak-calling summary.
  2369. \series default
  2370. For each histone mark, the number of peaks called using SICER at an IDR
  2371. threshold of ???, the mean width of those peaks, the fraction of the genome
  2372. covered by peaks, and the fraction of reads in peaks (FRiP).
  2373. \end_layout
  2374. \end_inset
  2375. \end_layout
  2376. \end_inset
  2377. \end_layout
  2378. \begin_layout Standard
  2379. Table
  2380. \begin_inset CommandInset ref
  2381. LatexCommand ref
  2382. reference "tab:peak-calling-summary"
  2383. plural "false"
  2384. caps "false"
  2385. noprefix "false"
  2386. \end_inset
  2387. gives a summary of the peak calling statistics for each histone mark.
  2388. Consistent with previous observations [CITATION NEEDED], all 3 histone
  2389. marks occur in broad regions spanning many consecutive nucleosomes, rather
  2390. than in sharp peaks as would be expected for a transcription factor or
  2391. other molecule that binds to specific sites.
  2392. This conclusion is further supported by Figure
  2393. \begin_inset CommandInset ref
  2394. LatexCommand ref
  2395. reference "fig:CCF-with-blacklist"
  2396. plural "false"
  2397. caps "false"
  2398. noprefix "false"
  2399. \end_inset
  2400. , in which a clear nucleosome-sized periodicity is visible in the cross-correlat
  2401. ion value for each sample, indicating that each time a given mark is present
  2402. on one histone, it is also likely to be found on adjacent histones as well.
  2403. H3K27me3 enrichment in particular is substantially more broad than either
  2404. H3K4 mark, with a mean peak width of almost 19,000 bp.
  2405. This is also reflected in the periodicity observed in Figure
  2406. \begin_inset CommandInset ref
  2407. LatexCommand ref
  2408. reference "fig:CCF-with-blacklist"
  2409. plural "false"
  2410. caps "false"
  2411. noprefix "false"
  2412. \end_inset
  2413. , which remains strong much farther out for H3K27me3 than the other marks,
  2414. showing H3K27me3 especially tends to be found on long runs of consecutive
  2415. histones.
  2416. \end_layout
  2417. \begin_layout Standard
  2418. \begin_inset Float figure
  2419. wide false
  2420. sideways false
  2421. status open
  2422. \begin_layout Plain Layout
  2423. \begin_inset Flex TODO Note (inline)
  2424. status open
  2425. \begin_layout Plain Layout
  2426. Ensure this figure uses the peak calls from the new analysis.
  2427. \end_layout
  2428. \end_inset
  2429. \end_layout
  2430. \begin_layout Plain Layout
  2431. \begin_inset Flex TODO Note (inline)
  2432. status open
  2433. \begin_layout Plain Layout
  2434. Need a control: shuffle all peaks and repeat, N times.
  2435. Do real vs shuffled control both in a top/bottom arrangement.
  2436. \end_layout
  2437. \end_inset
  2438. \end_layout
  2439. \begin_layout Plain Layout
  2440. \begin_inset Flex TODO Note (inline)
  2441. status open
  2442. \begin_layout Plain Layout
  2443. Consider counting TSS inside peaks as negative number indicating how far
  2444. \emph on
  2445. inside
  2446. \emph default
  2447. the peak the TSS is (i.e.
  2448. distance to nearest non-peak area).
  2449. \end_layout
  2450. \end_inset
  2451. \end_layout
  2452. \begin_layout Plain Layout
  2453. \begin_inset Flex TODO Note (inline)
  2454. status open
  2455. \begin_layout Plain Layout
  2456. The H3K4 part of this figure is included in
  2457. \begin_inset CommandInset citation
  2458. LatexCommand cite
  2459. key "LaMere2016"
  2460. literal "false"
  2461. \end_inset
  2462. as Fig.
  2463. S2.
  2464. Do I need to do anything about that?
  2465. \end_layout
  2466. \end_inset
  2467. \end_layout
  2468. \begin_layout Plain Layout
  2469. \align center
  2470. \begin_inset Graphics
  2471. filename graphics/CD4-csaw/Promoter Peak Distance Profile-PAGE1-CROP.pdf
  2472. lyxscale 50
  2473. width 80col%
  2474. \end_inset
  2475. \end_layout
  2476. \begin_layout Plain Layout
  2477. \begin_inset Caption Standard
  2478. \begin_layout Plain Layout
  2479. \series bold
  2480. \begin_inset CommandInset label
  2481. LatexCommand label
  2482. name "fig:near-promoter-peak-enrich"
  2483. \end_inset
  2484. Enrichment of peaks in promoter neighborhoods.
  2485. \series default
  2486. This plot shows the distribution of distances from each annotated transcription
  2487. start site in the genome to the nearest called peak.
  2488. Each line represents one combination of histone mark, cell type, and time
  2489. point.
  2490. Distributions are smoothed using kernel density estimation [CITE? see ggplot2
  2491. stat_density()].
  2492. Transcription start sites that occur
  2493. \emph on
  2494. within
  2495. \emph default
  2496. peaks were excluded from this plot to avoid a large spike at zero that
  2497. would overshadow the rest of the distribution.
  2498. \end_layout
  2499. \end_inset
  2500. \end_layout
  2501. \end_inset
  2502. \end_layout
  2503. \begin_layout Standard
  2504. \begin_inset Float table
  2505. wide false
  2506. sideways false
  2507. status collapsed
  2508. \begin_layout Plain Layout
  2509. \align center
  2510. \begin_inset Tabular
  2511. <lyxtabular version="3" rows="4" columns="2">
  2512. <features tabularvalignment="middle">
  2513. <column alignment="center" valignment="top">
  2514. <column alignment="center" valignment="top">
  2515. <row>
  2516. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  2517. \begin_inset Text
  2518. \begin_layout Plain Layout
  2519. Histone mark
  2520. \end_layout
  2521. \end_inset
  2522. </cell>
  2523. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  2524. \begin_inset Text
  2525. \begin_layout Plain Layout
  2526. Effective promoter radius
  2527. \end_layout
  2528. \end_inset
  2529. </cell>
  2530. </row>
  2531. <row>
  2532. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  2533. \begin_inset Text
  2534. \begin_layout Plain Layout
  2535. H3K4me2
  2536. \end_layout
  2537. \end_inset
  2538. </cell>
  2539. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  2540. \begin_inset Text
  2541. \begin_layout Plain Layout
  2542. 1 kb
  2543. \end_layout
  2544. \end_inset
  2545. </cell>
  2546. </row>
  2547. <row>
  2548. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  2549. \begin_inset Text
  2550. \begin_layout Plain Layout
  2551. H3K4me3
  2552. \end_layout
  2553. \end_inset
  2554. </cell>
  2555. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  2556. \begin_inset Text
  2557. \begin_layout Plain Layout
  2558. 1 kb
  2559. \end_layout
  2560. \end_inset
  2561. </cell>
  2562. </row>
  2563. <row>
  2564. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  2565. \begin_inset Text
  2566. \begin_layout Plain Layout
  2567. H3K27me3
  2568. \end_layout
  2569. \end_inset
  2570. </cell>
  2571. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  2572. \begin_inset Text
  2573. \begin_layout Plain Layout
  2574. 2.5 kb
  2575. \end_layout
  2576. \end_inset
  2577. </cell>
  2578. </row>
  2579. </lyxtabular>
  2580. \end_inset
  2581. \end_layout
  2582. \begin_layout Plain Layout
  2583. \begin_inset Caption Standard
  2584. \begin_layout Plain Layout
  2585. \series bold
  2586. \begin_inset CommandInset label
  2587. LatexCommand label
  2588. name "tab:effective-promoter-radius"
  2589. \end_inset
  2590. Effective promoter radius for each histone mark.
  2591. \series default
  2592. These values represent the approximate distance from transcription start
  2593. site positions within which an excess of peaks are found, as shown in Figure
  2594. \begin_inset CommandInset ref
  2595. LatexCommand ref
  2596. reference "fig:near-promoter-peak-enrich"
  2597. plural "false"
  2598. caps "false"
  2599. noprefix "false"
  2600. \end_inset
  2601. .
  2602. \end_layout
  2603. \end_inset
  2604. \end_layout
  2605. \begin_layout Plain Layout
  2606. \end_layout
  2607. \end_inset
  2608. \end_layout
  2609. \begin_layout Standard
  2610. All 3 histone marks tend to occur more often near promoter regions, as shown
  2611. in Figure
  2612. \begin_inset CommandInset ref
  2613. LatexCommand ref
  2614. reference "fig:near-promoter-peak-enrich"
  2615. plural "false"
  2616. caps "false"
  2617. noprefix "false"
  2618. \end_inset
  2619. .
  2620. The majority of each density distribution is flat, representing the background
  2621. density of peaks genome-wide.
  2622. Each distribution has a peak near zero, representing an enrichment of peaks
  2623. close transcription start site (TSS) positions relative to the remainder
  2624. of the genome.
  2625. Interestingly, the
  2626. \begin_inset Quotes eld
  2627. \end_inset
  2628. radius
  2629. \begin_inset Quotes erd
  2630. \end_inset
  2631. within which this enrichment occurs is not the same for every histone mark
  2632. (Table
  2633. \begin_inset CommandInset ref
  2634. LatexCommand ref
  2635. reference "tab:effective-promoter-radius"
  2636. plural "false"
  2637. caps "false"
  2638. noprefix "false"
  2639. \end_inset
  2640. ).
  2641. For H3K4me2 and H3K4me3, peaks are most enriched within 1
  2642. \begin_inset space ~
  2643. \end_inset
  2644. kbp of TSS positions, while for H3K27me3, enrichment is broader, extending
  2645. to 2.5
  2646. \begin_inset space ~
  2647. \end_inset
  2648. kbp.
  2649. These
  2650. \begin_inset Quotes eld
  2651. \end_inset
  2652. effective promoter radii
  2653. \begin_inset Quotes erd
  2654. \end_inset
  2655. remain approximately the same across all combinations of experimental condition
  2656. (cell type, time point, and donor), so they appear to be a property of
  2657. the histone mark itself.
  2658. Hence, these radii were used to define the promoter regions for each histone
  2659. mark in all further analyses.
  2660. \end_layout
  2661. \begin_layout Standard
  2662. \begin_inset Flex TODO Note (inline)
  2663. status open
  2664. \begin_layout Plain Layout
  2665. Consider also showing figure for distance to nearest peak center, and reference
  2666. median peak size once that is known.
  2667. \end_layout
  2668. \end_inset
  2669. \end_layout
  2670. \begin_layout Subsection
  2671. H3K4 and H3K27 promoter methylation has broadly the expected correlation
  2672. with gene expression
  2673. \end_layout
  2674. \begin_layout Standard
  2675. \begin_inset Float figure
  2676. wide false
  2677. sideways false
  2678. status collapsed
  2679. \begin_layout Plain Layout
  2680. \begin_inset Flex TODO Note (inline)
  2681. status open
  2682. \begin_layout Plain Layout
  2683. This figure is generated from the old analysis.
  2684. Eiher note that in some way or re-generate it from the new peak calls.
  2685. \end_layout
  2686. \end_inset
  2687. \end_layout
  2688. \begin_layout Plain Layout
  2689. \align center
  2690. \begin_inset Graphics
  2691. filename graphics/CD4-csaw/FPKM by Peak Violin Plots-CROP.pdf
  2692. lyxscale 50
  2693. width 100col%
  2694. \end_inset
  2695. \end_layout
  2696. \begin_layout Plain Layout
  2697. \begin_inset Caption Standard
  2698. \begin_layout Plain Layout
  2699. \series bold
  2700. \begin_inset CommandInset label
  2701. LatexCommand label
  2702. name "fig:fpkm-by-peak"
  2703. \end_inset
  2704. Expression distributions of genes with and without promoter peaks.
  2705. \end_layout
  2706. \end_inset
  2707. \end_layout
  2708. \end_inset
  2709. \end_layout
  2710. \begin_layout Standard
  2711. H3K4me2 and H3K4me2 have previously been reported as activating marks whose
  2712. presence in a gene's promoter is associated with higher gene expression,
  2713. while H3K27me3 has been reported as inactivating [CITE].
  2714. The data are consistent with this characterization: genes whose promoters
  2715. (as defined by the radii for each histone mark listed in
  2716. \begin_inset CommandInset ref
  2717. LatexCommand ref
  2718. reference "tab:effective-promoter-radius"
  2719. plural "false"
  2720. caps "false"
  2721. noprefix "false"
  2722. \end_inset
  2723. ) overlap with a H3K4me2 or H3K4me3 peak tend to have higher expression
  2724. than those that don't, while H3K27me3 is likewise associated with lower
  2725. gene expression, as shown in
  2726. \begin_inset CommandInset ref
  2727. LatexCommand ref
  2728. reference "fig:fpkm-by-peak"
  2729. plural "false"
  2730. caps "false"
  2731. noprefix "false"
  2732. \end_inset
  2733. .
  2734. This pattern holds across all combinations of cell type and time point
  2735. (Welch's
  2736. \emph on
  2737. t
  2738. \emph default
  2739. -test, all
  2740. \begin_inset Formula $p\mathrm{-values}\ll2.2\times10^{-16}$
  2741. \end_inset
  2742. ).
  2743. The difference in average log FPKM values when a peak overlaps the promoter
  2744. is about
  2745. \begin_inset Formula $+5.67$
  2746. \end_inset
  2747. for H3K4me2,
  2748. \begin_inset Formula $+5.76$
  2749. \end_inset
  2750. for H3K4me2, and
  2751. \begin_inset Formula $-4.00$
  2752. \end_inset
  2753. for H3K27me3.
  2754. \end_layout
  2755. \begin_layout Standard
  2756. \begin_inset Flex TODO Note (inline)
  2757. status open
  2758. \begin_layout Plain Layout
  2759. I also have some figures looking at interactions between marks (e.g.
  2760. what if a promoter has both H3K4me3 and H3K27me3), but I don't know if
  2761. that much detail is warranted here, since all the effects just seem approximate
  2762. ly additive anyway.
  2763. \end_layout
  2764. \end_inset
  2765. \end_layout
  2766. \begin_layout Subsection
  2767. Gene expression and promoter histone methylation patterns in naive and memory
  2768. show convergence at day 14
  2769. \end_layout
  2770. \begin_layout Standard
  2771. \begin_inset ERT
  2772. status open
  2773. \begin_layout Plain Layout
  2774. \backslash
  2775. afterpage{
  2776. \end_layout
  2777. \begin_layout Plain Layout
  2778. \backslash
  2779. begin{landscape}
  2780. \end_layout
  2781. \end_inset
  2782. \end_layout
  2783. \begin_layout Standard
  2784. \begin_inset Float table
  2785. wide false
  2786. sideways false
  2787. status open
  2788. \begin_layout Plain Layout
  2789. \align center
  2790. \begin_inset Tabular
  2791. <lyxtabular version="3" rows="6" columns="7">
  2792. <features tabularvalignment="middle">
  2793. <column alignment="center" valignment="top">
  2794. <column alignment="center" valignment="top">
  2795. <column alignment="center" valignment="top">
  2796. <column alignment="center" valignment="top">
  2797. <column alignment="center" valignment="top">
  2798. <column alignment="center" valignment="top">
  2799. <column alignment="center" valignment="top">
  2800. <row>
  2801. <cell alignment="center" valignment="top" usebox="none">
  2802. \begin_inset Text
  2803. \begin_layout Plain Layout
  2804. \end_layout
  2805. \end_inset
  2806. </cell>
  2807. <cell multicolumn="1" alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  2808. \begin_inset Text
  2809. \begin_layout Plain Layout
  2810. Number of significant promoters
  2811. \end_layout
  2812. \end_inset
  2813. </cell>
  2814. <cell multicolumn="2" alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  2815. \begin_inset Text
  2816. \begin_layout Plain Layout
  2817. \end_layout
  2818. \end_inset
  2819. </cell>
  2820. <cell multicolumn="2" alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  2821. \begin_inset Text
  2822. \begin_layout Plain Layout
  2823. \end_layout
  2824. \end_inset
  2825. </cell>
  2826. <cell multicolumn="1" alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  2827. \begin_inset Text
  2828. \begin_layout Plain Layout
  2829. Est.
  2830. differentially modified promoters
  2831. \end_layout
  2832. \end_inset
  2833. </cell>
  2834. <cell multicolumn="2" alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  2835. \begin_inset Text
  2836. \begin_layout Plain Layout
  2837. \end_layout
  2838. \end_inset
  2839. </cell>
  2840. <cell multicolumn="2" alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  2841. \begin_inset Text
  2842. \begin_layout Plain Layout
  2843. \end_layout
  2844. \end_inset
  2845. </cell>
  2846. </row>
  2847. <row>
  2848. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  2849. \begin_inset Text
  2850. \begin_layout Plain Layout
  2851. Time Point
  2852. \end_layout
  2853. \end_inset
  2854. </cell>
  2855. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  2856. \begin_inset Text
  2857. \begin_layout Plain Layout
  2858. H3K4me2
  2859. \end_layout
  2860. \end_inset
  2861. </cell>
  2862. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  2863. \begin_inset Text
  2864. \begin_layout Plain Layout
  2865. H3K4me3
  2866. \end_layout
  2867. \end_inset
  2868. </cell>
  2869. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  2870. \begin_inset Text
  2871. \begin_layout Plain Layout
  2872. H3K27me3
  2873. \end_layout
  2874. \end_inset
  2875. </cell>
  2876. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  2877. \begin_inset Text
  2878. \begin_layout Plain Layout
  2879. H3K4me2
  2880. \end_layout
  2881. \end_inset
  2882. </cell>
  2883. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  2884. \begin_inset Text
  2885. \begin_layout Plain Layout
  2886. H3K4me3
  2887. \end_layout
  2888. \end_inset
  2889. </cell>
  2890. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  2891. \begin_inset Text
  2892. \begin_layout Plain Layout
  2893. H3K27me3
  2894. \end_layout
  2895. \end_inset
  2896. </cell>
  2897. </row>
  2898. <row>
  2899. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  2900. \begin_inset Text
  2901. \begin_layout Plain Layout
  2902. Day 0
  2903. \end_layout
  2904. \end_inset
  2905. </cell>
  2906. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  2907. \begin_inset Text
  2908. \begin_layout Plain Layout
  2909. 4553
  2910. \end_layout
  2911. \end_inset
  2912. </cell>
  2913. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  2914. \begin_inset Text
  2915. \begin_layout Plain Layout
  2916. 927
  2917. \end_layout
  2918. \end_inset
  2919. </cell>
  2920. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  2921. \begin_inset Text
  2922. \begin_layout Plain Layout
  2923. 6
  2924. \end_layout
  2925. \end_inset
  2926. </cell>
  2927. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  2928. \begin_inset Text
  2929. \begin_layout Plain Layout
  2930. 9967
  2931. \end_layout
  2932. \end_inset
  2933. </cell>
  2934. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  2935. \begin_inset Text
  2936. \begin_layout Plain Layout
  2937. 4149
  2938. \end_layout
  2939. \end_inset
  2940. </cell>
  2941. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  2942. \begin_inset Text
  2943. \begin_layout Plain Layout
  2944. 2404
  2945. \end_layout
  2946. \end_inset
  2947. </cell>
  2948. </row>
  2949. <row>
  2950. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  2951. \begin_inset Text
  2952. \begin_layout Plain Layout
  2953. Day 1
  2954. \end_layout
  2955. \end_inset
  2956. </cell>
  2957. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  2958. \begin_inset Text
  2959. \begin_layout Plain Layout
  2960. 567
  2961. \end_layout
  2962. \end_inset
  2963. </cell>
  2964. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  2965. \begin_inset Text
  2966. \begin_layout Plain Layout
  2967. 278
  2968. \end_layout
  2969. \end_inset
  2970. </cell>
  2971. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  2972. \begin_inset Text
  2973. \begin_layout Plain Layout
  2974. 1570
  2975. \end_layout
  2976. \end_inset
  2977. </cell>
  2978. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  2979. \begin_inset Text
  2980. \begin_layout Plain Layout
  2981. 4370
  2982. \end_layout
  2983. \end_inset
  2984. </cell>
  2985. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  2986. \begin_inset Text
  2987. \begin_layout Plain Layout
  2988. 2145
  2989. \end_layout
  2990. \end_inset
  2991. </cell>
  2992. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  2993. \begin_inset Text
  2994. \begin_layout Plain Layout
  2995. 6598
  2996. \end_layout
  2997. \end_inset
  2998. </cell>
  2999. </row>
  3000. <row>
  3001. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  3002. \begin_inset Text
  3003. \begin_layout Plain Layout
  3004. Day 5
  3005. \end_layout
  3006. \end_inset
  3007. </cell>
  3008. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  3009. \begin_inset Text
  3010. \begin_layout Plain Layout
  3011. 2313
  3012. \end_layout
  3013. \end_inset
  3014. </cell>
  3015. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  3016. \begin_inset Text
  3017. \begin_layout Plain Layout
  3018. 139
  3019. \end_layout
  3020. \end_inset
  3021. </cell>
  3022. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  3023. \begin_inset Text
  3024. \begin_layout Plain Layout
  3025. 490
  3026. \end_layout
  3027. \end_inset
  3028. </cell>
  3029. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  3030. \begin_inset Text
  3031. \begin_layout Plain Layout
  3032. 9450
  3033. \end_layout
  3034. \end_inset
  3035. </cell>
  3036. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  3037. \begin_inset Text
  3038. \begin_layout Plain Layout
  3039. 1148
  3040. \end_layout
  3041. \end_inset
  3042. </cell>
  3043. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  3044. \begin_inset Text
  3045. \begin_layout Plain Layout
  3046. 4141
  3047. \end_layout
  3048. \end_inset
  3049. </cell>
  3050. </row>
  3051. <row>
  3052. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  3053. \begin_inset Text
  3054. \begin_layout Plain Layout
  3055. Day 14
  3056. \end_layout
  3057. \end_inset
  3058. </cell>
  3059. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  3060. \begin_inset Text
  3061. \begin_layout Plain Layout
  3062. 0
  3063. \end_layout
  3064. \end_inset
  3065. </cell>
  3066. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  3067. \begin_inset Text
  3068. \begin_layout Plain Layout
  3069. 0
  3070. \end_layout
  3071. \end_inset
  3072. </cell>
  3073. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  3074. \begin_inset Text
  3075. \begin_layout Plain Layout
  3076. 0
  3077. \end_layout
  3078. \end_inset
  3079. </cell>
  3080. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  3081. \begin_inset Text
  3082. \begin_layout Plain Layout
  3083. 0
  3084. \end_layout
  3085. \end_inset
  3086. </cell>
  3087. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  3088. \begin_inset Text
  3089. \begin_layout Plain Layout
  3090. 0
  3091. \end_layout
  3092. \end_inset
  3093. </cell>
  3094. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  3095. \begin_inset Text
  3096. \begin_layout Plain Layout
  3097. 0
  3098. \end_layout
  3099. \end_inset
  3100. </cell>
  3101. </row>
  3102. </lyxtabular>
  3103. \end_inset
  3104. \end_layout
  3105. \begin_layout Plain Layout
  3106. \begin_inset Caption Standard
  3107. \begin_layout Plain Layout
  3108. \series bold
  3109. \begin_inset CommandInset label
  3110. LatexCommand label
  3111. name "tab:Number-signif-promoters"
  3112. \end_inset
  3113. Number of differentially modified promoters between naive and memory cells
  3114. at each time point after activation.
  3115. \series default
  3116. This table shows both the number of differentially modified promoters detected
  3117. at a 10% FDR threshold (left half), and the total number of differentially
  3118. modified promoters as estimated using the method of
  3119. \begin_inset CommandInset citation
  3120. LatexCommand cite
  3121. key "Phipson2013"
  3122. literal "false"
  3123. \end_inset
  3124. (right half).
  3125. \end_layout
  3126. \end_inset
  3127. \end_layout
  3128. \end_inset
  3129. \end_layout
  3130. \begin_layout Standard
  3131. \begin_inset ERT
  3132. status open
  3133. \begin_layout Plain Layout
  3134. \backslash
  3135. end{landscape}
  3136. \end_layout
  3137. \begin_layout Plain Layout
  3138. }
  3139. \end_layout
  3140. \end_inset
  3141. \end_layout
  3142. \begin_layout Standard
  3143. \begin_inset Float figure
  3144. placement p
  3145. wide false
  3146. sideways false
  3147. status open
  3148. \begin_layout Plain Layout
  3149. \align center
  3150. \begin_inset Float figure
  3151. wide false
  3152. sideways false
  3153. status open
  3154. \begin_layout Plain Layout
  3155. \align center
  3156. \begin_inset Graphics
  3157. filename graphics/CD4-csaw/ChIP-seq/H3K4me2-promoter-PCA-group-CROP.png
  3158. lyxscale 25
  3159. width 45col%
  3160. groupId pcoa-prom-subfig
  3161. \end_inset
  3162. \end_layout
  3163. \begin_layout Plain Layout
  3164. \begin_inset Caption Standard
  3165. \begin_layout Plain Layout
  3166. \series bold
  3167. \begin_inset CommandInset label
  3168. LatexCommand label
  3169. name "fig:PCoA-H3K4me2-prom"
  3170. \end_inset
  3171. PCoA plot of H3K4me2 promoters, after subtracting surrogate variables
  3172. \end_layout
  3173. \end_inset
  3174. \end_layout
  3175. \end_inset
  3176. \begin_inset space \hfill{}
  3177. \end_inset
  3178. \begin_inset Float figure
  3179. wide false
  3180. sideways false
  3181. status open
  3182. \begin_layout Plain Layout
  3183. \align center
  3184. \begin_inset Graphics
  3185. filename graphics/CD4-csaw/ChIP-seq/H3K4me3-promoter-PCA-group-CROP.png
  3186. lyxscale 25
  3187. width 45col%
  3188. groupId pcoa-prom-subfig
  3189. \end_inset
  3190. \end_layout
  3191. \begin_layout Plain Layout
  3192. \begin_inset Caption Standard
  3193. \begin_layout Plain Layout
  3194. \series bold
  3195. \begin_inset CommandInset label
  3196. LatexCommand label
  3197. name "fig:PCoA-H3K4me3-prom"
  3198. \end_inset
  3199. PCoA plot of H3K4me3 promoters, after subtracting surrogate variables
  3200. \end_layout
  3201. \end_inset
  3202. \end_layout
  3203. \end_inset
  3204. \end_layout
  3205. \begin_layout Plain Layout
  3206. \align center
  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/CD4-csaw/ChIP-seq/H3K27me3-promoter-PCA-group-CROP.png
  3215. lyxscale 25
  3216. width 45col%
  3217. groupId pcoa-prom-subfig
  3218. \end_inset
  3219. \end_layout
  3220. \begin_layout Plain Layout
  3221. \begin_inset Caption Standard
  3222. \begin_layout Plain Layout
  3223. \series bold
  3224. \begin_inset CommandInset label
  3225. LatexCommand label
  3226. name "fig:PCoA-H3K27me3-prom"
  3227. \end_inset
  3228. PCoA plot of H3K27me3 promoters, after subtracting surrogate variables
  3229. \end_layout
  3230. \end_inset
  3231. \end_layout
  3232. \end_inset
  3233. \begin_inset space \hfill{}
  3234. \end_inset
  3235. \begin_inset Float figure
  3236. wide false
  3237. sideways false
  3238. status open
  3239. \begin_layout Plain Layout
  3240. \align center
  3241. \begin_inset Graphics
  3242. filename graphics/CD4-csaw/RNA-seq/PCA-final-23-CROP.png
  3243. lyxscale 25
  3244. width 45col%
  3245. groupId pcoa-prom-subfig
  3246. \end_inset
  3247. \end_layout
  3248. \begin_layout Plain Layout
  3249. \begin_inset Caption Standard
  3250. \begin_layout Plain Layout
  3251. \series bold
  3252. \begin_inset CommandInset label
  3253. LatexCommand label
  3254. name "fig:RNA-PCA-group"
  3255. \end_inset
  3256. RNA-seq PCoA showing principal coordiantes 2 and 3.
  3257. \end_layout
  3258. \end_inset
  3259. \end_layout
  3260. \end_inset
  3261. \end_layout
  3262. \begin_layout Plain Layout
  3263. \begin_inset Caption Standard
  3264. \begin_layout Plain Layout
  3265. \series bold
  3266. \begin_inset CommandInset label
  3267. LatexCommand label
  3268. name "fig:PCoA-promoters"
  3269. \end_inset
  3270. PCoA plots for promoter ChIP-seq and expression RNA-seq data
  3271. \end_layout
  3272. \end_inset
  3273. \end_layout
  3274. \end_inset
  3275. \end_layout
  3276. \begin_layout Standard
  3277. \begin_inset Flex TODO Note (inline)
  3278. status open
  3279. \begin_layout Plain Layout
  3280. Check up on figure refs in this paragraph
  3281. \end_layout
  3282. \end_inset
  3283. \end_layout
  3284. \begin_layout Standard
  3285. We hypothesized that if naive cells had differentiated into memory cells
  3286. by Day 14, then their patterns of expression and histone modification should
  3287. converge with those of memory cells at Day 14.
  3288. Figure
  3289. \begin_inset CommandInset ref
  3290. LatexCommand ref
  3291. reference "fig:PCoA-promoters"
  3292. plural "false"
  3293. caps "false"
  3294. noprefix "false"
  3295. \end_inset
  3296. shows the patterns of variation in all 3 histone marks in the promoter
  3297. regions of the genome using principal coordinate analysis.
  3298. All 3 marks show a noticeable convergence between the naive and memory
  3299. samples at day 14, visible as an overlapping of the day 14 groups on each
  3300. plot.
  3301. This is consistent with the counts of significantly differentially modified
  3302. promoters and estimates of the total numbers of differentially modified
  3303. promoters shown in Table
  3304. \begin_inset CommandInset ref
  3305. LatexCommand ref
  3306. reference "tab:Number-signif-promoters"
  3307. plural "false"
  3308. caps "false"
  3309. noprefix "false"
  3310. \end_inset
  3311. .
  3312. For all histone marks, evidence of differential modification between naive
  3313. and memory samples was detected at every time point except day 14.
  3314. The day 14 convergence pattern is also present in the RNA-seq data (Figure
  3315. \begin_inset CommandInset ref
  3316. LatexCommand ref
  3317. reference "fig:RNA-PCA-group"
  3318. plural "false"
  3319. caps "false"
  3320. noprefix "false"
  3321. \end_inset
  3322. ), albiet in the 2nd and 3rd principal coordinates, indicating that it is
  3323. not the most dominant pattern driving gene expression.
  3324. Taken together, the data show that promoter histone methylation for these
  3325. 3 histone marks and RNA expression for naive and memory cells are most
  3326. similar at day 14, the furthest time point after activation.
  3327. MOFA was also able to capture this day 14 convergence pattern in latent
  3328. factor 5 (Figure
  3329. \begin_inset CommandInset ref
  3330. LatexCommand ref
  3331. reference "fig:mofa-lf-scatter"
  3332. plural "false"
  3333. caps "false"
  3334. noprefix "false"
  3335. \end_inset
  3336. ), which accounts for shared variation across all 3 histone marks and the
  3337. RNA-seq data, confirming that this convergence is a coordinated pattern
  3338. across all 4 data sets.
  3339. While this observation does not prove that the naive cells have differentiated
  3340. into memory cells at Day 14, it is consistent with that hypothesis.
  3341. \end_layout
  3342. \begin_layout Subsection
  3343. Effect of H3K4me2 and H3K4me3 promoter coverage upstream vs downstream of
  3344. TSS
  3345. \end_layout
  3346. \begin_layout Standard
  3347. \begin_inset Flex TODO Note (inline)
  3348. status open
  3349. \begin_layout Plain Layout
  3350. Need a better section title, for this and the next one.
  3351. \end_layout
  3352. \end_inset
  3353. \end_layout
  3354. \begin_layout Standard
  3355. \begin_inset Flex TODO Note (inline)
  3356. status open
  3357. \begin_layout Plain Layout
  3358. Make sure use of coverage/abundance/whatever is consistent.
  3359. \end_layout
  3360. \end_inset
  3361. \end_layout
  3362. \begin_layout Standard
  3363. \begin_inset Flex TODO Note (inline)
  3364. status open
  3365. \begin_layout Plain Layout
  3366. For the figures in this section and the next, the group labels are arbitrary,
  3367. so if time allows, it would be good to manually reorder them in a logical
  3368. way, e.g.
  3369. most upstream to most downstream.
  3370. If this is done, make sure to update the text with the correct group labels.
  3371. \end_layout
  3372. \end_inset
  3373. \end_layout
  3374. \begin_layout Standard
  3375. \begin_inset ERT
  3376. status open
  3377. \begin_layout Plain Layout
  3378. \backslash
  3379. afterpage{
  3380. \end_layout
  3381. \begin_layout Plain Layout
  3382. \backslash
  3383. begin{landscape}
  3384. \end_layout
  3385. \end_inset
  3386. \end_layout
  3387. \begin_layout Standard
  3388. \begin_inset Float figure
  3389. wide false
  3390. sideways false
  3391. status open
  3392. \begin_layout Plain Layout
  3393. \align center
  3394. \begin_inset Float figure
  3395. wide false
  3396. sideways false
  3397. status open
  3398. \begin_layout Plain Layout
  3399. \align center
  3400. \begin_inset Graphics
  3401. filename graphics/CD4-csaw/ChIP-seq/H3K4me2-neighborhood-clusters-CROP.png
  3402. lyxscale 25
  3403. width 30col%
  3404. groupId covprof-subfig
  3405. \end_inset
  3406. \end_layout
  3407. \begin_layout Plain Layout
  3408. \begin_inset Caption Standard
  3409. \begin_layout Plain Layout
  3410. \series bold
  3411. \begin_inset CommandInset label
  3412. LatexCommand label
  3413. name "fig:H3K4me2-neighborhood-clusters"
  3414. \end_inset
  3415. Average relative coverage for each bin in each cluster
  3416. \end_layout
  3417. \end_inset
  3418. \end_layout
  3419. \end_inset
  3420. \begin_inset space \hfill{}
  3421. \end_inset
  3422. \begin_inset Float figure
  3423. wide false
  3424. sideways false
  3425. status open
  3426. \begin_layout Plain Layout
  3427. \align center
  3428. \begin_inset Graphics
  3429. filename graphics/CD4-csaw/ChIP-seq/H3K4me2-neighborhood-PCA-CROP.png
  3430. lyxscale 25
  3431. width 30col%
  3432. groupId covprof-subfig
  3433. \end_inset
  3434. \end_layout
  3435. \begin_layout Plain Layout
  3436. \begin_inset Caption Standard
  3437. \begin_layout Plain Layout
  3438. \series bold
  3439. \begin_inset CommandInset label
  3440. LatexCommand label
  3441. name "fig:H3K4me2-neighborhood-pca"
  3442. \end_inset
  3443. PCA of relative coverage depth, colored by K-means cluster membership.
  3444. \end_layout
  3445. \end_inset
  3446. \end_layout
  3447. \end_inset
  3448. \begin_inset space \hfill{}
  3449. \end_inset
  3450. \begin_inset Float figure
  3451. wide false
  3452. sideways false
  3453. status open
  3454. \begin_layout Plain Layout
  3455. \align center
  3456. \begin_inset Graphics
  3457. filename graphics/CD4-csaw/ChIP-seq/H3K4me2-neighborhood-expression-CROP.png
  3458. lyxscale 25
  3459. width 30col%
  3460. groupId covprof-subfig
  3461. \end_inset
  3462. \end_layout
  3463. \begin_layout Plain Layout
  3464. \begin_inset Caption Standard
  3465. \begin_layout Plain Layout
  3466. \series bold
  3467. \begin_inset CommandInset label
  3468. LatexCommand label
  3469. name "fig:H3K4me2-neighborhood-expression"
  3470. \end_inset
  3471. Gene expression grouped by promoter coverage clusters.
  3472. \end_layout
  3473. \end_inset
  3474. \end_layout
  3475. \end_inset
  3476. \end_layout
  3477. \begin_layout Plain Layout
  3478. \begin_inset Caption Standard
  3479. \begin_layout Plain Layout
  3480. \series bold
  3481. \begin_inset CommandInset label
  3482. LatexCommand label
  3483. name "fig:H3K4me2-neighborhood"
  3484. \end_inset
  3485. K-means clustering of promoter H3K4me2 relative coverage depth in naive
  3486. day 0 samples.
  3487. \series default
  3488. H3K4me2 ChIP-seq reads were binned into 500-bp windows tiled across each
  3489. promoter from 5
  3490. \begin_inset space ~
  3491. \end_inset
  3492. kbp upstream to 5
  3493. \begin_inset space ~
  3494. \end_inset
  3495. kbp downstream, and the logCPM values were normalized within each promoter
  3496. to an average of 0, yielding relative coverage depths.
  3497. These were then grouped using K-means clustering with
  3498. \begin_inset Formula $K=6$
  3499. \end_inset
  3500. ,
  3501. \series bold
  3502. \series default
  3503. and the average bin values were plotted for each cluster (a).
  3504. The
  3505. \begin_inset Formula $x$
  3506. \end_inset
  3507. -axis is the genomic coordinate of each bin relative to the the transcription
  3508. start site, and the
  3509. \begin_inset Formula $y$
  3510. \end_inset
  3511. -axis is the mean relative coverage depth of that bin across all promoters
  3512. in the cluster.
  3513. Each line represents the average
  3514. \begin_inset Quotes eld
  3515. \end_inset
  3516. shape
  3517. \begin_inset Quotes erd
  3518. \end_inset
  3519. of the promoter coverage for promoters in that cluster.
  3520. PCA was performed on the same data, and the first two principal components
  3521. were plotted, coloring each point by its K-means cluster identity (b).
  3522. For each cluster, the distribution of gene expression values was plotted
  3523. (c).
  3524. \end_layout
  3525. \end_inset
  3526. \end_layout
  3527. \end_inset
  3528. \end_layout
  3529. \begin_layout Standard
  3530. \begin_inset ERT
  3531. status open
  3532. \begin_layout Plain Layout
  3533. \backslash
  3534. end{landscape}
  3535. \end_layout
  3536. \begin_layout Plain Layout
  3537. }
  3538. \end_layout
  3539. \end_inset
  3540. \end_layout
  3541. \begin_layout Standard
  3542. To test whether the position of a histone mark relative to a gene's transcriptio
  3543. n start site (TSS) was important, we looked at the
  3544. \begin_inset Quotes eld
  3545. \end_inset
  3546. landscape
  3547. \begin_inset Quotes erd
  3548. \end_inset
  3549. of ChIP-seq read coverage in naive Day 0 samples within 5 kb of each gene's
  3550. TSS by binning reads into 500-bp windows tiled across each promoter LogCPM
  3551. values were calculated for the bins in each promoter and then the average
  3552. logCPM for each promoter's bins was normalized to zero, such that the values
  3553. represent coverage relative to other regions of the same promoter rather
  3554. than being proportional to absolute read count.
  3555. The promoters were then clustered based on the normalized bin abundances
  3556. using
  3557. \begin_inset Formula $k$
  3558. \end_inset
  3559. -means clustering with
  3560. \begin_inset Formula $K=6$
  3561. \end_inset
  3562. .
  3563. Different values of
  3564. \begin_inset Formula $K$
  3565. \end_inset
  3566. were also tested, but did not substantially change the interpretation of
  3567. the data.
  3568. \end_layout
  3569. \begin_layout Standard
  3570. For H3K4me2, plotting the average bin abundances for each cluster reveals
  3571. a simple pattern (Figure
  3572. \begin_inset CommandInset ref
  3573. LatexCommand ref
  3574. reference "fig:H3K4me2-neighborhood-clusters"
  3575. plural "false"
  3576. caps "false"
  3577. noprefix "false"
  3578. \end_inset
  3579. ): Cluster 5 represents a completely flat promoter coverage profile, likely
  3580. consisting of genes with no H3K4me2 methylation in the promoter.
  3581. All the other clusters represent a continuum of peak positions relative
  3582. to the TSS.
  3583. In order from must upstream to most downstream, they are Clusters 6, 4,
  3584. 3, 1, and 2.
  3585. There do not appear to be any clusters representing coverage patterns other
  3586. than lone peaks, such as coverage troughs or double peaks.
  3587. Next, all promoters were plotted in a PCA plot based on the same relative
  3588. bin abundance data, and colored based on cluster membership (Figure
  3589. \begin_inset CommandInset ref
  3590. LatexCommand ref
  3591. reference "fig:H3K4me2-neighborhood-pca"
  3592. plural "false"
  3593. caps "false"
  3594. noprefix "false"
  3595. \end_inset
  3596. ).
  3597. The PCA plot shows Cluster 5 (the
  3598. \begin_inset Quotes eld
  3599. \end_inset
  3600. no peak
  3601. \begin_inset Quotes erd
  3602. \end_inset
  3603. cluster) at the center, with the other clusters arranged in a counter-clockwise
  3604. arc around it in the order noted above, from most upstream peak to most
  3605. downstream.
  3606. Notably, the
  3607. \begin_inset Quotes eld
  3608. \end_inset
  3609. clusters
  3610. \begin_inset Quotes erd
  3611. \end_inset
  3612. form a single large
  3613. \begin_inset Quotes eld
  3614. \end_inset
  3615. cloud
  3616. \begin_inset Quotes erd
  3617. \end_inset
  3618. with no apparent separation between them, further supporting the conclusion
  3619. that these clusters represent an arbitrary partitioning of a continuous
  3620. distribution of promoter coverage landscapes.
  3621. While the clusters are a useful abstraction that aids in visualization,
  3622. they are ultimately not an accurate representation of the data.
  3623. A better representation might be something like a polar coordinate system
  3624. with the origin at the center of Cluster 5, where the radius represents
  3625. the peak height above the background and the angle represents the peak's
  3626. position upstream or downstream of the TSS.
  3627. The continuous nature of the distribution also explains why different values
  3628. of
  3629. \begin_inset Formula $K$
  3630. \end_inset
  3631. led to similar conclusions.
  3632. \end_layout
  3633. \begin_layout Standard
  3634. \begin_inset Flex TODO Note (inline)
  3635. status open
  3636. \begin_layout Plain Layout
  3637. RNA-seq values in the plots use logCPM but should really use logFPKM or
  3638. logTPM.
  3639. Fix if time allows.
  3640. \end_layout
  3641. \end_inset
  3642. \end_layout
  3643. \begin_layout Standard
  3644. \begin_inset Flex TODO Note (inline)
  3645. status open
  3646. \begin_layout Plain Layout
  3647. Should have a table of p-values on difference of means between Cluster 5
  3648. and the others.
  3649. \end_layout
  3650. \end_inset
  3651. \end_layout
  3652. \begin_layout Standard
  3653. To investigate the association between relative peak position and gene expressio
  3654. n, we plotted the Naive Day 0 expression for the genes in each cluster (Figure
  3655. \begin_inset CommandInset ref
  3656. LatexCommand ref
  3657. reference "fig:H3K4me2-neighborhood-expression"
  3658. plural "false"
  3659. caps "false"
  3660. noprefix "false"
  3661. \end_inset
  3662. ).
  3663. Most genes in Cluster 5, the
  3664. \begin_inset Quotes eld
  3665. \end_inset
  3666. no peak
  3667. \begin_inset Quotes erd
  3668. \end_inset
  3669. cluster, have low expression values.
  3670. Taking this as the
  3671. \begin_inset Quotes eld
  3672. \end_inset
  3673. baseline
  3674. \begin_inset Quotes erd
  3675. \end_inset
  3676. distribution when no H3K4me2 methylation is present, we can compare the
  3677. other clusters' distributions to determine which peak positions are associated
  3678. with elevated expression.
  3679. As might be expected, the 3 clusters representing peaks closest to the
  3680. TSS, Clusters 1, 3, and 4, show the highest average expression distributions.
  3681. Specifically, these clusters all have their highest ChIP-seq abundance
  3682. within 1kb of the TSS, consistent with the previously determined promoter
  3683. radius.
  3684. In contrast, cluster 6, which represents peaks several kb upstream of the
  3685. TSS, shows a slightly higher average expression than baseline, while Cluster
  3686. 2, which represents peaks several kb downstream, doesn't appear to show
  3687. any appreciable difference.
  3688. Interestingly, the cluster with the highest average expression is Cluster
  3689. 1, which represents peaks about 1 kb downstream of the TSS, rather than
  3690. Cluster 3, which represents peaks centered directly at the TSS.
  3691. This suggests that conceptualizing the promoter as a region centered on
  3692. the TSS with a certain
  3693. \begin_inset Quotes eld
  3694. \end_inset
  3695. radius
  3696. \begin_inset Quotes erd
  3697. \end_inset
  3698. may be an oversimplification – a peak that is a specific distance from
  3699. the TSS may have a different degree of influence depending on whether it
  3700. is upstream or downstream of the TSS.
  3701. \end_layout
  3702. \begin_layout Standard
  3703. \begin_inset ERT
  3704. status open
  3705. \begin_layout Plain Layout
  3706. \backslash
  3707. afterpage{
  3708. \end_layout
  3709. \begin_layout Plain Layout
  3710. \backslash
  3711. begin{landscape}
  3712. \end_layout
  3713. \end_inset
  3714. \end_layout
  3715. \begin_layout Standard
  3716. \begin_inset Float figure
  3717. wide false
  3718. sideways false
  3719. status open
  3720. \begin_layout Plain Layout
  3721. \align center
  3722. \begin_inset Float figure
  3723. wide false
  3724. sideways false
  3725. status open
  3726. \begin_layout Plain Layout
  3727. \align center
  3728. \begin_inset Graphics
  3729. filename graphics/CD4-csaw/ChIP-seq/H3K4me3-neighborhood-clusters-CROP.png
  3730. lyxscale 25
  3731. width 30col%
  3732. groupId covprof-subfig
  3733. \end_inset
  3734. \end_layout
  3735. \begin_layout Plain Layout
  3736. \begin_inset Caption Standard
  3737. \begin_layout Plain Layout
  3738. \series bold
  3739. \begin_inset CommandInset label
  3740. LatexCommand label
  3741. name "fig:H3K4me3-neighborhood-clusters"
  3742. \end_inset
  3743. Average relative coverage for each bin in each cluster
  3744. \end_layout
  3745. \end_inset
  3746. \end_layout
  3747. \end_inset
  3748. \begin_inset space \hfill{}
  3749. \end_inset
  3750. \begin_inset Float figure
  3751. wide false
  3752. sideways false
  3753. status open
  3754. \begin_layout Plain Layout
  3755. \align center
  3756. \begin_inset Graphics
  3757. filename graphics/CD4-csaw/ChIP-seq/H3K4me3-neighborhood-PCA-CROP.png
  3758. lyxscale 25
  3759. width 30col%
  3760. groupId covprof-subfig
  3761. \end_inset
  3762. \end_layout
  3763. \begin_layout Plain Layout
  3764. \begin_inset Caption Standard
  3765. \begin_layout Plain Layout
  3766. \series bold
  3767. \begin_inset CommandInset label
  3768. LatexCommand label
  3769. name "fig:H3K4me3-neighborhood-pca"
  3770. \end_inset
  3771. PCA of relative coverage depth, colored by K-means cluster membership.
  3772. \end_layout
  3773. \end_inset
  3774. \end_layout
  3775. \end_inset
  3776. \begin_inset space \hfill{}
  3777. \end_inset
  3778. \begin_inset Float figure
  3779. wide false
  3780. sideways false
  3781. status open
  3782. \begin_layout Plain Layout
  3783. \align center
  3784. \begin_inset Graphics
  3785. filename graphics/CD4-csaw/ChIP-seq/H3K4me3-neighborhood-expression-CROP.png
  3786. lyxscale 25
  3787. width 30col%
  3788. groupId covprof-subfig
  3789. \end_inset
  3790. \end_layout
  3791. \begin_layout Plain Layout
  3792. \begin_inset Caption Standard
  3793. \begin_layout Plain Layout
  3794. \series bold
  3795. \begin_inset CommandInset label
  3796. LatexCommand label
  3797. name "fig:H3K4me3-neighborhood-expression"
  3798. \end_inset
  3799. Gene expression grouped by promoter coverage clusters.
  3800. \end_layout
  3801. \end_inset
  3802. \end_layout
  3803. \end_inset
  3804. \end_layout
  3805. \begin_layout Plain Layout
  3806. \begin_inset Caption Standard
  3807. \begin_layout Plain Layout
  3808. \series bold
  3809. \begin_inset CommandInset label
  3810. LatexCommand label
  3811. name "fig:H3K4me3-neighborhood"
  3812. \end_inset
  3813. K-means clustering of promoter H3K4me3 relative coverage depth in naive
  3814. day 0 samples.
  3815. \series default
  3816. H3K4me2 ChIP-seq reads were binned into 500-bp windows tiled across each
  3817. promoter from 5
  3818. \begin_inset space ~
  3819. \end_inset
  3820. kbp upstream to 5
  3821. \begin_inset space ~
  3822. \end_inset
  3823. kbp downstream, and the logCPM values were normalized within each promoter
  3824. to an average of 0, yielding relative coverage depths.
  3825. These were then grouped using K-means clustering with
  3826. \begin_inset Formula $K=6$
  3827. \end_inset
  3828. ,
  3829. \series bold
  3830. \series default
  3831. and the average bin values were plotted for each cluster (a).
  3832. The
  3833. \begin_inset Formula $x$
  3834. \end_inset
  3835. -axis is the genomic coordinate of each bin relative to the the transcription
  3836. start site, and the
  3837. \begin_inset Formula $y$
  3838. \end_inset
  3839. -axis is the mean relative coverage depth of that bin across all promoters
  3840. in the cluster.
  3841. Each line represents the average
  3842. \begin_inset Quotes eld
  3843. \end_inset
  3844. shape
  3845. \begin_inset Quotes erd
  3846. \end_inset
  3847. of the promoter coverage for promoters in that cluster.
  3848. PCA was performed on the same data, and the first two principal components
  3849. were plotted, coloring each point by its K-means cluster identity (b).
  3850. For each cluster, the distribution of gene expression values was plotted
  3851. (c).
  3852. \end_layout
  3853. \end_inset
  3854. \end_layout
  3855. \end_inset
  3856. \end_layout
  3857. \begin_layout Standard
  3858. \begin_inset ERT
  3859. status open
  3860. \begin_layout Plain Layout
  3861. \backslash
  3862. end{landscape}
  3863. \end_layout
  3864. \begin_layout Plain Layout
  3865. }
  3866. \end_layout
  3867. \end_inset
  3868. \end_layout
  3869. \begin_layout Standard
  3870. \begin_inset Flex TODO Note (inline)
  3871. status open
  3872. \begin_layout Plain Layout
  3873. Is there more to say here?
  3874. \end_layout
  3875. \end_inset
  3876. \end_layout
  3877. \begin_layout Standard
  3878. All observations described above for H3K4me2 ChIP-seq also appear to hold
  3879. for H3K4me3 as well (Figure
  3880. \begin_inset CommandInset ref
  3881. LatexCommand ref
  3882. reference "fig:H3K4me3-neighborhood"
  3883. plural "false"
  3884. caps "false"
  3885. noprefix "false"
  3886. \end_inset
  3887. ).
  3888. This is expected, since there is a high correlation between the positions
  3889. where both histone marks occur.
  3890. \end_layout
  3891. \begin_layout Subsection
  3892. Promoter coverage H3K27me3
  3893. \end_layout
  3894. \begin_layout Standard
  3895. \begin_inset ERT
  3896. status open
  3897. \begin_layout Plain Layout
  3898. \backslash
  3899. afterpage{
  3900. \end_layout
  3901. \begin_layout Plain Layout
  3902. \backslash
  3903. begin{landscape}
  3904. \end_layout
  3905. \end_inset
  3906. \end_layout
  3907. \begin_layout Standard
  3908. \begin_inset Float figure
  3909. wide false
  3910. sideways false
  3911. status collapsed
  3912. \begin_layout Plain Layout
  3913. \align center
  3914. \begin_inset Float figure
  3915. wide false
  3916. sideways false
  3917. status open
  3918. \begin_layout Plain Layout
  3919. \align center
  3920. \begin_inset Graphics
  3921. filename graphics/CD4-csaw/ChIP-seq/H3K27me3-neighborhood-clusters-CROP.png
  3922. lyxscale 25
  3923. width 30col%
  3924. groupId covprof-subfig
  3925. \end_inset
  3926. \end_layout
  3927. \begin_layout Plain Layout
  3928. \begin_inset Caption Standard
  3929. \begin_layout Plain Layout
  3930. \series bold
  3931. \begin_inset CommandInset label
  3932. LatexCommand label
  3933. name "fig:H3K27me3-neighborhood-clusters"
  3934. \end_inset
  3935. Average relative coverage for each bin in each cluster
  3936. \end_layout
  3937. \end_inset
  3938. \end_layout
  3939. \end_inset
  3940. \begin_inset space \hfill{}
  3941. \end_inset
  3942. \begin_inset Float figure
  3943. wide false
  3944. sideways false
  3945. status open
  3946. \begin_layout Plain Layout
  3947. \align center
  3948. \begin_inset Graphics
  3949. filename graphics/CD4-csaw/ChIP-seq/H3K27me3-neighborhood-PCA-CROP.png
  3950. lyxscale 25
  3951. width 30col%
  3952. groupId covprof-subfig
  3953. \end_inset
  3954. \end_layout
  3955. \begin_layout Plain Layout
  3956. \begin_inset Caption Standard
  3957. \begin_layout Plain Layout
  3958. \series bold
  3959. \begin_inset CommandInset label
  3960. LatexCommand label
  3961. name "fig:H3K27me3-neighborhood-pca"
  3962. \end_inset
  3963. PCA of relative coverage depth, colored by K-means cluster membership.
  3964. \series default
  3965. Note that Cluster 6 is hidden behind all the other clusters.
  3966. \end_layout
  3967. \end_inset
  3968. \end_layout
  3969. \end_inset
  3970. \begin_inset space \hfill{}
  3971. \end_inset
  3972. \begin_inset Float figure
  3973. wide false
  3974. sideways false
  3975. status open
  3976. \begin_layout Plain Layout
  3977. \align center
  3978. \begin_inset Graphics
  3979. filename graphics/CD4-csaw/ChIP-seq/H3K27me3-neighborhood-expression-CROP.png
  3980. lyxscale 25
  3981. width 30col%
  3982. groupId covprof-subfig
  3983. \end_inset
  3984. \end_layout
  3985. \begin_layout Plain Layout
  3986. \begin_inset Caption Standard
  3987. \begin_layout Plain Layout
  3988. \series bold
  3989. \begin_inset CommandInset label
  3990. LatexCommand label
  3991. name "fig:H3K27me3-neighborhood-expression"
  3992. \end_inset
  3993. Gene expression grouped by promoter coverage clusters.
  3994. \end_layout
  3995. \end_inset
  3996. \end_layout
  3997. \end_inset
  3998. \end_layout
  3999. \begin_layout Plain Layout
  4000. \begin_inset Flex TODO Note (inline)
  4001. status open
  4002. \begin_layout Plain Layout
  4003. Repeated figure legends are kind of an issue here.
  4004. What to do?
  4005. \end_layout
  4006. \end_inset
  4007. \end_layout
  4008. \begin_layout Plain Layout
  4009. \begin_inset Caption Standard
  4010. \begin_layout Plain Layout
  4011. \series bold
  4012. \begin_inset CommandInset label
  4013. LatexCommand label
  4014. name "fig:H3K27me3-neighborhood"
  4015. \end_inset
  4016. K-means clustering of promoter H3K27me3 relative coverage depth in naive
  4017. day 0 samples.
  4018. \series default
  4019. H3K27me3 ChIP-seq reads were binned into 500-bp windows tiled across each
  4020. promoter from 5
  4021. \begin_inset space ~
  4022. \end_inset
  4023. kbp upstream to 5
  4024. \begin_inset space ~
  4025. \end_inset
  4026. kbp downstream, and the logCPM values were normalized within each promoter
  4027. to an average of 0, yielding relative coverage depths.
  4028. These were then grouped using
  4029. \begin_inset Formula $k$
  4030. \end_inset
  4031. -means clustering with
  4032. \begin_inset Formula $K=6$
  4033. \end_inset
  4034. ,
  4035. \series bold
  4036. \series default
  4037. and the average bin values were plotted for each cluster (a).
  4038. The
  4039. \begin_inset Formula $x$
  4040. \end_inset
  4041. -axis is the genomic coordinate of each bin relative to the the transcription
  4042. start site, and the
  4043. \begin_inset Formula $y$
  4044. \end_inset
  4045. -axis is the mean relative coverage depth of that bin across all promoters
  4046. in the cluster.
  4047. Each line represents the average
  4048. \begin_inset Quotes eld
  4049. \end_inset
  4050. shape
  4051. \begin_inset Quotes erd
  4052. \end_inset
  4053. of the promoter coverage for promoters in that cluster.
  4054. PCA was performed on the same data, and the first two principal components
  4055. were plotted, coloring each point by its K-means cluster identity (b).
  4056. For each cluster, the distribution of gene expression values was plotted
  4057. (c).
  4058. \end_layout
  4059. \end_inset
  4060. \end_layout
  4061. \end_inset
  4062. \end_layout
  4063. \begin_layout Standard
  4064. \begin_inset ERT
  4065. status open
  4066. \begin_layout Plain Layout
  4067. \backslash
  4068. end{landscape}
  4069. \end_layout
  4070. \begin_layout Plain Layout
  4071. }
  4072. \end_layout
  4073. \end_inset
  4074. \end_layout
  4075. \begin_layout Standard
  4076. \begin_inset Flex TODO Note (inline)
  4077. status open
  4078. \begin_layout Plain Layout
  4079. Should maybe re-explain what was done or refer back to the previous section.
  4080. \end_layout
  4081. \end_inset
  4082. \end_layout
  4083. \begin_layout Standard
  4084. Unlike both H3K4 marks, whose main patterns of variation appear directly
  4085. related to the size and position of a single peak within the promoter,
  4086. the patterns of H3K27me3 methylation in promoters are more complex (Figure
  4087. \begin_inset CommandInset ref
  4088. LatexCommand ref
  4089. reference "fig:H3K27me3-neighborhood"
  4090. plural "false"
  4091. caps "false"
  4092. noprefix "false"
  4093. \end_inset
  4094. ).
  4095. Once again looking at the relative coverage in a 500-bp wide bins in a
  4096. 5kb radius around each TSS, promoters were clustered based on the normalized
  4097. relative coverage values in each bin using
  4098. \begin_inset Formula $k$
  4099. \end_inset
  4100. -means clustering with
  4101. \begin_inset Formula $K=6$
  4102. \end_inset
  4103. (Figure
  4104. \begin_inset CommandInset ref
  4105. LatexCommand ref
  4106. reference "fig:H3K27me3-neighborhood-clusters"
  4107. plural "false"
  4108. caps "false"
  4109. noprefix "false"
  4110. \end_inset
  4111. ).
  4112. This time, 3
  4113. \begin_inset Quotes eld
  4114. \end_inset
  4115. axes
  4116. \begin_inset Quotes erd
  4117. \end_inset
  4118. of variation can be observed, each represented by 2 clusters with opposing
  4119. patterns.
  4120. The first axis is greater upstream coverage (Cluster 1) vs.
  4121. greater downstream coverage (Cluster 3); the second axis is the coverage
  4122. at the TSS itself: peak (Cluster 4) or trough (Cluster 2); lastly, the
  4123. third axis represents a trough upstream of the TSS (Cluster 5) vs.
  4124. downstream of the TSS (Cluster 6).
  4125. Referring to these opposing pairs of clusters as axes of variation is justified
  4126. , because they correspond precisely to the first 3 principal components
  4127. in the PCA plot of the relative coverage values (Figure
  4128. \begin_inset CommandInset ref
  4129. LatexCommand ref
  4130. reference "fig:H3K27me3-neighborhood-pca"
  4131. plural "false"
  4132. caps "false"
  4133. noprefix "false"
  4134. \end_inset
  4135. ).
  4136. The PCA plot reveals that as in the case of H3K4me2, all the
  4137. \begin_inset Quotes eld
  4138. \end_inset
  4139. clusters
  4140. \begin_inset Quotes erd
  4141. \end_inset
  4142. are really just sections of a single connected cloud rather than discrete
  4143. clusters.
  4144. The cloud is approximately ellipsoid-shaped, with each PC being an axis
  4145. of the ellipse, and each cluster consisting of a pyrimidal section of the
  4146. ellipsoid.
  4147. \end_layout
  4148. \begin_layout Standard
  4149. In Figure
  4150. \begin_inset CommandInset ref
  4151. LatexCommand ref
  4152. reference "fig:H3K27me3-neighborhood-expression"
  4153. plural "false"
  4154. caps "false"
  4155. noprefix "false"
  4156. \end_inset
  4157. , we can see that Clusters 1 and 2 are the only clusters with higher gene
  4158. expression than the others.
  4159. For Cluster 2, this is expected, since this cluster represents genes with
  4160. depletion of H3K27me3 near the promoter.
  4161. Hence, elevated expression in cluster 2 is consistent with the conventional
  4162. view of H3K27me3 as a deactivating mark.
  4163. However, Cluster 1, the cluster with the most elevated gene expression,
  4164. represents genes with elevated coverage upstream of the TSS, or equivalently,
  4165. decreased coverage downstream, inside the gene body.
  4166. The opposite pattern, in which H3K27me3 is more abundant within the gene
  4167. body and less abundance in the upstream promoter region, does not show
  4168. any elevation in gene expression.
  4169. As with H3K4me2, this shows that the location of H3K27 trimethylation relative
  4170. to the TSS is potentially an important factor beyond simple proximity.
  4171. \end_layout
  4172. \begin_layout Standard
  4173. \begin_inset Flex TODO Note (inline)
  4174. status open
  4175. \begin_layout Plain Layout
  4176. Show the figures where the negative result ended this line of inquiry.
  4177. I need to debug some errors resulting from an R upgrade to do this.
  4178. \end_layout
  4179. \end_inset
  4180. \end_layout
  4181. \begin_layout Subsection
  4182. Defined pattern analysis
  4183. \end_layout
  4184. \begin_layout Standard
  4185. \begin_inset Flex TODO Note (inline)
  4186. status open
  4187. \begin_layout Plain Layout
  4188. This was where I defined interesting expression patterns and then looked
  4189. at initial relative promoter coverage for each expression pattern.
  4190. Negative result.
  4191. I forgot about this until recently.
  4192. Worth including? Remember to also write methods.
  4193. \end_layout
  4194. \end_inset
  4195. \end_layout
  4196. \begin_layout Subsection
  4197. Promoter CpG islands?
  4198. \end_layout
  4199. \begin_layout Standard
  4200. \begin_inset Flex TODO Note (inline)
  4201. status open
  4202. \begin_layout Plain Layout
  4203. I forgot until recently about the work I did on this.
  4204. Worth including? Remember to also write methods.
  4205. \end_layout
  4206. \end_inset
  4207. \end_layout
  4208. \begin_layout Section
  4209. Discussion
  4210. \end_layout
  4211. \begin_layout Standard
  4212. \begin_inset Flex TODO Note (inline)
  4213. status open
  4214. \begin_layout Plain Layout
  4215. Write better section headers
  4216. \end_layout
  4217. \end_inset
  4218. \end_layout
  4219. \begin_layout Subsection
  4220. Effective promoter radius
  4221. \end_layout
  4222. \begin_layout Standard
  4223. Figure
  4224. \begin_inset CommandInset ref
  4225. LatexCommand ref
  4226. reference "fig:near-promoter-peak-enrich"
  4227. plural "false"
  4228. caps "false"
  4229. noprefix "false"
  4230. \end_inset
  4231. shows that H3K4me2, H3K4me3, and H3K27me3 are all enriched near promoters,
  4232. relative to the rest of the genome, consistent with their conventionally
  4233. understood role in regulating gene transcription.
  4234. Interestingly, the radius within this enrichment occurs is not the same
  4235. for each histone mark.
  4236. H3K4me2 and H3K4me3 are enriched within a 1
  4237. \begin_inset space \thinspace{}
  4238. \end_inset
  4239. kb radius, while H3K27me3 is enriched within 2.5
  4240. \begin_inset space \thinspace{}
  4241. \end_inset
  4242. kb.
  4243. Notably, the determined promoter radius was consistent across all experimental
  4244. conditions, varying only between different histone marks.
  4245. This suggests that the conventional
  4246. \begin_inset Quotes eld
  4247. \end_inset
  4248. one size fits all
  4249. \begin_inset Quotes erd
  4250. \end_inset
  4251. approach of defining a single promoter region for each gene (or each TSS)
  4252. and using that same promoter region for analyzing all types of genomic
  4253. data within an experiment may not be appropriate, and a better approach
  4254. may be to use a separate promoter radius for each kind of data, with each
  4255. radius being derived from the data itself.
  4256. Furthermore, the apparent assymetry of upstream and downstream promoter
  4257. histone modification with respect to gene expression, seen in Figures
  4258. \begin_inset CommandInset ref
  4259. LatexCommand ref
  4260. reference "fig:H3K4me2-neighborhood"
  4261. plural "false"
  4262. caps "false"
  4263. noprefix "false"
  4264. \end_inset
  4265. ,
  4266. \begin_inset CommandInset ref
  4267. LatexCommand ref
  4268. reference "fig:H3K4me3-neighborhood"
  4269. plural "false"
  4270. caps "false"
  4271. noprefix "false"
  4272. \end_inset
  4273. , and
  4274. \begin_inset CommandInset ref
  4275. LatexCommand ref
  4276. reference "fig:H3K27me3-neighborhood"
  4277. plural "false"
  4278. caps "false"
  4279. noprefix "false"
  4280. \end_inset
  4281. , shows that even the concept of a promoter
  4282. \begin_inset Quotes eld
  4283. \end_inset
  4284. radius
  4285. \begin_inset Quotes erd
  4286. \end_inset
  4287. is likely an oversimplification.
  4288. At a minimum, nearby enrichment of peaks should be evaluated separately
  4289. for both upstream and downstream peaks, and an appropriate
  4290. \begin_inset Quotes eld
  4291. \end_inset
  4292. radius
  4293. \begin_inset Quotes erd
  4294. \end_inset
  4295. should be selected for each direction.
  4296. \end_layout
  4297. \begin_layout Standard
  4298. Figures
  4299. \begin_inset CommandInset ref
  4300. LatexCommand ref
  4301. reference "fig:H3K4me2-neighborhood"
  4302. plural "false"
  4303. caps "false"
  4304. noprefix "false"
  4305. \end_inset
  4306. and
  4307. \begin_inset CommandInset ref
  4308. LatexCommand ref
  4309. reference "fig:H3K4me3-neighborhood"
  4310. plural "false"
  4311. caps "false"
  4312. noprefix "false"
  4313. \end_inset
  4314. show that the determined promoter radius of 1
  4315. \begin_inset space ~
  4316. \end_inset
  4317. kb is approximately consistent with the distance from the TSS at which enrichmen
  4318. t of H3K4 methylationis correlates with increased expression, showing that
  4319. this radius, which was determined by a simple analysis of measuring the
  4320. distance from each TSS to the nearest peak, also has functional significance.
  4321. For H3K27me3, the correlation between histone modification near the promoter
  4322. and gene expression is more complex, involving non-peak variations such
  4323. as troughs in coverage at the TSS and asymmetric coverage upstream and
  4324. downstream, so it is difficult in this case to evaluate whether the 2.5
  4325. \begin_inset space ~
  4326. \end_inset
  4327. kb radius determined from TSS-to-peak distances is functionally significant.
  4328. However, the two patterns of coverage associated with elevated expression
  4329. levels both have interesting features within this radius.
  4330. \end_layout
  4331. \begin_layout Standard
  4332. \begin_inset Flex TODO Note (inline)
  4333. status open
  4334. \begin_layout Plain Layout
  4335. My instinct is to say
  4336. \begin_inset Quotes eld
  4337. \end_inset
  4338. further study is needed
  4339. \begin_inset Quotes erd
  4340. \end_inset
  4341. here, but that goes in Chapter 5, right?
  4342. \end_layout
  4343. \end_inset
  4344. \end_layout
  4345. \begin_layout Subsection
  4346. Convergence
  4347. \end_layout
  4348. \begin_layout Standard
  4349. \begin_inset Flex TODO Note (inline)
  4350. status open
  4351. \begin_layout Plain Layout
  4352. Look up some more references for these histone marks being involved in memory
  4353. differentiation.
  4354. (Ask Sarah)
  4355. \end_layout
  4356. \end_inset
  4357. \end_layout
  4358. \begin_layout Standard
  4359. We have observed that all 3 histone marks and the gene expression data all
  4360. exhibit evidence of convergence in abundance between naive and memory cells
  4361. by day 14 after activation (Figure
  4362. \begin_inset CommandInset ref
  4363. LatexCommand ref
  4364. reference "fig:PCoA-promoters"
  4365. plural "false"
  4366. caps "false"
  4367. noprefix "false"
  4368. \end_inset
  4369. , Table
  4370. \begin_inset CommandInset ref
  4371. LatexCommand ref
  4372. reference "tab:Number-signif-promoters"
  4373. plural "false"
  4374. caps "false"
  4375. noprefix "false"
  4376. \end_inset
  4377. ).
  4378. The MOFA latent factor scatter plots (Figure
  4379. \begin_inset CommandInset ref
  4380. LatexCommand ref
  4381. reference "fig:mofa-lf-scatter"
  4382. plural "false"
  4383. caps "false"
  4384. noprefix "false"
  4385. \end_inset
  4386. ) show that this pattern of convergence is captured in latent factor 5.
  4387. Like all the latent factors in this plot, this factor explains a substantial
  4388. portion of the variance in all 4 data sets, indicating a coordinated pattern
  4389. of variation shared across all histone marks and gene expression.
  4390. This, of course, is consistent with the expectation that any naive CD4
  4391. T-cells remaining at day 14 should have differentiated into memory cells
  4392. by that time, and should therefore have a genomic state similar to memory
  4393. cells.
  4394. This convergence is evidence that these histone marks all play an important
  4395. role in the naive-to-memory differentiation process.
  4396. A histone mark that was not involved in naive-to-memory differentiation
  4397. would not be expected to converge in this way after activation.
  4398. \end_layout
  4399. \begin_layout Standard
  4400. \begin_inset Float figure
  4401. wide false
  4402. sideways false
  4403. status collapsed
  4404. \begin_layout Plain Layout
  4405. \align center
  4406. \begin_inset Graphics
  4407. filename graphics/CD4-csaw/LaMere2016_fig8.pdf
  4408. lyxscale 50
  4409. width 60col%
  4410. groupId colwidth
  4411. \end_inset
  4412. \end_layout
  4413. \begin_layout Plain Layout
  4414. \begin_inset Caption Standard
  4415. \begin_layout Plain Layout
  4416. \series bold
  4417. \begin_inset CommandInset label
  4418. LatexCommand label
  4419. name "fig:Lamere2016-Fig8"
  4420. \end_inset
  4421. Lamere 2016 Figure 8
  4422. \begin_inset CommandInset citation
  4423. LatexCommand cite
  4424. key "LaMere2016"
  4425. literal "false"
  4426. \end_inset
  4427. ,
  4428. \begin_inset Quotes eld
  4429. \end_inset
  4430. Model for the role of H3K4 methylation during CD4 T-cell activation.
  4431. \begin_inset Quotes erd
  4432. \end_inset
  4433. \series default
  4434. Reproduced with permission.
  4435. \end_layout
  4436. \end_inset
  4437. \end_layout
  4438. \end_inset
  4439. \end_layout
  4440. \begin_layout Standard
  4441. In H3K4me2, H3K4me3, and RNA-seq, this convergence appears to be in progress
  4442. already by Day 5, shown by the smaller distance between naive and memory
  4443. cells at day 5 along the
  4444. \begin_inset Formula $y$
  4445. \end_inset
  4446. -axes in Figures
  4447. \begin_inset CommandInset ref
  4448. LatexCommand ref
  4449. reference "fig:PCoA-H3K4me2-prom"
  4450. plural "false"
  4451. caps "false"
  4452. noprefix "false"
  4453. \end_inset
  4454. ,
  4455. \begin_inset CommandInset ref
  4456. LatexCommand ref
  4457. reference "fig:PCoA-H3K4me3-prom"
  4458. plural "false"
  4459. caps "false"
  4460. noprefix "false"
  4461. \end_inset
  4462. , and
  4463. \begin_inset CommandInset ref
  4464. LatexCommand ref
  4465. reference "fig:RNA-PCA-group"
  4466. plural "false"
  4467. caps "false"
  4468. noprefix "false"
  4469. \end_inset
  4470. .
  4471. This agrees with the model proposed by Sarah Lamere based on an prior analysis
  4472. of the same data, shown in Figure
  4473. \begin_inset CommandInset ref
  4474. LatexCommand ref
  4475. reference "fig:Lamere2016-Fig8"
  4476. plural "false"
  4477. caps "false"
  4478. noprefix "false"
  4479. \end_inset
  4480. , which shows the pattern of H3K4 methylation and expression for naive cells
  4481. and memory cells converging at day 5.
  4482. This model was developed without the benefit of the PCoA plots in Figure
  4483. \begin_inset CommandInset ref
  4484. LatexCommand ref
  4485. reference "fig:PCoA-promoters"
  4486. plural "false"
  4487. caps "false"
  4488. noprefix "false"
  4489. \end_inset
  4490. , which have been corrected for confounding factors by ComBat and SVA.
  4491. This shows that proper batch correction assists in extracting meaningful
  4492. patterns in the data while eliminating systematic sources of irrelevant
  4493. variation in the data, allowing simple automated procedures like PCoA to
  4494. reveal interesting behaviors in the data that were previously only detectable
  4495. by a detailed manual analysis.
  4496. \end_layout
  4497. \begin_layout Standard
  4498. While the ideal comparison to demonstrate this convergence would be naive
  4499. cells at day 14 to memory cells at day 0, this is not feasible in this
  4500. experimental system, since neither naive nor memory cells are able to fully
  4501. return to their pre-activation state, as shown by the lack of overlap between
  4502. days 0 and 14 for either naive or memory cells in Figure
  4503. \begin_inset CommandInset ref
  4504. LatexCommand ref
  4505. reference "fig:PCoA-promoters"
  4506. plural "false"
  4507. caps "false"
  4508. noprefix "false"
  4509. \end_inset
  4510. .
  4511. \end_layout
  4512. \begin_layout Subsection
  4513. Positional
  4514. \end_layout
  4515. \begin_layout Standard
  4516. When looking at patterns in the relative coverage of each histone mark near
  4517. the TSS of each gene, several interesting patterns were apparent.
  4518. For H3K4me2 and H3K4me3, the pattern was straightforward: the consistent
  4519. pattern across all promoters was a single peak a few kb wide, with the
  4520. main axis of variation being the position of this peak relative to the
  4521. TSS (Figures
  4522. \begin_inset CommandInset ref
  4523. LatexCommand ref
  4524. reference "fig:H3K4me2-neighborhood"
  4525. plural "false"
  4526. caps "false"
  4527. noprefix "false"
  4528. \end_inset
  4529. &
  4530. \begin_inset CommandInset ref
  4531. LatexCommand ref
  4532. reference "fig:H3K4me3-neighborhood"
  4533. plural "false"
  4534. caps "false"
  4535. noprefix "false"
  4536. \end_inset
  4537. ).
  4538. There were no obvious
  4539. \begin_inset Quotes eld
  4540. \end_inset
  4541. preferred
  4542. \begin_inset Quotes erd
  4543. \end_inset
  4544. positions, but rather a continuous distribution of relative positions ranging
  4545. all across the promoter region.
  4546. The association with gene expression was also straightforward: peaks closer
  4547. to the TSS were more strongly associated with elevated gene expression.
  4548. Coverage downstream of the TSS appears to be more strongly associated with
  4549. elevated expression than coverage the same distance upstream, indicating
  4550. that the
  4551. \begin_inset Quotes eld
  4552. \end_inset
  4553. effective promoter region
  4554. \begin_inset Quotes erd
  4555. \end_inset
  4556. for H3K4me2 and H3K4me3 may be centered downstream of the TSS.
  4557. \end_layout
  4558. \begin_layout Standard
  4559. The relative promoter coverage for H3K27me3 had a more complex pattern,
  4560. with two specific patterns of promoter coverage associated with elevated
  4561. expression: a sharp depletion of H3K27me3 around the TSS relative to the
  4562. surrounding area, and a depletion of H3K27me3 downstream of the TSS relative
  4563. to upstream (Figure
  4564. \begin_inset CommandInset ref
  4565. LatexCommand ref
  4566. reference "fig:H3K27me3-neighborhood"
  4567. plural "false"
  4568. caps "false"
  4569. noprefix "false"
  4570. \end_inset
  4571. ).
  4572. A previous study found that H3K27me3 depletion within the gene body was
  4573. associated with elevated gene expression in 4 different cell types in mice
  4574. \begin_inset CommandInset citation
  4575. LatexCommand cite
  4576. key "Young2011"
  4577. literal "false"
  4578. \end_inset
  4579. .
  4580. This is consistent with the second pattern described here.
  4581. This study also reported that a spike in coverage at the TSS was associated
  4582. with
  4583. \emph on
  4584. lower
  4585. \emph default
  4586. expression, which is indirectly consistent with the first pattern described
  4587. here, in the sense that it associates lower H3K27me3 levels near the TSS
  4588. with higher expression.
  4589. \end_layout
  4590. \begin_layout Subsection
  4591. Workflow
  4592. \end_layout
  4593. \begin_layout Standard
  4594. \begin_inset ERT
  4595. status open
  4596. \begin_layout Plain Layout
  4597. \backslash
  4598. afterpage{
  4599. \end_layout
  4600. \begin_layout Plain Layout
  4601. \backslash
  4602. begin{landscape}
  4603. \end_layout
  4604. \end_inset
  4605. \end_layout
  4606. \begin_layout Standard
  4607. \begin_inset Float figure
  4608. wide false
  4609. sideways false
  4610. status open
  4611. \begin_layout Plain Layout
  4612. \align center
  4613. \begin_inset Graphics
  4614. filename graphics/CD4-csaw/rulegraphs/rulegraph-all.pdf
  4615. lyxscale 50
  4616. width 100col%
  4617. height 95theight%
  4618. \end_inset
  4619. \end_layout
  4620. \begin_layout Plain Layout
  4621. \begin_inset Caption Standard
  4622. \begin_layout Plain Layout
  4623. \begin_inset CommandInset label
  4624. LatexCommand label
  4625. name "fig:rulegraph"
  4626. \end_inset
  4627. \series bold
  4628. Dependency graph of steps in reproducible workflow.
  4629. \end_layout
  4630. \end_inset
  4631. \end_layout
  4632. \end_inset
  4633. \end_layout
  4634. \begin_layout Standard
  4635. \begin_inset ERT
  4636. status open
  4637. \begin_layout Plain Layout
  4638. \backslash
  4639. end{landscape}
  4640. \end_layout
  4641. \begin_layout Plain Layout
  4642. }
  4643. \end_layout
  4644. \end_inset
  4645. \end_layout
  4646. \begin_layout Standard
  4647. The analyses described in this chapter were organized into a reproducible
  4648. workflow using the Snakemake workflow management system.
  4649. As shown in Figure
  4650. \begin_inset CommandInset ref
  4651. LatexCommand ref
  4652. reference "fig:rulegraph"
  4653. plural "false"
  4654. caps "false"
  4655. noprefix "false"
  4656. \end_inset
  4657. , the workflow includes many steps with complex dependencies between them.
  4658. For example, the step that counts the number of ChIP-seq reads in 500
  4659. \begin_inset space ~
  4660. \end_inset
  4661. bp windows in each promoter (the starting point for Figures
  4662. \begin_inset CommandInset ref
  4663. LatexCommand ref
  4664. reference "fig:H3K4me2-neighborhood"
  4665. plural "false"
  4666. caps "false"
  4667. noprefix "false"
  4668. \end_inset
  4669. ,
  4670. \begin_inset CommandInset ref
  4671. LatexCommand ref
  4672. reference "fig:H3K4me3-neighborhood"
  4673. plural "false"
  4674. caps "false"
  4675. noprefix "false"
  4676. \end_inset
  4677. , and
  4678. \begin_inset CommandInset ref
  4679. LatexCommand ref
  4680. reference "fig:H3K27me3-neighborhood"
  4681. plural "false"
  4682. caps "false"
  4683. noprefix "false"
  4684. \end_inset
  4685. ), named
  4686. \begin_inset Formula $\texttt{chipseq\_count\_tss\_neighborhoods}$
  4687. \end_inset
  4688. , depends on the RNA-seq abundance estimates in order to select the most-used
  4689. TSS for each gene, the aligned ChIP-seq reads, the index for those reads,
  4690. and the blacklist of regions to be excluded from ChIP-seq analysis.
  4691. Each step declares its inputs and outputs, and Snakemake uses these to
  4692. determine the dependencies between steps.
  4693. Each step is marked as depending on all the steps whose outputs match its
  4694. inputs, generating the workflow graph in Figure
  4695. \begin_inset CommandInset ref
  4696. LatexCommand ref
  4697. reference "fig:rulegraph"
  4698. plural "false"
  4699. caps "false"
  4700. noprefix "false"
  4701. \end_inset
  4702. , which Snakemake uses to determine order in which to execute each step
  4703. so that each step is executed only after all of the steps it depends on
  4704. have completed, thereby automating the entire workflow from start to finish.
  4705. \end_layout
  4706. \begin_layout Standard
  4707. In addition to simply making it easier to organize the steps in the analysis,
  4708. structuring the analysis as a workflow allowed for some analysis strategies
  4709. that would not have been practical otherwise.
  4710. For example, 5 different RNA-seq quantification methods were tested against
  4711. two different reference transcriptome annotations for a total of 10 different
  4712. quantifications of the same RNA-seq data.
  4713. These were then compared against each other in the exploratory data analysis
  4714. step, to determine that the results were not very sensitive to either the
  4715. choice of quantification method or the choice of annotation.
  4716. This was possible with a single script for the exploratory data analysis,
  4717. because Snakemake was able to automate running this script for every combinatio
  4718. n of method and reference.
  4719. In a similar manner, two different peak calling methods were tested against
  4720. each other, and in this case it was determined that SICER was unambiguously
  4721. superior to MACS for all histone marks studied.
  4722. By enabling these types of comparisons, structuring the analysis as an
  4723. automated workflow allowed important analysis decisions to be made in a
  4724. data-driven way, by running every reasonable option through the downstream
  4725. steps, seeing the consequences of choosing each option, and deciding accordingl
  4726. y.
  4727. \end_layout
  4728. \begin_layout Subsection
  4729. Data quality issues limit conclusions
  4730. \end_layout
  4731. \begin_layout Standard
  4732. \begin_inset Flex TODO Note (inline)
  4733. status open
  4734. \begin_layout Plain Layout
  4735. Is this needed?
  4736. \end_layout
  4737. \end_inset
  4738. \end_layout
  4739. \begin_layout Chapter
  4740. Improving array-based diagnostics for transplant rejection by optimizing
  4741. data preprocessing
  4742. \end_layout
  4743. \begin_layout Standard
  4744. \begin_inset Note Note
  4745. status open
  4746. \begin_layout Plain Layout
  4747. Chapter author list: Me, Sunil, Tom, Padma, Dan
  4748. \end_layout
  4749. \end_inset
  4750. \end_layout
  4751. \begin_layout Section
  4752. Approach
  4753. \end_layout
  4754. \begin_layout Subsection
  4755. Proper pre-processing is essential for array data
  4756. \end_layout
  4757. \begin_layout Standard
  4758. \begin_inset Flex TODO Note (inline)
  4759. status open
  4760. \begin_layout Plain Layout
  4761. This section could probably use some citations
  4762. \end_layout
  4763. \end_inset
  4764. \end_layout
  4765. \begin_layout Standard
  4766. Microarrays, bead arrays, and similar assays produce raw data in the form
  4767. of fluorescence intensity measurements, with the each intensity measurement
  4768. proportional to the abundance of some fluorescently-labelled target DNA
  4769. or RNA sequence that base pairs to a specific probe sequence.
  4770. However, these measurements for each probe are also affected my many technical
  4771. confounding factors, such as the concentration of target material, strength
  4772. of off-target binding, and the sensitivity of the imaging sensor.
  4773. Some array designs also use multiple probe sequences for each target.
  4774. Hence, extensive pre-processing of array data is necessary to normalize
  4775. out the effects of these technical factors and summarize the information
  4776. from multiple probes to arrive at a single usable estimate of abundance
  4777. or other relevant quantity, such as a ratio of two abundances, for each
  4778. target.
  4779. \end_layout
  4780. \begin_layout Standard
  4781. The choice of pre-processing algorithms used in the analysis of an array
  4782. data set can have a large effect on the results of that analysis.
  4783. However, despite their importance, these steps are often neglected or rushed
  4784. in order to get to the more scientifically interesting analysis steps involving
  4785. the actual biology of the system under study.
  4786. Hence, it is often possible to achieve substantial gains in statistical
  4787. power, model goodness-of-fit, or other relevant performance measures, by
  4788. checking the assumptions made by each preprocessing step and choosing specific
  4789. normalization methods tailored to the specific goals of the current analysis.
  4790. \end_layout
  4791. \begin_layout Subsection
  4792. Clinical diagnostic applications for microarrays require single-channel
  4793. normalization
  4794. \end_layout
  4795. \begin_layout Standard
  4796. As the cost of performing microarray assays falls, there is increasing interest
  4797. in using genomic assays for diagnostic purposes, such as distinguishing
  4798. healthy transplants (TX) from transplants undergoing acute rejection (AR)
  4799. or acute dysfunction with no rejection (ADNR).
  4800. However, the the standard normalization algorithm used for microarray data,
  4801. Robust Multi-chip Average (RMA)
  4802. \begin_inset CommandInset citation
  4803. LatexCommand cite
  4804. key "Irizarry2003a"
  4805. literal "false"
  4806. \end_inset
  4807. , is not applicable in a clinical setting.
  4808. Two of the steps in RMA, quantile normalization and probe summarization
  4809. by median polish, depend on every array in the data set being normalized.
  4810. This means that adding or removing any arrays from a data set changes the
  4811. normalized values for all arrays, and data sets that have been normalized
  4812. separately cannot be compared to each other.
  4813. Hence, when using RMA, any arrays to be analyzed together must also be
  4814. normalized together, and the set of arrays included in the data set must
  4815. be held constant throughout an analysis.
  4816. \end_layout
  4817. \begin_layout Standard
  4818. These limitations present serious impediments to the use of arrays as a
  4819. diagnostic tool.
  4820. When training a classifier, the samples to be classified must not be involved
  4821. in any step of the training process, lest their inclusion bias the training
  4822. process.
  4823. Once a classifier is deployed in a clinical setting, the samples to be
  4824. classified will not even
  4825. \emph on
  4826. exist
  4827. \emph default
  4828. at the time of training, so including them would be impossible even if
  4829. it were statistically justifiable.
  4830. Therefore, any machine learning application for microarrays demands that
  4831. the normalized expression values computed for an array must depend only
  4832. on information contained within that array.
  4833. This would ensure that each array's normalization is independent of every
  4834. other array, and that arrays normalized separately can still be compared
  4835. to each other without bias.
  4836. Such a normalization is commonly referred to as
  4837. \begin_inset Quotes eld
  4838. \end_inset
  4839. single-channel normalization
  4840. \begin_inset Quotes erd
  4841. \end_inset
  4842. .
  4843. \end_layout
  4844. \begin_layout Standard
  4845. Frozen RMA (fRMA) addresses these concerns by replacing the quantile normalizati
  4846. on and median polish with alternatives that do not introduce inter-array
  4847. dependence, allowing each array to be normalized independently of all others
  4848. \begin_inset CommandInset citation
  4849. LatexCommand cite
  4850. key "McCall2010"
  4851. literal "false"
  4852. \end_inset
  4853. .
  4854. Quantile normalization is performed against a pre-generated set of quantiles
  4855. learned from a collection of 850 publically available arrays sampled from
  4856. a wide variety of tissues in the Gene Expression Omnibus (GEO).
  4857. Each array's probe intensity distribution is normalized against these pre-gener
  4858. ated quantiles.
  4859. The median polish step is replaced with a robust weighted average of probe
  4860. intensities, using inverse variance weights learned from the same public
  4861. GEO data.
  4862. The result is a normalization that satisfies the requirements mentioned
  4863. above: each array is normalized independently of all others, and any two
  4864. normalized arrays can be compared directly to each other.
  4865. \end_layout
  4866. \begin_layout Standard
  4867. One important limitation of fRMA is that it requires a separate reference
  4868. data set from which to learn the parameters (reference quantiles and probe
  4869. weights) that will be used to normalize each array.
  4870. These parameters are specific to a given array platform, and pre-generated
  4871. parameters are only provided for the most common platforms, such as Affymetrix
  4872. hgu133plus2.
  4873. For a less common platform, such as hthgu133pluspm, is is necessary to
  4874. learn custom parameters from in-house data before fRMA can be used to normalize
  4875. samples on that platform
  4876. \begin_inset CommandInset citation
  4877. LatexCommand cite
  4878. key "McCall2011"
  4879. literal "false"
  4880. \end_inset
  4881. .
  4882. \end_layout
  4883. \begin_layout Standard
  4884. One other option is the aptly-named Single Channel Array Normalization (SCAN),
  4885. which adapts a normalization method originally designed for tiling arrays
  4886. \begin_inset CommandInset citation
  4887. LatexCommand cite
  4888. key "Piccolo2012"
  4889. literal "false"
  4890. \end_inset
  4891. .
  4892. SCAN is truly single-channel in that it does not require a set of normalization
  4893. paramters estimated from an external set of reference samples like fRMA
  4894. does.
  4895. \end_layout
  4896. \begin_layout Subsection
  4897. Heteroskedasticity must be accounted for in methylation array data
  4898. \end_layout
  4899. \begin_layout Standard
  4900. DNA methylation arrays are a relatively new kind of assay that uses microarrays
  4901. to measure the degree of methylation on cytosines in specific regions arrayed
  4902. across the genome.
  4903. First, bisulfite treatment converts all unmethylated cytosines to uracil
  4904. (which then become thymine after amplication) while leaving methylated
  4905. cytosines unaffected.
  4906. Then, each target region is interrogated with two probes: one binds to
  4907. the original genomic sequence and interrogates the level of methylated
  4908. DNA, and the other binds to the same sequence with all cytosines replaced
  4909. by thymidines and interrogates the level of unmethylated DNA.
  4910. \end_layout
  4911. \begin_layout Standard
  4912. \begin_inset Float figure
  4913. wide false
  4914. sideways false
  4915. status collapsed
  4916. \begin_layout Plain Layout
  4917. \align center
  4918. \begin_inset Graphics
  4919. filename graphics/methylvoom/sigmoid.pdf
  4920. lyxscale 50
  4921. width 60col%
  4922. groupId colwidth
  4923. \end_inset
  4924. \end_layout
  4925. \begin_layout Plain Layout
  4926. \begin_inset Caption Standard
  4927. \begin_layout Plain Layout
  4928. \begin_inset CommandInset label
  4929. LatexCommand label
  4930. name "fig:Sigmoid-beta-m-mapping"
  4931. \end_inset
  4932. \series bold
  4933. Sigmoid shape of the mapping between β and M values
  4934. \end_layout
  4935. \end_inset
  4936. \end_layout
  4937. \end_inset
  4938. \end_layout
  4939. \begin_layout Standard
  4940. After normalization, these two probe intensities are summarized in one of
  4941. two ways, each with advantages and disadvantages.
  4942. β
  4943. \series bold
  4944. \series default
  4945. values, interpreted as fraction of DNA copies methylated, range from 0 to
  4946. 1.
  4947. β
  4948. \series bold
  4949. \series default
  4950. values are conceptually easy to interpret, but the constrained range makes
  4951. them unsuitable for linear modeling, and their error distributions are
  4952. highly non-normal, which also frustrates linear modeling.
  4953. M-values, interpreted as the log ratio of methylated to unmethylated copies,
  4954. are computed by mapping the beta values from
  4955. \begin_inset Formula $[0,1]$
  4956. \end_inset
  4957. onto
  4958. \begin_inset Formula $(-\infty,+\infty)$
  4959. \end_inset
  4960. using a sigmoid curve (Figure
  4961. \begin_inset CommandInset ref
  4962. LatexCommand ref
  4963. reference "fig:Sigmoid-beta-m-mapping"
  4964. plural "false"
  4965. caps "false"
  4966. noprefix "false"
  4967. \end_inset
  4968. ).
  4969. This transformation results in values with better statistical perperties:
  4970. the unconstrained range is suitable for linear modeling, and the error
  4971. distributions are more normal.
  4972. Hence, most linear modeling and other statistical testing on methylation
  4973. arrays is performed using M-values.
  4974. \end_layout
  4975. \begin_layout Standard
  4976. However, the steep slope of the sigmoid transformation near 0 and 1 tends
  4977. to over-exaggerate small differences in β values near those extremes, which
  4978. in turn amplifies the error in those values, leading to a U-shaped trend
  4979. in the mean-variance curve: extreme values have higher variances than values
  4980. near the middle.
  4981. This mean-variance dependency must be accounted for when fitting the linear
  4982. model for differential methylation, or else the variance will be systematically
  4983. overestimated for probes with moderate M-values and underestimated for
  4984. probes with extreme M-values.
  4985. This is particularly undesirable for methylation data because the intermediate
  4986. M-values are the ones of most interest, since they are more likely to represent
  4987. areas of varying methylation, whereas extreme M-values typically represent
  4988. complete methylation or complete lack of methylation.
  4989. \end_layout
  4990. \begin_layout Standard
  4991. RNA-seq read count data are also known to show heteroskedasticity, and the
  4992. voom method was introduced for modeling this heteroskedasticity by estimating
  4993. the mean-variance trend in the data and using this trend to assign precision
  4994. weights to each observation
  4995. \begin_inset CommandInset citation
  4996. LatexCommand cite
  4997. key "Law2013"
  4998. literal "false"
  4999. \end_inset
  5000. .
  5001. While methylation array data are not derived from counts and have a very
  5002. different mean-variance relationship from that of typical RNA-seq data,
  5003. the voom method makes no specific assumptions on the shape of the mean-variance
  5004. relationship – it only assumes that the relationship can be modeled as
  5005. a smooth curve.
  5006. Hence, the method is sufficiently general to model the mean-variance relationsh
  5007. ip in methylation array data.
  5008. However, the standard implementation of voom assumes that the input is
  5009. given in raw read counts, and it must be adapted to run on methylation
  5010. M-values.
  5011. \end_layout
  5012. \begin_layout Section
  5013. Methods
  5014. \end_layout
  5015. \begin_layout Subsection
  5016. Evaluation of classifier performance with different normalization methods
  5017. \end_layout
  5018. \begin_layout Standard
  5019. For testing different expression microarray normalizations, a data set of
  5020. 157 hgu133plus2 arrays was used, consisting of blood samples from kidney
  5021. transplant patients whose grafts had been graded as TX, AR, or ADNR via
  5022. biopsy and histology (46 TX, 69 AR, 42 ADNR)
  5023. \begin_inset CommandInset citation
  5024. LatexCommand cite
  5025. key "Kurian2014"
  5026. literal "true"
  5027. \end_inset
  5028. .
  5029. Additionally, an external validation set of 75 samples was gathered from
  5030. public GEO data (37 TX, 38 AR, no ADNR).
  5031. \end_layout
  5032. \begin_layout Standard
  5033. \begin_inset Flex TODO Note (inline)
  5034. status open
  5035. \begin_layout Plain Layout
  5036. Find appropriate GEO identifiers if possible.
  5037. Kurian 2014 says GSE15296, but this seems to be different data.
  5038. I also need to look up the GEO accession for the external validation set.
  5039. \end_layout
  5040. \end_inset
  5041. \end_layout
  5042. \begin_layout Standard
  5043. To evaluate the effect of each normalization on classifier performance,
  5044. the same classifier training and validation procedure was used after each
  5045. normalization method.
  5046. The PAM package was used to train a nearest shrunken centroid classifier
  5047. on the training set and select the appropriate threshold for centroid shrinking.
  5048. Then the trained classifier was used to predict the class probabilities
  5049. of each validation sample.
  5050. From these class probabilities, ROC curves and area-under-curve (AUC) values
  5051. were generated
  5052. \begin_inset CommandInset citation
  5053. LatexCommand cite
  5054. key "Turck2011"
  5055. literal "false"
  5056. \end_inset
  5057. .
  5058. Each normalization was tested on two different sets of training and validation
  5059. samples.
  5060. For internal validation, the 115 TX and AR arrays in the internal set were
  5061. split at random into two equal sized sets, one for training and one for
  5062. validation, each containing the same numbers of TX and AR samples as the
  5063. other set.
  5064. For external validation, the full set of 115 TX and AR samples were used
  5065. as a training set, and the 75 external TX and AR samples were used as the
  5066. validation set.
  5067. Thus, 2 ROC curves and AUC values were generated for each normalization
  5068. method: one internal and one external.
  5069. Because the external validation set contains no ADNR samples, only classificati
  5070. on of TX and AR samples was considered.
  5071. The ADNR samples were included during normalization but excluded from all
  5072. classifier training and validation.
  5073. This ensures that the performance on internal and external validation sets
  5074. is directly comparable, since both are performing the same task: distinguising
  5075. TX from AR.
  5076. \end_layout
  5077. \begin_layout Standard
  5078. \begin_inset Flex TODO Note (inline)
  5079. status open
  5080. \begin_layout Plain Layout
  5081. Summarize the get.best.threshold algorithm for PAM threshold selection, or
  5082. just put the code online?
  5083. \end_layout
  5084. \end_inset
  5085. \end_layout
  5086. \begin_layout Standard
  5087. Six different normalization strategies were evaluated.
  5088. First, 2 well-known non-single-channel normalization methods were considered:
  5089. RMA and dChip
  5090. \begin_inset CommandInset citation
  5091. LatexCommand cite
  5092. key "Li2001,Irizarry2003a"
  5093. literal "false"
  5094. \end_inset
  5095. .
  5096. Since RMA produces expression values on a log2 scale and dChip does not,
  5097. the values from dChip were log2 transformed after normalization.
  5098. Next, RMA and dChip followed by Global Rank-invariant Set Normalization
  5099. (GRSN) were tested
  5100. \begin_inset CommandInset citation
  5101. LatexCommand cite
  5102. key "Pelz2008"
  5103. literal "false"
  5104. \end_inset
  5105. .
  5106. Post-processing with GRSN does not turn RMA or dChip into single-channel
  5107. methods, but it may help mitigate batch effects and is therefore useful
  5108. as a benchmark.
  5109. Lastly, the two single-channel normalization methods, fRMA and SCAN, were
  5110. tested
  5111. \begin_inset CommandInset citation
  5112. LatexCommand cite
  5113. key "McCall2010,Piccolo2012"
  5114. literal "false"
  5115. \end_inset
  5116. .
  5117. When evaluting internal validation performance, only the 157 internal samples
  5118. were normalized; when evaluating external validation performance, all 157
  5119. internal samples and 75 external samples were normalized together.
  5120. \end_layout
  5121. \begin_layout Standard
  5122. For demonstrating the problem with separate normalization of training and
  5123. validation data, one additional normalization was performed: the internal
  5124. and external sets were each normalized separately using RMA, and the normalized
  5125. data for each set were combined into a single set with no further attempts
  5126. at normalizing between the two sets.
  5127. The represents approximately how RMA would have to be used in a clinical
  5128. setting, where the samples to be classified are not available at the time
  5129. the classifier is trained.
  5130. \end_layout
  5131. \begin_layout Subsection
  5132. Generating custom fRMA vectors for hthgu133pluspm array platform
  5133. \end_layout
  5134. \begin_layout Standard
  5135. In order to enable fRMA normalization for the hthgu133pluspm array platform,
  5136. custom fRMA normalization vectors were trained using the frmaTools package
  5137. \begin_inset CommandInset citation
  5138. LatexCommand cite
  5139. key "McCall2011"
  5140. literal "false"
  5141. \end_inset
  5142. .
  5143. Separate vectors were created for two types of samples: kidney graft biopsy
  5144. samples and blood samples from graft recipients.
  5145. For training, a 341 kidney biopsy samples from 2 data sets and 965 blood
  5146. samples from 5 data sets were used as the reference set.
  5147. Arrays were groups into batches based on unique combinations of sample
  5148. type (blood or biopsy), diagnosis (TX, AR, etc.), data set, and scan date.
  5149. Thus, each batch represents arrays of the same kind that were run together
  5150. on the same day.
  5151. For estimating the probe inverse variance weights, frmaTools requires equal-siz
  5152. ed batches, which means a batch size must be chosen, and then batches smaller
  5153. than that size must be ignored, while batches larger than the chosen size
  5154. must be downsampled.
  5155. This downsampling is performed randomly, so the sampling process is repeated
  5156. 5 times and the resulting normalizations are compared to each other.
  5157. \end_layout
  5158. \begin_layout Standard
  5159. To evaluate the consistency of the generated normalization vectors, the
  5160. 5 fRMA vector sets generated from 5 random batch samplings were each used
  5161. to normalize the same 20 randomly selected samples from each tissue.
  5162. Then the normalized expression values for each probe on each array were
  5163. compared across all normalizations.
  5164. Each fRMA normalization was also compared against the normalized expression
  5165. values obtained by normalizing the same 20 samples with ordinary RMA.
  5166. \end_layout
  5167. \begin_layout Subsection
  5168. Modeling methylation array M-value heteroskedasticy in linear models with
  5169. modified voom implementation
  5170. \end_layout
  5171. \begin_layout Standard
  5172. \begin_inset Flex TODO Note (inline)
  5173. status open
  5174. \begin_layout Plain Layout
  5175. Put code on Github and reference it.
  5176. \end_layout
  5177. \end_inset
  5178. \end_layout
  5179. \begin_layout Standard
  5180. To investigate the whether DNA methylation could be used to distinguish
  5181. between healthy and dysfunctional transplants, a data set of 78 Illumina
  5182. 450k methylation arrays from human kidney graft biopsies was analyzed for
  5183. differential metylation between 4 transplant statuses: healthy transplant
  5184. (TX), transplants undergoing acute rejection (AR), acute dysfunction with
  5185. no rejection (ADNR), and chronic allograpft nephropathy (CAN).
  5186. The data consisted of 33 TX, 9 AR, 8 ADNR, and 28 CAN samples.
  5187. The uneven group sizes are a result of taking the biopsy samples before
  5188. the eventual fate of the transplant was known.
  5189. Each sample was additionally annotated with a donor ID (anonymized), Sex,
  5190. Age, Ethnicity, Creatinine Level, and Diabetes diagnosois (all samples
  5191. in this data set came from patients with either Type 1 or Type 2 diabetes).
  5192. \end_layout
  5193. \begin_layout Standard
  5194. The intensity data were first normalized using subset-quantile within array
  5195. normalization (SWAN)
  5196. \begin_inset CommandInset citation
  5197. LatexCommand cite
  5198. key "Maksimovic2012"
  5199. literal "false"
  5200. \end_inset
  5201. , then converted to intensity ratios (beta values)
  5202. \begin_inset CommandInset citation
  5203. LatexCommand cite
  5204. key "Aryee2014"
  5205. literal "false"
  5206. \end_inset
  5207. .
  5208. Any probes binding to loci that overlapped annotated SNPs were dropped,
  5209. and the annotated sex of each sample was verified against the sex inferred
  5210. from the ratio of median probe intensities for the X and Y chromosomes.
  5211. Then, the ratios were transformed to M-values.
  5212. \end_layout
  5213. \begin_layout Standard
  5214. \begin_inset Float table
  5215. wide false
  5216. sideways false
  5217. status open
  5218. \begin_layout Plain Layout
  5219. \align center
  5220. \begin_inset Tabular
  5221. <lyxtabular version="3" rows="4" columns="6">
  5222. <features tabularvalignment="middle">
  5223. <column alignment="center" valignment="top">
  5224. <column alignment="center" valignment="top">
  5225. <column alignment="center" valignment="top">
  5226. <column alignment="center" valignment="top">
  5227. <column alignment="center" valignment="top">
  5228. <column alignment="center" valignment="top">
  5229. <row>
  5230. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  5231. \begin_inset Text
  5232. \begin_layout Plain Layout
  5233. Analysis
  5234. \end_layout
  5235. \end_inset
  5236. </cell>
  5237. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  5238. \begin_inset Text
  5239. \begin_layout Plain Layout
  5240. random effect
  5241. \end_layout
  5242. \end_inset
  5243. </cell>
  5244. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  5245. \begin_inset Text
  5246. \begin_layout Plain Layout
  5247. eBayes
  5248. \end_layout
  5249. \end_inset
  5250. </cell>
  5251. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  5252. \begin_inset Text
  5253. \begin_layout Plain Layout
  5254. SVA
  5255. \end_layout
  5256. \end_inset
  5257. </cell>
  5258. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  5259. \begin_inset Text
  5260. \begin_layout Plain Layout
  5261. weights
  5262. \end_layout
  5263. \end_inset
  5264. </cell>
  5265. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  5266. \begin_inset Text
  5267. \begin_layout Plain Layout
  5268. voom
  5269. \end_layout
  5270. \end_inset
  5271. </cell>
  5272. </row>
  5273. <row>
  5274. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  5275. \begin_inset Text
  5276. \begin_layout Plain Layout
  5277. A
  5278. \end_layout
  5279. \end_inset
  5280. </cell>
  5281. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  5282. \begin_inset Text
  5283. \begin_layout Plain Layout
  5284. Yes
  5285. \end_layout
  5286. \end_inset
  5287. </cell>
  5288. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  5289. \begin_inset Text
  5290. \begin_layout Plain Layout
  5291. Yes
  5292. \end_layout
  5293. \end_inset
  5294. </cell>
  5295. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  5296. \begin_inset Text
  5297. \begin_layout Plain Layout
  5298. No
  5299. \end_layout
  5300. \end_inset
  5301. </cell>
  5302. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  5303. \begin_inset Text
  5304. \begin_layout Plain Layout
  5305. No
  5306. \end_layout
  5307. \end_inset
  5308. </cell>
  5309. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  5310. \begin_inset Text
  5311. \begin_layout Plain Layout
  5312. No
  5313. \end_layout
  5314. \end_inset
  5315. </cell>
  5316. </row>
  5317. <row>
  5318. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  5319. \begin_inset Text
  5320. \begin_layout Plain Layout
  5321. B
  5322. \end_layout
  5323. \end_inset
  5324. </cell>
  5325. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  5326. \begin_inset Text
  5327. \begin_layout Plain Layout
  5328. Yes
  5329. \end_layout
  5330. \end_inset
  5331. </cell>
  5332. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  5333. \begin_inset Text
  5334. \begin_layout Plain Layout
  5335. Yes
  5336. \end_layout
  5337. \end_inset
  5338. </cell>
  5339. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  5340. \begin_inset Text
  5341. \begin_layout Plain Layout
  5342. Yes
  5343. \end_layout
  5344. \end_inset
  5345. </cell>
  5346. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  5347. \begin_inset Text
  5348. \begin_layout Plain Layout
  5349. Yes
  5350. \end_layout
  5351. \end_inset
  5352. </cell>
  5353. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  5354. \begin_inset Text
  5355. \begin_layout Plain Layout
  5356. No
  5357. \end_layout
  5358. \end_inset
  5359. </cell>
  5360. </row>
  5361. <row>
  5362. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  5363. \begin_inset Text
  5364. \begin_layout Plain Layout
  5365. C
  5366. \end_layout
  5367. \end_inset
  5368. </cell>
  5369. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  5370. \begin_inset Text
  5371. \begin_layout Plain Layout
  5372. Yes
  5373. \end_layout
  5374. \end_inset
  5375. </cell>
  5376. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  5377. \begin_inset Text
  5378. \begin_layout Plain Layout
  5379. Yes
  5380. \end_layout
  5381. \end_inset
  5382. </cell>
  5383. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  5384. \begin_inset Text
  5385. \begin_layout Plain Layout
  5386. Yes
  5387. \end_layout
  5388. \end_inset
  5389. </cell>
  5390. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  5391. \begin_inset Text
  5392. \begin_layout Plain Layout
  5393. Yes
  5394. \end_layout
  5395. \end_inset
  5396. </cell>
  5397. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  5398. \begin_inset Text
  5399. \begin_layout Plain Layout
  5400. Yes
  5401. \end_layout
  5402. \end_inset
  5403. </cell>
  5404. </row>
  5405. </lyxtabular>
  5406. \end_inset
  5407. \end_layout
  5408. \begin_layout Plain Layout
  5409. \begin_inset Caption Standard
  5410. \begin_layout Plain Layout
  5411. \series bold
  5412. \begin_inset CommandInset label
  5413. LatexCommand label
  5414. name "tab:Summary-of-meth-analysis"
  5415. \end_inset
  5416. Summary of analysis variants for methylation array data.
  5417. \series default
  5418. Each analysis included a different set of steps to adjust or account for
  5419. various systematic features of the data.
  5420. Random effect: The model included a random effect accounting for correlation
  5421. between samples from the same patient
  5422. \begin_inset CommandInset citation
  5423. LatexCommand cite
  5424. key "Smyth2005a"
  5425. literal "false"
  5426. \end_inset
  5427. ; eBayes: Empirical bayes squeezing of per-probe variances toward the mean-varia
  5428. nce trend
  5429. \begin_inset CommandInset citation
  5430. LatexCommand cite
  5431. key "Ritchie2015"
  5432. literal "false"
  5433. \end_inset
  5434. ; SVA: Surrogate variable analysis to account for unobserved confounders
  5435. \begin_inset CommandInset citation
  5436. LatexCommand cite
  5437. key "Leek2007"
  5438. literal "false"
  5439. \end_inset
  5440. ; Weights: Estimate sample weights to account for differences in sample
  5441. quality
  5442. \begin_inset CommandInset citation
  5443. LatexCommand cite
  5444. key "Liu2015,Ritchie2006"
  5445. literal "false"
  5446. \end_inset
  5447. ; voom: Use mean-variance trend to assign individual sample weights
  5448. \begin_inset CommandInset citation
  5449. LatexCommand cite
  5450. key "Law2013"
  5451. literal "false"
  5452. \end_inset
  5453. .
  5454. See the text for a more detailed explanation of each step.
  5455. \end_layout
  5456. \end_inset
  5457. \end_layout
  5458. \end_inset
  5459. \end_layout
  5460. \begin_layout Standard
  5461. From the M-values, a series of parallel analyses was performed, each adding
  5462. additional steps into the model fit to accomodate a feature of the data
  5463. (see Table
  5464. \begin_inset CommandInset ref
  5465. LatexCommand ref
  5466. reference "tab:Summary-of-meth-analysis"
  5467. plural "false"
  5468. caps "false"
  5469. noprefix "false"
  5470. \end_inset
  5471. ).
  5472. For analysis A, a
  5473. \begin_inset Quotes eld
  5474. \end_inset
  5475. basic
  5476. \begin_inset Quotes erd
  5477. \end_inset
  5478. linear modeling analysis was performed, compensating for known confounders
  5479. by including terms for the factor of interest (transplant status) as well
  5480. as the known biological confounders: sex, age, ethnicity, and diabetes.
  5481. Since some samples came from the same patients at different times, the
  5482. intra-patient correlation was modeled as a random effect, estimating a
  5483. shared correlation value across all probes
  5484. \begin_inset CommandInset citation
  5485. LatexCommand cite
  5486. key "Smyth2005a"
  5487. literal "false"
  5488. \end_inset
  5489. .
  5490. Then the linear model was fit, and the variance was modeled using empirical
  5491. Bayes squeezing toward the mean-variance trend
  5492. \begin_inset CommandInset citation
  5493. LatexCommand cite
  5494. key "Ritchie2015"
  5495. literal "false"
  5496. \end_inset
  5497. .
  5498. Finally, t-tests or F-tests were performed as appropriate for each test:
  5499. t-tests for single contrasts, and F-tests for multiple contrasts.
  5500. P-values were corrected for multiple testing using the Benjamini-Hochberg
  5501. procedure for FDR control
  5502. \begin_inset CommandInset citation
  5503. LatexCommand cite
  5504. key "Benjamini1995"
  5505. literal "false"
  5506. \end_inset
  5507. .
  5508. \end_layout
  5509. \begin_layout Standard
  5510. For the analysis B, surrogate variable analysis (SVA) was used to infer
  5511. additional unobserved sources of heterogeneity in the data
  5512. \begin_inset CommandInset citation
  5513. LatexCommand cite
  5514. key "Leek2007"
  5515. literal "false"
  5516. \end_inset
  5517. .
  5518. These surrogate variables were added to the design matrix before fitting
  5519. the linear model.
  5520. In addition, sample quality weights were estimated from the data and used
  5521. during linear modeling to down-weight the contribution of highly variable
  5522. arrays while increasing the weight to arrays with lower variability
  5523. \begin_inset CommandInset citation
  5524. LatexCommand cite
  5525. key "Ritchie2006"
  5526. literal "false"
  5527. \end_inset
  5528. .
  5529. The remainder of the analysis proceeded as in analysis A.
  5530. For analysis C, the voom method was adapted to run on methylation array
  5531. data and used to model and correct for the mean-variance trend using individual
  5532. observation weights
  5533. \begin_inset CommandInset citation
  5534. LatexCommand cite
  5535. key "Law2013"
  5536. literal "false"
  5537. \end_inset
  5538. , which were combined with the sample weights
  5539. \begin_inset CommandInset citation
  5540. LatexCommand cite
  5541. key "Liu2015,Ritchie2006"
  5542. literal "false"
  5543. \end_inset
  5544. .
  5545. Each time weights were used, they were estimated once before estimating
  5546. the random effect correlation value, and then the weights were re-estimated
  5547. taking the random effect into account.
  5548. The remainder of the analysis proceeded as in analysis B.
  5549. \end_layout
  5550. \begin_layout Section
  5551. Results
  5552. \end_layout
  5553. \begin_layout Standard
  5554. \begin_inset Flex TODO Note (inline)
  5555. status open
  5556. \begin_layout Plain Layout
  5557. Improve subsection titles in this section
  5558. \end_layout
  5559. \end_inset
  5560. \end_layout
  5561. \begin_layout Subsection
  5562. Separate normalization with RMA introduces unwanted biases in classification
  5563. \end_layout
  5564. \begin_layout Standard
  5565. \begin_inset Float figure
  5566. wide false
  5567. sideways false
  5568. status open
  5569. \begin_layout Plain Layout
  5570. \align center
  5571. \begin_inset Graphics
  5572. filename graphics/PAM/predplot.pdf
  5573. lyxscale 50
  5574. width 60col%
  5575. groupId colwidth
  5576. \end_inset
  5577. \end_layout
  5578. \begin_layout Plain Layout
  5579. \begin_inset Caption Standard
  5580. \begin_layout Plain Layout
  5581. \begin_inset CommandInset label
  5582. LatexCommand label
  5583. name "fig:Classifier-probabilities-RMA"
  5584. \end_inset
  5585. \series bold
  5586. Classifier probabilities on validation samples when normalized with RMA
  5587. together vs.
  5588. separately.
  5589. \series default
  5590. The PAM classifier algorithm was trained on the training set of arrays to
  5591. distinguish AR from TX and then used to assign class probabilities to the
  5592. validation set.
  5593. The process was performed after normalizing all samples together and after
  5594. normalizing the training and test sets separately, and the class probabilities
  5595. assigned to each sample in the validation set were plotted against each
  5596. other (PP(AR), posterior probability of being AR).
  5597. The color of each point indicates the true classification of that sample.
  5598. \end_layout
  5599. \end_inset
  5600. \end_layout
  5601. \end_inset
  5602. \end_layout
  5603. \begin_layout Standard
  5604. To demonstrate the problem with non-single-channel normalization methods,
  5605. we considered the problem of training a classifier to distinguish TX from
  5606. AR using the samples from the internal set as training data, evaluating
  5607. performance on the external set.
  5608. First, training and evaluation were performed after normalizing all array
  5609. samples together as a single set using RMA, and second, the internal samples
  5610. were normalized separately from the external samples and the training and
  5611. evaluation were repeated.
  5612. For each sample in the validation set, the classifier probabilities from
  5613. both classifiers were plotted against each other (Fig.
  5614. \begin_inset CommandInset ref
  5615. LatexCommand ref
  5616. reference "fig:Classifier-probabilities-RMA"
  5617. plural "false"
  5618. caps "false"
  5619. noprefix "false"
  5620. \end_inset
  5621. ).
  5622. As expected, separate normalization biases the classifier probabilities,
  5623. resulting in several misclassifications.
  5624. In this case, the bias from separate normalization causes the classifier
  5625. to assign a lower probability of AR to every sample.
  5626. \end_layout
  5627. \begin_layout Subsection
  5628. fRMA and SCAN maintain classification performance while eliminating dependence
  5629. on normalization strategy
  5630. \end_layout
  5631. \begin_layout Standard
  5632. \begin_inset Float figure
  5633. wide false
  5634. sideways false
  5635. status open
  5636. \begin_layout Plain Layout
  5637. \align center
  5638. \begin_inset Float figure
  5639. placement tb
  5640. wide false
  5641. sideways false
  5642. status open
  5643. \begin_layout Plain Layout
  5644. \align center
  5645. \begin_inset Graphics
  5646. filename graphics/PAM/ROC-TXvsAR-internal.pdf
  5647. lyxscale 50
  5648. height 40theight%
  5649. groupId roc-pam
  5650. \end_inset
  5651. \end_layout
  5652. \begin_layout Plain Layout
  5653. \begin_inset Caption Standard
  5654. \begin_layout Plain Layout
  5655. \begin_inset CommandInset label
  5656. LatexCommand label
  5657. name "fig:ROC-PAM-int"
  5658. \end_inset
  5659. ROC curves for PAM on internal validation data
  5660. \end_layout
  5661. \end_inset
  5662. \end_layout
  5663. \end_inset
  5664. \end_layout
  5665. \begin_layout Plain Layout
  5666. \align center
  5667. \begin_inset Float figure
  5668. placement tb
  5669. wide false
  5670. sideways false
  5671. status open
  5672. \begin_layout Plain Layout
  5673. \align center
  5674. \begin_inset Graphics
  5675. filename graphics/PAM/ROC-TXvsAR-external.pdf
  5676. lyxscale 50
  5677. height 40theight%
  5678. groupId roc-pam
  5679. \end_inset
  5680. \end_layout
  5681. \begin_layout Plain Layout
  5682. \begin_inset Caption Standard
  5683. \begin_layout Plain Layout
  5684. \begin_inset CommandInset label
  5685. LatexCommand label
  5686. name "fig:ROC-PAM-ext"
  5687. \end_inset
  5688. ROC curves for PAM on external validation data
  5689. \end_layout
  5690. \end_inset
  5691. \end_layout
  5692. \end_inset
  5693. \end_layout
  5694. \begin_layout Plain Layout
  5695. \begin_inset Caption Standard
  5696. \begin_layout Plain Layout
  5697. \series bold
  5698. \begin_inset CommandInset label
  5699. LatexCommand label
  5700. name "fig:ROC-PAM-main"
  5701. \end_inset
  5702. ROC curves for PAM using different normalization strategies.
  5703. \series default
  5704. ROC curves were generated for PAM classification of AR vs TX after 6 different
  5705. normalization strategies applied to the same data sets.
  5706. Only fRMA and SCAN are single-channel normalizations.
  5707. The other normalizations are for comparison.
  5708. \end_layout
  5709. \end_inset
  5710. \end_layout
  5711. \end_inset
  5712. \end_layout
  5713. \begin_layout Standard
  5714. \begin_inset Float table
  5715. wide false
  5716. sideways false
  5717. status open
  5718. \begin_layout Plain Layout
  5719. \align center
  5720. \begin_inset Tabular
  5721. <lyxtabular version="3" rows="7" columns="4">
  5722. <features tabularvalignment="middle">
  5723. <column alignment="center" valignment="top">
  5724. <column alignment="center" valignment="top">
  5725. <column alignment="center" valignment="top">
  5726. <column alignment="center" valignment="top">
  5727. <row>
  5728. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  5729. \begin_inset Text
  5730. \begin_layout Plain Layout
  5731. \family roman
  5732. \series medium
  5733. \shape up
  5734. \size normal
  5735. \emph off
  5736. \bar no
  5737. \strikeout off
  5738. \xout off
  5739. \uuline off
  5740. \uwave off
  5741. \noun off
  5742. \color none
  5743. Normalization
  5744. \end_layout
  5745. \end_inset
  5746. </cell>
  5747. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  5748. \begin_inset Text
  5749. \begin_layout Plain Layout
  5750. Single-channel?
  5751. \end_layout
  5752. \end_inset
  5753. </cell>
  5754. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  5755. \begin_inset Text
  5756. \begin_layout Plain Layout
  5757. \family roman
  5758. \series medium
  5759. \shape up
  5760. \size normal
  5761. \emph off
  5762. \bar no
  5763. \strikeout off
  5764. \xout off
  5765. \uuline off
  5766. \uwave off
  5767. \noun off
  5768. \color none
  5769. Internal Val.
  5770. AUC
  5771. \end_layout
  5772. \end_inset
  5773. </cell>
  5774. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  5775. \begin_inset Text
  5776. \begin_layout Plain Layout
  5777. External Val.
  5778. AUC
  5779. \end_layout
  5780. \end_inset
  5781. </cell>
  5782. </row>
  5783. <row>
  5784. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  5785. \begin_inset Text
  5786. \begin_layout Plain Layout
  5787. \family roman
  5788. \series medium
  5789. \shape up
  5790. \size normal
  5791. \emph off
  5792. \bar no
  5793. \strikeout off
  5794. \xout off
  5795. \uuline off
  5796. \uwave off
  5797. \noun off
  5798. \color none
  5799. RMA
  5800. \end_layout
  5801. \end_inset
  5802. </cell>
  5803. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  5804. \begin_inset Text
  5805. \begin_layout Plain Layout
  5806. No
  5807. \end_layout
  5808. \end_inset
  5809. </cell>
  5810. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  5811. \begin_inset Text
  5812. \begin_layout Plain Layout
  5813. \family roman
  5814. \series medium
  5815. \shape up
  5816. \size normal
  5817. \emph off
  5818. \bar no
  5819. \strikeout off
  5820. \xout off
  5821. \uuline off
  5822. \uwave off
  5823. \noun off
  5824. \color none
  5825. 0.852
  5826. \end_layout
  5827. \end_inset
  5828. </cell>
  5829. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  5830. \begin_inset Text
  5831. \begin_layout Plain Layout
  5832. \family roman
  5833. \series medium
  5834. \shape up
  5835. \size normal
  5836. \emph off
  5837. \bar no
  5838. \strikeout off
  5839. \xout off
  5840. \uuline off
  5841. \uwave off
  5842. \noun off
  5843. \color none
  5844. 0.713
  5845. \end_layout
  5846. \end_inset
  5847. </cell>
  5848. </row>
  5849. <row>
  5850. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  5851. \begin_inset Text
  5852. \begin_layout Plain Layout
  5853. \family roman
  5854. \series medium
  5855. \shape up
  5856. \size normal
  5857. \emph off
  5858. \bar no
  5859. \strikeout off
  5860. \xout off
  5861. \uuline off
  5862. \uwave off
  5863. \noun off
  5864. \color none
  5865. dChip
  5866. \end_layout
  5867. \end_inset
  5868. </cell>
  5869. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  5870. \begin_inset Text
  5871. \begin_layout Plain Layout
  5872. No
  5873. \end_layout
  5874. \end_inset
  5875. </cell>
  5876. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  5877. \begin_inset Text
  5878. \begin_layout Plain Layout
  5879. \family roman
  5880. \series medium
  5881. \shape up
  5882. \size normal
  5883. \emph off
  5884. \bar no
  5885. \strikeout off
  5886. \xout off
  5887. \uuline off
  5888. \uwave off
  5889. \noun off
  5890. \color none
  5891. 0.891
  5892. \end_layout
  5893. \end_inset
  5894. </cell>
  5895. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  5896. \begin_inset Text
  5897. \begin_layout Plain Layout
  5898. \family roman
  5899. \series medium
  5900. \shape up
  5901. \size normal
  5902. \emph off
  5903. \bar no
  5904. \strikeout off
  5905. \xout off
  5906. \uuline off
  5907. \uwave off
  5908. \noun off
  5909. \color none
  5910. 0.657
  5911. \end_layout
  5912. \end_inset
  5913. </cell>
  5914. </row>
  5915. <row>
  5916. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  5917. \begin_inset Text
  5918. \begin_layout Plain Layout
  5919. \family roman
  5920. \series medium
  5921. \shape up
  5922. \size normal
  5923. \emph off
  5924. \bar no
  5925. \strikeout off
  5926. \xout off
  5927. \uuline off
  5928. \uwave off
  5929. \noun off
  5930. \color none
  5931. RMA + GRSN
  5932. \end_layout
  5933. \end_inset
  5934. </cell>
  5935. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  5936. \begin_inset Text
  5937. \begin_layout Plain Layout
  5938. No
  5939. \end_layout
  5940. \end_inset
  5941. </cell>
  5942. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  5943. \begin_inset Text
  5944. \begin_layout Plain Layout
  5945. \family roman
  5946. \series medium
  5947. \shape up
  5948. \size normal
  5949. \emph off
  5950. \bar no
  5951. \strikeout off
  5952. \xout off
  5953. \uuline off
  5954. \uwave off
  5955. \noun off
  5956. \color none
  5957. 0.816
  5958. \end_layout
  5959. \end_inset
  5960. </cell>
  5961. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  5962. \begin_inset Text
  5963. \begin_layout Plain Layout
  5964. \family roman
  5965. \series medium
  5966. \shape up
  5967. \size normal
  5968. \emph off
  5969. \bar no
  5970. \strikeout off
  5971. \xout off
  5972. \uuline off
  5973. \uwave off
  5974. \noun off
  5975. \color none
  5976. 0.750
  5977. \end_layout
  5978. \end_inset
  5979. </cell>
  5980. </row>
  5981. <row>
  5982. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  5983. \begin_inset Text
  5984. \begin_layout Plain Layout
  5985. \family roman
  5986. \series medium
  5987. \shape up
  5988. \size normal
  5989. \emph off
  5990. \bar no
  5991. \strikeout off
  5992. \xout off
  5993. \uuline off
  5994. \uwave off
  5995. \noun off
  5996. \color none
  5997. dChip + GRSN
  5998. \end_layout
  5999. \end_inset
  6000. </cell>
  6001. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  6002. \begin_inset Text
  6003. \begin_layout Plain Layout
  6004. No
  6005. \end_layout
  6006. \end_inset
  6007. </cell>
  6008. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  6009. \begin_inset Text
  6010. \begin_layout Plain Layout
  6011. \family roman
  6012. \series medium
  6013. \shape up
  6014. \size normal
  6015. \emph off
  6016. \bar no
  6017. \strikeout off
  6018. \xout off
  6019. \uuline off
  6020. \uwave off
  6021. \noun off
  6022. \color none
  6023. 0.875
  6024. \end_layout
  6025. \end_inset
  6026. </cell>
  6027. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  6028. \begin_inset Text
  6029. \begin_layout Plain Layout
  6030. \family roman
  6031. \series medium
  6032. \shape up
  6033. \size normal
  6034. \emph off
  6035. \bar no
  6036. \strikeout off
  6037. \xout off
  6038. \uuline off
  6039. \uwave off
  6040. \noun off
  6041. \color none
  6042. 0.642
  6043. \end_layout
  6044. \end_inset
  6045. </cell>
  6046. </row>
  6047. <row>
  6048. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  6049. \begin_inset Text
  6050. \begin_layout Plain Layout
  6051. \family roman
  6052. \series medium
  6053. \shape up
  6054. \size normal
  6055. \emph off
  6056. \bar no
  6057. \strikeout off
  6058. \xout off
  6059. \uuline off
  6060. \uwave off
  6061. \noun off
  6062. \color none
  6063. fRMA
  6064. \end_layout
  6065. \end_inset
  6066. </cell>
  6067. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  6068. \begin_inset Text
  6069. \begin_layout Plain Layout
  6070. Yes
  6071. \end_layout
  6072. \end_inset
  6073. </cell>
  6074. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  6075. \begin_inset Text
  6076. \begin_layout Plain Layout
  6077. \family roman
  6078. \series medium
  6079. \shape up
  6080. \size normal
  6081. \emph off
  6082. \bar no
  6083. \strikeout off
  6084. \xout off
  6085. \uuline off
  6086. \uwave off
  6087. \noun off
  6088. \color none
  6089. 0.863
  6090. \end_layout
  6091. \end_inset
  6092. </cell>
  6093. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  6094. \begin_inset Text
  6095. \begin_layout Plain Layout
  6096. \family roman
  6097. \series medium
  6098. \shape up
  6099. \size normal
  6100. \emph off
  6101. \bar no
  6102. \strikeout off
  6103. \xout off
  6104. \uuline off
  6105. \uwave off
  6106. \noun off
  6107. \color none
  6108. 0.718
  6109. \end_layout
  6110. \end_inset
  6111. </cell>
  6112. </row>
  6113. <row>
  6114. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  6115. \begin_inset Text
  6116. \begin_layout Plain Layout
  6117. \family roman
  6118. \series medium
  6119. \shape up
  6120. \size normal
  6121. \emph off
  6122. \bar no
  6123. \strikeout off
  6124. \xout off
  6125. \uuline off
  6126. \uwave off
  6127. \noun off
  6128. \color none
  6129. SCAN
  6130. \end_layout
  6131. \end_inset
  6132. </cell>
  6133. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  6134. \begin_inset Text
  6135. \begin_layout Plain Layout
  6136. Yes
  6137. \end_layout
  6138. \end_inset
  6139. </cell>
  6140. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  6141. \begin_inset Text
  6142. \begin_layout Plain Layout
  6143. \family roman
  6144. \series medium
  6145. \shape up
  6146. \size normal
  6147. \emph off
  6148. \bar no
  6149. \strikeout off
  6150. \xout off
  6151. \uuline off
  6152. \uwave off
  6153. \noun off
  6154. \color none
  6155. 0.853
  6156. \end_layout
  6157. \end_inset
  6158. </cell>
  6159. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  6160. \begin_inset Text
  6161. \begin_layout Plain Layout
  6162. \family roman
  6163. \series medium
  6164. \shape up
  6165. \size normal
  6166. \emph off
  6167. \bar no
  6168. \strikeout off
  6169. \xout off
  6170. \uuline off
  6171. \uwave off
  6172. \noun off
  6173. \color none
  6174. 0.689
  6175. \end_layout
  6176. \end_inset
  6177. </cell>
  6178. </row>
  6179. </lyxtabular>
  6180. \end_inset
  6181. \end_layout
  6182. \begin_layout Plain Layout
  6183. \begin_inset Caption Standard
  6184. \begin_layout Plain Layout
  6185. \begin_inset CommandInset label
  6186. LatexCommand label
  6187. name "tab:AUC-PAM"
  6188. \end_inset
  6189. \series bold
  6190. ROC curve AUC values for internal and external validation with 6 different
  6191. normalization strategies.
  6192. \series default
  6193. These AUC values correspond to the ROC curves in Figure
  6194. \begin_inset CommandInset ref
  6195. LatexCommand ref
  6196. reference "fig:ROC-PAM-main"
  6197. plural "false"
  6198. caps "false"
  6199. noprefix "false"
  6200. \end_inset
  6201. .
  6202. \end_layout
  6203. \end_inset
  6204. \end_layout
  6205. \end_inset
  6206. \end_layout
  6207. \begin_layout Standard
  6208. For internal validation, the 6 methods' AUC values ranged from 0.816 to 0.891,
  6209. as shown in Table
  6210. \begin_inset CommandInset ref
  6211. LatexCommand ref
  6212. reference "tab:AUC-PAM"
  6213. plural "false"
  6214. caps "false"
  6215. noprefix "false"
  6216. \end_inset
  6217. .
  6218. Among the non-single-channel normalizations, dChip outperformed RMA, while
  6219. GRSN reduced the AUC values for both dChip and RMA.
  6220. Both single-channel methods, fRMA and SCAN, slightly outperformed RMA,
  6221. with fRMA ahead of SCAN.
  6222. However, the difference between RMA and fRMA is still quite small.
  6223. Figure
  6224. \begin_inset CommandInset ref
  6225. LatexCommand ref
  6226. reference "fig:ROC-PAM-int"
  6227. plural "false"
  6228. caps "false"
  6229. noprefix "false"
  6230. \end_inset
  6231. shows that the ROC curves for RMA, dChip, and fRMA look very similar and
  6232. relatively smooth, while both GRSN curves and the curve for SCAN have a
  6233. more jagged appearance.
  6234. \end_layout
  6235. \begin_layout Standard
  6236. For external validation, as expected, all the AUC values are lower than
  6237. the internal validations, ranging from 0.642 to 0.750 (Table
  6238. \begin_inset CommandInset ref
  6239. LatexCommand ref
  6240. reference "tab:AUC-PAM"
  6241. plural "false"
  6242. caps "false"
  6243. noprefix "false"
  6244. \end_inset
  6245. ).
  6246. With or without GRSN, RMA shows its dominance over dChip in this more challengi
  6247. ng test.
  6248. Unlike in the internal validation, GRSN actually improves the classifier
  6249. performance for RMA, although it does not for dChip.
  6250. Once again, both single-channel methods perform about on par with RMA,
  6251. with fRMA performing slightly better and SCAN performing a bit worse.
  6252. Figure
  6253. \begin_inset CommandInset ref
  6254. LatexCommand ref
  6255. reference "fig:ROC-PAM-ext"
  6256. plural "false"
  6257. caps "false"
  6258. noprefix "false"
  6259. \end_inset
  6260. shows the ROC curves for the external validation test.
  6261. As expected, none of them are as clean-looking as the internal validation
  6262. ROC curves.
  6263. The curves for RMA, RMA+GRSN, and fRMA all look similar, while the other
  6264. curves look more divergent.
  6265. \end_layout
  6266. \begin_layout Subsection
  6267. fRMA with custom-generated vectors enables single-channel normalization
  6268. on hthgu133pluspm platform
  6269. \end_layout
  6270. \begin_layout Standard
  6271. \begin_inset Float figure
  6272. wide false
  6273. sideways false
  6274. status open
  6275. \begin_layout Plain Layout
  6276. \align center
  6277. \begin_inset Float figure
  6278. placement tb
  6279. wide false
  6280. sideways false
  6281. status collapsed
  6282. \begin_layout Plain Layout
  6283. \align center
  6284. \begin_inset Graphics
  6285. filename graphics/frma-pax-bx/batchsize_batches.pdf
  6286. lyxscale 50
  6287. height 35theight%
  6288. groupId frmatools-subfig
  6289. \end_inset
  6290. \end_layout
  6291. \begin_layout Plain Layout
  6292. \begin_inset Caption Standard
  6293. \begin_layout Plain Layout
  6294. \begin_inset CommandInset label
  6295. LatexCommand label
  6296. name "fig:batch-size-batches"
  6297. \end_inset
  6298. \series bold
  6299. Number of batches usable in fRMA probe weight learning as a function of
  6300. batch size.
  6301. \end_layout
  6302. \end_inset
  6303. \end_layout
  6304. \end_inset
  6305. \end_layout
  6306. \begin_layout Plain Layout
  6307. \align center
  6308. \begin_inset Float figure
  6309. placement tb
  6310. wide false
  6311. sideways false
  6312. status collapsed
  6313. \begin_layout Plain Layout
  6314. \align center
  6315. \begin_inset Graphics
  6316. filename graphics/frma-pax-bx/batchsize_samples.pdf
  6317. lyxscale 50
  6318. height 35theight%
  6319. groupId frmatools-subfig
  6320. \end_inset
  6321. \end_layout
  6322. \begin_layout Plain Layout
  6323. \begin_inset Caption Standard
  6324. \begin_layout Plain Layout
  6325. \begin_inset CommandInset label
  6326. LatexCommand label
  6327. name "fig:batch-size-samples"
  6328. \end_inset
  6329. \series bold
  6330. Number of samples usable in fRMA probe weight learning as a function of
  6331. batch size.
  6332. \end_layout
  6333. \end_inset
  6334. \end_layout
  6335. \end_inset
  6336. \end_layout
  6337. \begin_layout Plain Layout
  6338. \begin_inset Caption Standard
  6339. \begin_layout Plain Layout
  6340. \series bold
  6341. \begin_inset CommandInset label
  6342. LatexCommand label
  6343. name "fig:frmatools-batch-size"
  6344. \end_inset
  6345. Effect of batch size selection on number of batches and number of samples
  6346. included in fRMA probe weight learning.
  6347. \series default
  6348. For batch sizes ranging from 3 to 15, the number of batches (a) and samples
  6349. (b) included in probe weight training were plotted for biopsy (BX) and
  6350. blood (PAX) samples.
  6351. The selected batch size, 5, is marked with a dotted vertical line.
  6352. \end_layout
  6353. \end_inset
  6354. \end_layout
  6355. \end_inset
  6356. \end_layout
  6357. \begin_layout Standard
  6358. In order to enable use of fRMA to normalize hthgu133pluspm, a custom set
  6359. of fRMA vectors was created.
  6360. First, an appropriate batch size was chosen by looking at the number of
  6361. batches and number of samples included as a function of batch size (Figure
  6362. \begin_inset CommandInset ref
  6363. LatexCommand ref
  6364. reference "fig:frmatools-batch-size"
  6365. plural "false"
  6366. caps "false"
  6367. noprefix "false"
  6368. \end_inset
  6369. ).
  6370. For a given batch size, all batches with fewer samples that the chosen
  6371. size must be ignored during training, while larger batches must be randomly
  6372. downsampled to the chosen size.
  6373. Hence, the number of samples included for a given batch size equals the
  6374. batch size times the number of batches with at least that many samples.
  6375. From Figure
  6376. \begin_inset CommandInset ref
  6377. LatexCommand ref
  6378. reference "fig:batch-size-samples"
  6379. plural "false"
  6380. caps "false"
  6381. noprefix "false"
  6382. \end_inset
  6383. , it is apparent that that a batch size of 8 maximizes the number of samples
  6384. included in training.
  6385. Increasing the batch size beyond this causes too many smaller batches to
  6386. be excluded, reducing the total number of samples for both tissue types.
  6387. However, a batch size of 8 is not necessarily optimal.
  6388. The article introducing frmaTools concluded that it was highly advantageous
  6389. to use a smaller batch size in order to include more batches, even at the
  6390. expense of including fewer total samples in training
  6391. \begin_inset CommandInset citation
  6392. LatexCommand cite
  6393. key "McCall2011"
  6394. literal "false"
  6395. \end_inset
  6396. .
  6397. To strike an appropriate balance between more batches and more samples,
  6398. a batch size of 5 was chosen.
  6399. For both blood and biopsy samples, this increased the number of batches
  6400. included by 10, with only a modest reduction in the number of samples compared
  6401. to a batch size of 8.
  6402. With a batch size of 5, 26 batches of biopsy samples and 46 batches of
  6403. blood samples were available.
  6404. \end_layout
  6405. \begin_layout Standard
  6406. \begin_inset Float figure
  6407. wide false
  6408. sideways false
  6409. status collapsed
  6410. \begin_layout Plain Layout
  6411. \begin_inset Float figure
  6412. wide false
  6413. sideways false
  6414. status open
  6415. \begin_layout Plain Layout
  6416. \align center
  6417. \begin_inset Graphics
  6418. filename graphics/frma-pax-bx/M-BX-violin.pdf
  6419. lyxscale 40
  6420. width 45col%
  6421. groupId m-violin
  6422. \end_inset
  6423. \end_layout
  6424. \begin_layout Plain Layout
  6425. \begin_inset Caption Standard
  6426. \begin_layout Plain Layout
  6427. \begin_inset CommandInset label
  6428. LatexCommand label
  6429. name "fig:m-bx-violin"
  6430. \end_inset
  6431. \series bold
  6432. Violin plot of inter-normalization log ratios for biopsy samples.
  6433. \end_layout
  6434. \end_inset
  6435. \end_layout
  6436. \end_inset
  6437. \begin_inset space \hfill{}
  6438. \end_inset
  6439. \begin_inset Float figure
  6440. wide false
  6441. sideways false
  6442. status collapsed
  6443. \begin_layout Plain Layout
  6444. \align center
  6445. \begin_inset Graphics
  6446. filename graphics/frma-pax-bx/M-PAX-violin.pdf
  6447. lyxscale 40
  6448. width 45col%
  6449. groupId m-violin
  6450. \end_inset
  6451. \end_layout
  6452. \begin_layout Plain Layout
  6453. \begin_inset Caption Standard
  6454. \begin_layout Plain Layout
  6455. \begin_inset CommandInset label
  6456. LatexCommand label
  6457. name "fig:m-pax-violin"
  6458. \end_inset
  6459. \series bold
  6460. Violin plot of inter-normalization log ratios for blood samples.
  6461. \end_layout
  6462. \end_inset
  6463. \end_layout
  6464. \end_inset
  6465. \end_layout
  6466. \begin_layout Plain Layout
  6467. \begin_inset Caption Standard
  6468. \begin_layout Plain Layout
  6469. \begin_inset CommandInset label
  6470. LatexCommand label
  6471. name "fig:frma-violin"
  6472. \end_inset
  6473. \series bold
  6474. Violin plot of log ratios between normalizations for 20 biopsy samples.
  6475. \series default
  6476. Each of 20 randomly selected samples was normalized with RMA and with 5
  6477. different sets of fRMA vectors.
  6478. The distribution of log ratios between normalized expression values, aggregated
  6479. across all 20 arrays, was plotted for each pair of normalizations.
  6480. \end_layout
  6481. \end_inset
  6482. \end_layout
  6483. \end_inset
  6484. \end_layout
  6485. \begin_layout Standard
  6486. Since fRMA training requires equal-size batches, larger batches are downsampled
  6487. randomly.
  6488. This introduces a nondeterministic step in the generation of normalization
  6489. vectors.
  6490. To show that this randomness does not substantially change the outcome,
  6491. the random downsampling and subsequent vector learning was repeated 5 times,
  6492. with a different random seed each time.
  6493. 20 samples were selected at random as a test set and normalized with each
  6494. of the 5 sets of fRMA normalization vectors as well as ordinary RMA, and
  6495. the normalized expression values were compared across normalizations.
  6496. Figure
  6497. \begin_inset CommandInset ref
  6498. LatexCommand ref
  6499. reference "fig:m-bx-violin"
  6500. plural "false"
  6501. caps "false"
  6502. noprefix "false"
  6503. \end_inset
  6504. shows a summary of these comparisons for biopsy samples.
  6505. Comparing RMA to each of the 5 fRMA normalizations, the distribution of
  6506. log ratios is somewhat wide, indicating that the normalizations disagree
  6507. on the expression values of a fair number of probe sets.
  6508. In contrast, comparisons of fRMA against fRMA, the vast mojority of probe
  6509. sets have very small log ratios, indicating a very high agreement between
  6510. the normalized values generated by the two normalizations.
  6511. This shows that the fRMA normalization's behavior is not very sensitive
  6512. to the random downsampling of larger batches during training.
  6513. \end_layout
  6514. \begin_layout Standard
  6515. \begin_inset Float figure
  6516. wide false
  6517. sideways false
  6518. status open
  6519. \begin_layout Plain Layout
  6520. \align center
  6521. \begin_inset Float figure
  6522. wide false
  6523. sideways false
  6524. status collapsed
  6525. \begin_layout Plain Layout
  6526. \align center
  6527. \begin_inset Graphics
  6528. filename graphics/frma-pax-bx/MA-BX-RMA.fRMA-RASTER.png
  6529. lyxscale 10
  6530. width 45col%
  6531. groupId ma-frma
  6532. \end_inset
  6533. \end_layout
  6534. \begin_layout Plain Layout
  6535. \begin_inset Caption Standard
  6536. \begin_layout Plain Layout
  6537. \begin_inset CommandInset label
  6538. LatexCommand label
  6539. name "fig:ma-bx-rma-frma"
  6540. \end_inset
  6541. RMA vs.
  6542. fRMA for biopsy samples.
  6543. \end_layout
  6544. \end_inset
  6545. \end_layout
  6546. \end_inset
  6547. \begin_inset space \hfill{}
  6548. \end_inset
  6549. \begin_inset Float figure
  6550. wide false
  6551. sideways false
  6552. status collapsed
  6553. \begin_layout Plain Layout
  6554. \align center
  6555. \begin_inset Graphics
  6556. filename graphics/frma-pax-bx/MA-BX-fRMA.fRMA-RASTER.png
  6557. lyxscale 10
  6558. width 45col%
  6559. groupId ma-frma
  6560. \end_inset
  6561. \end_layout
  6562. \begin_layout Plain Layout
  6563. \begin_inset Caption Standard
  6564. \begin_layout Plain Layout
  6565. \begin_inset CommandInset label
  6566. LatexCommand label
  6567. name "fig:ma-bx-frma-frma"
  6568. \end_inset
  6569. fRMA vs fRMA for biopsy samples.
  6570. \end_layout
  6571. \end_inset
  6572. \end_layout
  6573. \end_inset
  6574. \end_layout
  6575. \begin_layout Plain Layout
  6576. \align center
  6577. \begin_inset Float figure
  6578. wide false
  6579. sideways false
  6580. status collapsed
  6581. \begin_layout Plain Layout
  6582. \align center
  6583. \begin_inset Graphics
  6584. filename graphics/frma-pax-bx/MA-PAX-RMA.fRMA-RASTER.png
  6585. lyxscale 10
  6586. width 45col%
  6587. groupId ma-frma
  6588. \end_inset
  6589. \end_layout
  6590. \begin_layout Plain Layout
  6591. \begin_inset Caption Standard
  6592. \begin_layout Plain Layout
  6593. \begin_inset CommandInset label
  6594. LatexCommand label
  6595. name "fig:MA-PAX-rma-frma"
  6596. \end_inset
  6597. RMA vs.
  6598. fRMA for blood samples.
  6599. \end_layout
  6600. \end_inset
  6601. \end_layout
  6602. \end_inset
  6603. \begin_inset space \hfill{}
  6604. \end_inset
  6605. \begin_inset Float figure
  6606. wide false
  6607. sideways false
  6608. status collapsed
  6609. \begin_layout Plain Layout
  6610. \align center
  6611. \begin_inset Graphics
  6612. filename graphics/frma-pax-bx/MA-PAX-fRMA.fRMA-RASTER.png
  6613. lyxscale 10
  6614. width 45col%
  6615. groupId ma-frma
  6616. \end_inset
  6617. \end_layout
  6618. \begin_layout Plain Layout
  6619. \begin_inset Caption Standard
  6620. \begin_layout Plain Layout
  6621. \begin_inset CommandInset label
  6622. LatexCommand label
  6623. name "fig:MA-PAX-frma-frma"
  6624. \end_inset
  6625. fRMA vs fRMA for blood samples.
  6626. \end_layout
  6627. \end_inset
  6628. \end_layout
  6629. \end_inset
  6630. \end_layout
  6631. \begin_layout Plain Layout
  6632. \begin_inset Caption Standard
  6633. \begin_layout Plain Layout
  6634. \series bold
  6635. \begin_inset CommandInset label
  6636. LatexCommand label
  6637. name "fig:Representative-MA-plots"
  6638. \end_inset
  6639. Representative MA plots comparing RMA and custom fRMA normalizations.
  6640. \series default
  6641. For each plot, 20 samples were normalized using 2 different normalizations,
  6642. and then averages (A) and log ratios (M) were plotted between the two different
  6643. normalizations for every probe.
  6644. For the
  6645. \begin_inset Quotes eld
  6646. \end_inset
  6647. fRMA vs fRMA
  6648. \begin_inset Quotes erd
  6649. \end_inset
  6650. plots (b & d), two different fRMA normalizations using vectors from two
  6651. independent batch samplings were compared.
  6652. Density of points is represented by blue shading, and individual outlier
  6653. points are plotted.
  6654. \end_layout
  6655. \end_inset
  6656. \end_layout
  6657. \end_inset
  6658. \end_layout
  6659. \begin_layout Standard
  6660. Figure
  6661. \begin_inset CommandInset ref
  6662. LatexCommand ref
  6663. reference "fig:ma-bx-rma-frma"
  6664. plural "false"
  6665. caps "false"
  6666. noprefix "false"
  6667. \end_inset
  6668. shows an MA plot of the RMA-normalized values against the fRMA-normalized
  6669. values for the same probe sets and arrays, corresponding to the first row
  6670. of Figure
  6671. \begin_inset CommandInset ref
  6672. LatexCommand ref
  6673. reference "fig:m-bx-violin"
  6674. plural "false"
  6675. caps "false"
  6676. noprefix "false"
  6677. \end_inset
  6678. .
  6679. This MA plot shows that not only is there a wide distribution of M-values,
  6680. but the trend of M-values is dependent on the average normalized intensity.
  6681. This is expected, since the overall trend represents the differences in
  6682. the quantile normalization step.
  6683. When running RMA, only the quantiles for these specific 20 arrays are used,
  6684. while for fRMA the quantile distribution is taking from all arrays used
  6685. in training.
  6686. Figure
  6687. \begin_inset CommandInset ref
  6688. LatexCommand ref
  6689. reference "fig:ma-bx-frma-frma"
  6690. plural "false"
  6691. caps "false"
  6692. noprefix "false"
  6693. \end_inset
  6694. shows a similar MA plot comparing 2 different fRMA normalizations, correspondin
  6695. g to the 6th row of Figure
  6696. \begin_inset CommandInset ref
  6697. LatexCommand ref
  6698. reference "fig:m-bx-violin"
  6699. plural "false"
  6700. caps "false"
  6701. noprefix "false"
  6702. \end_inset
  6703. .
  6704. The MA plot is very tightly centered around zero with no visible trend.
  6705. Figures
  6706. \begin_inset CommandInset ref
  6707. LatexCommand ref
  6708. reference "fig:m-pax-violin"
  6709. plural "false"
  6710. caps "false"
  6711. noprefix "false"
  6712. \end_inset
  6713. ,
  6714. \begin_inset CommandInset ref
  6715. LatexCommand ref
  6716. reference "fig:MA-PAX-rma-frma"
  6717. plural "false"
  6718. caps "false"
  6719. noprefix "false"
  6720. \end_inset
  6721. , and
  6722. \begin_inset CommandInset ref
  6723. LatexCommand ref
  6724. reference "fig:ma-bx-frma-frma"
  6725. plural "false"
  6726. caps "false"
  6727. noprefix "false"
  6728. \end_inset
  6729. show exactly the same information for the blood samples, once again comparing
  6730. the normalized expression values between normalizations for all probe sets
  6731. across 20 randomly selected test arrays.
  6732. Once again, there is a wider distribution of log ratios between RMA-normalized
  6733. values and fRMA-normalized, and a much tighter distribution when comparing
  6734. different fRMA normalizations to each other, indicating that the fRMA training
  6735. process is robust to random batch downsampling for the blood samples as
  6736. well.
  6737. \end_layout
  6738. \begin_layout Subsection
  6739. SVA, voom, and array weights improve model fit for methylation array data
  6740. \end_layout
  6741. \begin_layout Standard
  6742. \begin_inset ERT
  6743. status open
  6744. \begin_layout Plain Layout
  6745. \backslash
  6746. afterpage{
  6747. \end_layout
  6748. \begin_layout Plain Layout
  6749. \backslash
  6750. begin{landscape}
  6751. \end_layout
  6752. \end_inset
  6753. \end_layout
  6754. \begin_layout Standard
  6755. \begin_inset Float figure
  6756. wide false
  6757. sideways false
  6758. status open
  6759. \begin_layout Plain Layout
  6760. \begin_inset Flex TODO Note (inline)
  6761. status open
  6762. \begin_layout Plain Layout
  6763. Fix axis labels:
  6764. \begin_inset Quotes eld
  6765. \end_inset
  6766. log2 M-value
  6767. \begin_inset Quotes erd
  6768. \end_inset
  6769. is redundant because M-values are already log scale
  6770. \end_layout
  6771. \end_inset
  6772. \end_layout
  6773. \begin_layout Plain Layout
  6774. \begin_inset Float figure
  6775. wide false
  6776. sideways false
  6777. status collapsed
  6778. \begin_layout Plain Layout
  6779. \align center
  6780. \begin_inset Graphics
  6781. filename graphics/methylvoom/unadj.dupcor/meanvar-trends-PAGE1-CROP-RASTER.png
  6782. lyxscale 15
  6783. width 30col%
  6784. groupId voomaw-subfig
  6785. \end_inset
  6786. \end_layout
  6787. \begin_layout Plain Layout
  6788. \begin_inset Caption Standard
  6789. \begin_layout Plain Layout
  6790. \begin_inset CommandInset label
  6791. LatexCommand label
  6792. name "fig:meanvar-basic"
  6793. \end_inset
  6794. Mean-variance trend for analysis A.
  6795. \end_layout
  6796. \end_inset
  6797. \end_layout
  6798. \end_inset
  6799. \begin_inset space \hfill{}
  6800. \end_inset
  6801. \begin_inset Float figure
  6802. wide false
  6803. sideways false
  6804. status collapsed
  6805. \begin_layout Plain Layout
  6806. \align center
  6807. \begin_inset Graphics
  6808. filename graphics/methylvoom/unadj.dupcor.sva.aw/meanvar-trends-PAGE1-CROP-RASTER.png
  6809. lyxscale 15
  6810. width 30col%
  6811. groupId voomaw-subfig
  6812. \end_inset
  6813. \end_layout
  6814. \begin_layout Plain Layout
  6815. \begin_inset Caption Standard
  6816. \begin_layout Plain Layout
  6817. \begin_inset CommandInset label
  6818. LatexCommand label
  6819. name "fig:meanvar-sva-aw"
  6820. \end_inset
  6821. Mean-variance trend for analysis B.
  6822. \end_layout
  6823. \end_inset
  6824. \end_layout
  6825. \end_inset
  6826. \begin_inset space \hfill{}
  6827. \end_inset
  6828. \begin_inset Float figure
  6829. wide false
  6830. sideways false
  6831. status collapsed
  6832. \begin_layout Plain Layout
  6833. \align center
  6834. \begin_inset Graphics
  6835. filename graphics/methylvoom/unadj.dupcor.sva.voomaw/meanvar-trends-PAGE2-CROP-RASTER.png
  6836. lyxscale 15
  6837. width 30col%
  6838. groupId voomaw-subfig
  6839. \end_inset
  6840. \end_layout
  6841. \begin_layout Plain Layout
  6842. \begin_inset Caption Standard
  6843. \begin_layout Plain Layout
  6844. \begin_inset CommandInset label
  6845. LatexCommand label
  6846. name "fig:meanvar-sva-voomaw"
  6847. \end_inset
  6848. Mean-variance trend after voom modeling in analysis C.
  6849. \end_layout
  6850. \end_inset
  6851. \end_layout
  6852. \end_inset
  6853. \end_layout
  6854. \begin_layout Plain Layout
  6855. \begin_inset Caption Standard
  6856. \begin_layout Plain Layout
  6857. \series bold
  6858. Mean-variance trend modeling in methylation array data.
  6859. \series default
  6860. The estimated log2(standard deviation) for each probe is plotted against
  6861. the probe's average M-value across all samples as a black point, with some
  6862. transparency to make overplotting more visible, since there are about 450,000
  6863. points.
  6864. Density of points is also indicated by the dark blue contour lines.
  6865. The prior variance trend estimated by eBayes is shown in light blue, while
  6866. the lowess trend of the points is shown in red.
  6867. \end_layout
  6868. \end_inset
  6869. \end_layout
  6870. \end_inset
  6871. \end_layout
  6872. \begin_layout Standard
  6873. \begin_inset ERT
  6874. status open
  6875. \begin_layout Plain Layout
  6876. \backslash
  6877. end{landscape}
  6878. \end_layout
  6879. \begin_layout Plain Layout
  6880. }
  6881. \end_layout
  6882. \end_inset
  6883. \end_layout
  6884. \begin_layout Standard
  6885. Figure
  6886. \begin_inset CommandInset ref
  6887. LatexCommand ref
  6888. reference "fig:meanvar-basic"
  6889. plural "false"
  6890. caps "false"
  6891. noprefix "false"
  6892. \end_inset
  6893. shows the relationship between the mean M-value and the standard deviation
  6894. calculated for each probe in the methylation array data set.
  6895. A few features of the data are apparent.
  6896. First, the data are very strongly bimodal, with peaks in the density around
  6897. M-values of +4 and -4.
  6898. These modes correspond to methylation sites that are nearly 100% methylated
  6899. and nearly 100% unmethylated, respectively.
  6900. The strong bomodality indicates that a majority of probes interrogate sites
  6901. that fall into one of these two categories.
  6902. The points in between these modes represent sites that are either partially
  6903. methylated in many samples, or are fully methylated in some samples and
  6904. fully unmethylated in other samples, or some combination.
  6905. The next visible feature of the data is the W-shaped variance trend.
  6906. The upticks in the variance trend on either side are expected, based on
  6907. the sigmoid transformation exaggerating small differences at extreme M-values
  6908. (Figure
  6909. \begin_inset CommandInset ref
  6910. LatexCommand ref
  6911. reference "fig:Sigmoid-beta-m-mapping"
  6912. plural "false"
  6913. caps "false"
  6914. noprefix "false"
  6915. \end_inset
  6916. ).
  6917. However, the uptick in the center is interesting: it indicates that sites
  6918. that are not constitutitively methylated or unmethylated have a higher
  6919. variance.
  6920. This could be a genuine biological effect, or it could be spurious noise
  6921. that is only observable at sites with varying methylation.
  6922. \end_layout
  6923. \begin_layout Standard
  6924. In Figure
  6925. \begin_inset CommandInset ref
  6926. LatexCommand ref
  6927. reference "fig:meanvar-sva-aw"
  6928. plural "false"
  6929. caps "false"
  6930. noprefix "false"
  6931. \end_inset
  6932. , we see the mean-variance trend for the same methylation array data, this
  6933. time with surrogate variables and sample quality weights estimated from
  6934. the data and included in the model.
  6935. As expected, the overall average variance is smaller, since the surrogate
  6936. variables account for some of the variance.
  6937. In addition, the uptick in variance in the middle of the M-value range
  6938. has disappeared, turning the W shape into a wide U shape.
  6939. This indicates that the excess variance in the probes with intermediate
  6940. M-values was explained by systematic variations not correlated with known
  6941. covariates, and these variations were modeled by the surrogate variables.
  6942. The result is a nearly flat variance trend for the entire intermediate
  6943. M-value range from about -3 to +3.
  6944. Note that this corresponds closely to the range within which the M-value
  6945. transformation shown in Figure
  6946. \begin_inset CommandInset ref
  6947. LatexCommand ref
  6948. reference "fig:Sigmoid-beta-m-mapping"
  6949. plural "false"
  6950. caps "false"
  6951. noprefix "false"
  6952. \end_inset
  6953. is nearly linear.
  6954. In contrast, the excess variance at the extremes (greater than +3 and less
  6955. than -3) was not
  6956. \begin_inset Quotes eld
  6957. \end_inset
  6958. absorbed
  6959. \begin_inset Quotes erd
  6960. \end_inset
  6961. by the surrogate variables and remains in the plot, indicating that this
  6962. variation has no systematic component: probes with extreme M-values are
  6963. uniformly more variable across all samples, as expected.
  6964. \end_layout
  6965. \begin_layout Standard
  6966. Figure
  6967. \begin_inset CommandInset ref
  6968. LatexCommand ref
  6969. reference "fig:meanvar-sva-voomaw"
  6970. plural "false"
  6971. caps "false"
  6972. noprefix "false"
  6973. \end_inset
  6974. shows the mean-variance trend after fitting the model with the observation
  6975. weights assigned by voom based on the mean-variance trend shown in Figure
  6976. \begin_inset CommandInset ref
  6977. LatexCommand ref
  6978. reference "fig:meanvar-sva-aw"
  6979. plural "false"
  6980. caps "false"
  6981. noprefix "false"
  6982. \end_inset
  6983. .
  6984. As expected, the weights exactly counteract the trend in the data, resulting
  6985. in a nearly flat trend centered vertically at 1 (i.e.
  6986. 0 on the log scale).
  6987. This shows that the observations with extreme M-values have been appropriately
  6988. down-weighted to account for the fact that the noise in those observations
  6989. has been amplified by the non-linear M-value transformation.
  6990. In turn, this gives relatively more weight to observervations in the middle
  6991. region, which are more likely to correspond to probes measuring interesting
  6992. biology (not constitutively methylated or unmethylated).
  6993. \end_layout
  6994. \begin_layout Standard
  6995. \begin_inset Float table
  6996. wide false
  6997. sideways false
  6998. status open
  6999. \begin_layout Plain Layout
  7000. \align center
  7001. \begin_inset Tabular
  7002. <lyxtabular version="3" rows="5" columns="3">
  7003. <features tabularvalignment="middle">
  7004. <column alignment="center" valignment="top">
  7005. <column alignment="center" valignment="top">
  7006. <column alignment="center" valignment="top">
  7007. <row>
  7008. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  7009. \begin_inset Text
  7010. \begin_layout Plain Layout
  7011. Covariate
  7012. \end_layout
  7013. \end_inset
  7014. </cell>
  7015. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  7016. \begin_inset Text
  7017. \begin_layout Plain Layout
  7018. Test used
  7019. \end_layout
  7020. \end_inset
  7021. </cell>
  7022. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  7023. \begin_inset Text
  7024. \begin_layout Plain Layout
  7025. p-value
  7026. \end_layout
  7027. \end_inset
  7028. </cell>
  7029. </row>
  7030. <row>
  7031. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  7032. \begin_inset Text
  7033. \begin_layout Plain Layout
  7034. Transplant Status
  7035. \end_layout
  7036. \end_inset
  7037. </cell>
  7038. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  7039. \begin_inset Text
  7040. \begin_layout Plain Layout
  7041. F-test
  7042. \end_layout
  7043. \end_inset
  7044. </cell>
  7045. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  7046. \begin_inset Text
  7047. \begin_layout Plain Layout
  7048. 0.404
  7049. \end_layout
  7050. \end_inset
  7051. </cell>
  7052. </row>
  7053. <row>
  7054. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  7055. \begin_inset Text
  7056. \begin_layout Plain Layout
  7057. Diabetes Diagnosis
  7058. \end_layout
  7059. \end_inset
  7060. </cell>
  7061. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  7062. \begin_inset Text
  7063. \begin_layout Plain Layout
  7064. \emph on
  7065. t
  7066. \emph default
  7067. -test
  7068. \end_layout
  7069. \end_inset
  7070. </cell>
  7071. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  7072. \begin_inset Text
  7073. \begin_layout Plain Layout
  7074. 0.00106
  7075. \end_layout
  7076. \end_inset
  7077. </cell>
  7078. </row>
  7079. <row>
  7080. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  7081. \begin_inset Text
  7082. \begin_layout Plain Layout
  7083. Sex
  7084. \end_layout
  7085. \end_inset
  7086. </cell>
  7087. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  7088. \begin_inset Text
  7089. \begin_layout Plain Layout
  7090. \emph on
  7091. t
  7092. \emph default
  7093. -test
  7094. \end_layout
  7095. \end_inset
  7096. </cell>
  7097. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  7098. \begin_inset Text
  7099. \begin_layout Plain Layout
  7100. 0.148
  7101. \end_layout
  7102. \end_inset
  7103. </cell>
  7104. </row>
  7105. <row>
  7106. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  7107. \begin_inset Text
  7108. \begin_layout Plain Layout
  7109. Age
  7110. \end_layout
  7111. \end_inset
  7112. </cell>
  7113. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  7114. \begin_inset Text
  7115. \begin_layout Plain Layout
  7116. linear regression
  7117. \end_layout
  7118. \end_inset
  7119. </cell>
  7120. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  7121. \begin_inset Text
  7122. \begin_layout Plain Layout
  7123. 0.212
  7124. \end_layout
  7125. \end_inset
  7126. </cell>
  7127. </row>
  7128. </lyxtabular>
  7129. \end_inset
  7130. \end_layout
  7131. \begin_layout Plain Layout
  7132. \begin_inset Caption Standard
  7133. \begin_layout Plain Layout
  7134. \series bold
  7135. \begin_inset CommandInset label
  7136. LatexCommand label
  7137. name "tab:weight-covariate-tests"
  7138. \end_inset
  7139. Association of sample weights with clinical covariates in methylation array
  7140. data.
  7141. \series default
  7142. Computed sample quality log weights were tested for significant association
  7143. with each of the variables in the model (1st column).
  7144. An appropriate test was selected for each variable based on whether the
  7145. variable had 2 categories (
  7146. \emph on
  7147. t
  7148. \emph default
  7149. -test), had more than 2 categories (F-test), or was numeric (linear regression).
  7150. The test selected is shown in the 2nd column.
  7151. P-values for association with the log weights are shown in the 3rd column.
  7152. No multiple testing adjustment was performed for these p-values.
  7153. \end_layout
  7154. \end_inset
  7155. \end_layout
  7156. \end_inset
  7157. \end_layout
  7158. \begin_layout Standard
  7159. \begin_inset Float figure
  7160. wide false
  7161. sideways false
  7162. status open
  7163. \begin_layout Plain Layout
  7164. \begin_inset Flex TODO Note (inline)
  7165. status open
  7166. \begin_layout Plain Layout
  7167. Redo the sample weight boxplot with notches, and remove fill colors
  7168. \end_layout
  7169. \end_inset
  7170. \end_layout
  7171. \begin_layout Plain Layout
  7172. \align center
  7173. \begin_inset Graphics
  7174. filename graphics/methylvoom/unadj.dupcor.sva.voomaw/sample-weights-PAGE3-CROP.pdf
  7175. lyxscale 50
  7176. width 60col%
  7177. groupId colwidth
  7178. \end_inset
  7179. \end_layout
  7180. \begin_layout Plain Layout
  7181. \begin_inset Caption Standard
  7182. \begin_layout Plain Layout
  7183. \begin_inset CommandInset label
  7184. LatexCommand label
  7185. name "fig:diabetes-sample-weights"
  7186. \end_inset
  7187. \series bold
  7188. Box-and-whiskers plot of sample quality weights grouped by diabetes diagnosis.
  7189. \series default
  7190. Samples were grouped based on diabetes diagnosis, and the distribution of
  7191. sample quality weights for each diagnosis was plotted as a box-and-whiskers
  7192. plot
  7193. \begin_inset CommandInset citation
  7194. LatexCommand cite
  7195. key "McGill1978"
  7196. literal "false"
  7197. \end_inset
  7198. .
  7199. \end_layout
  7200. \end_inset
  7201. \end_layout
  7202. \begin_layout Plain Layout
  7203. \end_layout
  7204. \end_inset
  7205. \end_layout
  7206. \begin_layout Standard
  7207. To determine whether any of the known experimental factors had an impact
  7208. on data quality, the sample quality weights estimated from the data were
  7209. tested for association with each of the experimental factors (Table
  7210. \begin_inset CommandInset ref
  7211. LatexCommand ref
  7212. reference "tab:weight-covariate-tests"
  7213. plural "false"
  7214. caps "false"
  7215. noprefix "false"
  7216. \end_inset
  7217. ).
  7218. Diabetes diagnosis was found to have a potentially significant association
  7219. with the sample weights, with a t-test p-value of
  7220. \begin_inset Formula $1.06\times10^{-3}$
  7221. \end_inset
  7222. .
  7223. Figure
  7224. \begin_inset CommandInset ref
  7225. LatexCommand ref
  7226. reference "fig:diabetes-sample-weights"
  7227. plural "false"
  7228. caps "false"
  7229. noprefix "false"
  7230. \end_inset
  7231. shows the distribution of sample weights grouped by diabetes diagnosis.
  7232. The samples from patients with Type 2 diabetes were assigned significantly
  7233. lower weights than those from patients with Type 1 diabetes.
  7234. This indicates that the type 2 diabetes samples had an overall higher variance
  7235. on average across all probes.
  7236. \end_layout
  7237. \begin_layout Standard
  7238. \begin_inset Float table
  7239. wide false
  7240. sideways false
  7241. status open
  7242. \begin_layout Plain Layout
  7243. \align center
  7244. \begin_inset Flex TODO Note (inline)
  7245. status open
  7246. \begin_layout Plain Layout
  7247. Consider transposing these tables
  7248. \end_layout
  7249. \end_inset
  7250. \end_layout
  7251. \begin_layout Plain Layout
  7252. \begin_inset Float table
  7253. wide false
  7254. sideways false
  7255. status open
  7256. \begin_layout Plain Layout
  7257. \align center
  7258. \begin_inset Tabular
  7259. <lyxtabular version="3" rows="5" columns="4">
  7260. <features tabularvalignment="middle">
  7261. <column alignment="center" valignment="top">
  7262. <column alignment="center" valignment="top">
  7263. <column alignment="center" valignment="top">
  7264. <column alignment="center" valignment="top">
  7265. <row>
  7266. <cell alignment="center" valignment="top" usebox="none">
  7267. \begin_inset Text
  7268. \begin_layout Plain Layout
  7269. \end_layout
  7270. \end_inset
  7271. </cell>
  7272. <cell multicolumn="1" alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  7273. \begin_inset Text
  7274. \begin_layout Plain Layout
  7275. Analysis
  7276. \end_layout
  7277. \end_inset
  7278. </cell>
  7279. <cell multicolumn="2" alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  7280. \begin_inset Text
  7281. \begin_layout Plain Layout
  7282. \end_layout
  7283. \end_inset
  7284. </cell>
  7285. <cell multicolumn="2" alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  7286. \begin_inset Text
  7287. \begin_layout Plain Layout
  7288. \end_layout
  7289. \end_inset
  7290. </cell>
  7291. </row>
  7292. <row>
  7293. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  7294. \begin_inset Text
  7295. \begin_layout Plain Layout
  7296. Contrast
  7297. \end_layout
  7298. \end_inset
  7299. </cell>
  7300. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  7301. \begin_inset Text
  7302. \begin_layout Plain Layout
  7303. A
  7304. \end_layout
  7305. \end_inset
  7306. </cell>
  7307. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  7308. \begin_inset Text
  7309. \begin_layout Plain Layout
  7310. B
  7311. \end_layout
  7312. \end_inset
  7313. </cell>
  7314. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  7315. \begin_inset Text
  7316. \begin_layout Plain Layout
  7317. C
  7318. \end_layout
  7319. \end_inset
  7320. </cell>
  7321. </row>
  7322. <row>
  7323. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  7324. \begin_inset Text
  7325. \begin_layout Plain Layout
  7326. TX vs AR
  7327. \end_layout
  7328. \end_inset
  7329. </cell>
  7330. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  7331. \begin_inset Text
  7332. \begin_layout Plain Layout
  7333. 0
  7334. \end_layout
  7335. \end_inset
  7336. </cell>
  7337. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  7338. \begin_inset Text
  7339. \begin_layout Plain Layout
  7340. 25
  7341. \end_layout
  7342. \end_inset
  7343. </cell>
  7344. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  7345. \begin_inset Text
  7346. \begin_layout Plain Layout
  7347. 22
  7348. \end_layout
  7349. \end_inset
  7350. </cell>
  7351. </row>
  7352. <row>
  7353. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  7354. \begin_inset Text
  7355. \begin_layout Plain Layout
  7356. TX vs ADNR
  7357. \end_layout
  7358. \end_inset
  7359. </cell>
  7360. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  7361. \begin_inset Text
  7362. \begin_layout Plain Layout
  7363. 7
  7364. \end_layout
  7365. \end_inset
  7366. </cell>
  7367. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  7368. \begin_inset Text
  7369. \begin_layout Plain Layout
  7370. 338
  7371. \end_layout
  7372. \end_inset
  7373. </cell>
  7374. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  7375. \begin_inset Text
  7376. \begin_layout Plain Layout
  7377. 369
  7378. \end_layout
  7379. \end_inset
  7380. </cell>
  7381. </row>
  7382. <row>
  7383. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  7384. \begin_inset Text
  7385. \begin_layout Plain Layout
  7386. TX vs CAN
  7387. \end_layout
  7388. \end_inset
  7389. </cell>
  7390. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  7391. \begin_inset Text
  7392. \begin_layout Plain Layout
  7393. 0
  7394. \end_layout
  7395. \end_inset
  7396. </cell>
  7397. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  7398. \begin_inset Text
  7399. \begin_layout Plain Layout
  7400. 231
  7401. \end_layout
  7402. \end_inset
  7403. </cell>
  7404. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  7405. \begin_inset Text
  7406. \begin_layout Plain Layout
  7407. 278
  7408. \end_layout
  7409. \end_inset
  7410. </cell>
  7411. </row>
  7412. </lyxtabular>
  7413. \end_inset
  7414. \end_layout
  7415. \begin_layout Plain Layout
  7416. \begin_inset Caption Standard
  7417. \begin_layout Plain Layout
  7418. \begin_inset CommandInset label
  7419. LatexCommand label
  7420. name "tab:methyl-num-signif"
  7421. \end_inset
  7422. Number of probes significant at 10% FDR.
  7423. \end_layout
  7424. \end_inset
  7425. \end_layout
  7426. \end_inset
  7427. \begin_inset space \hfill{}
  7428. \end_inset
  7429. \begin_inset Float table
  7430. wide false
  7431. sideways false
  7432. status open
  7433. \begin_layout Plain Layout
  7434. \align center
  7435. \begin_inset Tabular
  7436. <lyxtabular version="3" rows="5" columns="4">
  7437. <features tabularvalignment="middle">
  7438. <column alignment="center" valignment="top">
  7439. <column alignment="center" valignment="top">
  7440. <column alignment="center" valignment="top">
  7441. <column alignment="center" valignment="top">
  7442. <row>
  7443. <cell alignment="center" valignment="top" usebox="none">
  7444. \begin_inset Text
  7445. \begin_layout Plain Layout
  7446. \end_layout
  7447. \end_inset
  7448. </cell>
  7449. <cell multicolumn="1" alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  7450. \begin_inset Text
  7451. \begin_layout Plain Layout
  7452. Analysis
  7453. \end_layout
  7454. \end_inset
  7455. </cell>
  7456. <cell multicolumn="2" alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  7457. \begin_inset Text
  7458. \begin_layout Plain Layout
  7459. \end_layout
  7460. \end_inset
  7461. </cell>
  7462. <cell multicolumn="2" alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  7463. \begin_inset Text
  7464. \begin_layout Plain Layout
  7465. \end_layout
  7466. \end_inset
  7467. </cell>
  7468. </row>
  7469. <row>
  7470. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  7471. \begin_inset Text
  7472. \begin_layout Plain Layout
  7473. Contrast
  7474. \end_layout
  7475. \end_inset
  7476. </cell>
  7477. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  7478. \begin_inset Text
  7479. \begin_layout Plain Layout
  7480. A
  7481. \end_layout
  7482. \end_inset
  7483. </cell>
  7484. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  7485. \begin_inset Text
  7486. \begin_layout Plain Layout
  7487. B
  7488. \end_layout
  7489. \end_inset
  7490. </cell>
  7491. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  7492. \begin_inset Text
  7493. \begin_layout Plain Layout
  7494. C
  7495. \end_layout
  7496. \end_inset
  7497. </cell>
  7498. </row>
  7499. <row>
  7500. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  7501. \begin_inset Text
  7502. \begin_layout Plain Layout
  7503. TX vs AR
  7504. \end_layout
  7505. \end_inset
  7506. </cell>
  7507. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  7508. \begin_inset Text
  7509. \begin_layout Plain Layout
  7510. 0
  7511. \end_layout
  7512. \end_inset
  7513. </cell>
  7514. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  7515. \begin_inset Text
  7516. \begin_layout Plain Layout
  7517. 10,063
  7518. \end_layout
  7519. \end_inset
  7520. </cell>
  7521. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  7522. \begin_inset Text
  7523. \begin_layout Plain Layout
  7524. 11,225
  7525. \end_layout
  7526. \end_inset
  7527. </cell>
  7528. </row>
  7529. <row>
  7530. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  7531. \begin_inset Text
  7532. \begin_layout Plain Layout
  7533. TX vs ADNR
  7534. \end_layout
  7535. \end_inset
  7536. </cell>
  7537. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  7538. \begin_inset Text
  7539. \begin_layout Plain Layout
  7540. 27
  7541. \end_layout
  7542. \end_inset
  7543. </cell>
  7544. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  7545. \begin_inset Text
  7546. \begin_layout Plain Layout
  7547. 12,674
  7548. \end_layout
  7549. \end_inset
  7550. </cell>
  7551. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  7552. \begin_inset Text
  7553. \begin_layout Plain Layout
  7554. 13,086
  7555. \end_layout
  7556. \end_inset
  7557. </cell>
  7558. </row>
  7559. <row>
  7560. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  7561. \begin_inset Text
  7562. \begin_layout Plain Layout
  7563. TX vs CAN
  7564. \end_layout
  7565. \end_inset
  7566. </cell>
  7567. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  7568. \begin_inset Text
  7569. \begin_layout Plain Layout
  7570. 966
  7571. \end_layout
  7572. \end_inset
  7573. </cell>
  7574. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  7575. \begin_inset Text
  7576. \begin_layout Plain Layout
  7577. 20,039
  7578. \end_layout
  7579. \end_inset
  7580. </cell>
  7581. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  7582. \begin_inset Text
  7583. \begin_layout Plain Layout
  7584. 20,955
  7585. \end_layout
  7586. \end_inset
  7587. </cell>
  7588. </row>
  7589. </lyxtabular>
  7590. \end_inset
  7591. \end_layout
  7592. \begin_layout Plain Layout
  7593. \begin_inset Caption Standard
  7594. \begin_layout Plain Layout
  7595. \begin_inset CommandInset label
  7596. LatexCommand label
  7597. name "tab:methyl-est-nonnull"
  7598. \end_inset
  7599. Estimated number of non-null tests, using the method of averaging local
  7600. FDR values
  7601. \begin_inset CommandInset citation
  7602. LatexCommand cite
  7603. key "Phipson2013Thesis"
  7604. literal "false"
  7605. \end_inset
  7606. .
  7607. \end_layout
  7608. \end_inset
  7609. \end_layout
  7610. \end_inset
  7611. \end_layout
  7612. \begin_layout Plain Layout
  7613. \begin_inset Caption Standard
  7614. \begin_layout Plain Layout
  7615. \series bold
  7616. Estimates of degree of differential methylation in for each contrast in
  7617. each analysis.
  7618. \series default
  7619. For each of the analyses in Table
  7620. \begin_inset CommandInset ref
  7621. LatexCommand ref
  7622. reference "tab:Summary-of-meth-analysis"
  7623. plural "false"
  7624. caps "false"
  7625. noprefix "false"
  7626. \end_inset
  7627. , these tables show the number of probes called significantly differentially
  7628. methylated at a threshold of 10% FDR for each comparison between TX and
  7629. the other 3 transplant statuses (a) and the estimated total number of probes
  7630. that are differentially methylated (b).
  7631. \end_layout
  7632. \end_inset
  7633. \end_layout
  7634. \end_inset
  7635. \end_layout
  7636. \begin_layout Standard
  7637. \begin_inset Float figure
  7638. wide false
  7639. sideways false
  7640. status open
  7641. \begin_layout Plain Layout
  7642. \align center
  7643. \series bold
  7644. \begin_inset Float figure
  7645. wide false
  7646. sideways false
  7647. status collapsed
  7648. \begin_layout Plain Layout
  7649. \align center
  7650. \begin_inset Graphics
  7651. filename graphics/methylvoom/unadj.dupcor/pval-histograms-PAGE1.pdf
  7652. lyxscale 33
  7653. width 30col%
  7654. groupId meth-pval-hist
  7655. \end_inset
  7656. \end_layout
  7657. \begin_layout Plain Layout
  7658. \series bold
  7659. \begin_inset Caption Standard
  7660. \begin_layout Plain Layout
  7661. AR vs.
  7662. TX, Analysis A
  7663. \end_layout
  7664. \end_inset
  7665. \end_layout
  7666. \begin_layout Plain Layout
  7667. \end_layout
  7668. \end_inset
  7669. \begin_inset space \hfill{}
  7670. \end_inset
  7671. \begin_inset Float figure
  7672. wide false
  7673. sideways false
  7674. status collapsed
  7675. \begin_layout Plain Layout
  7676. \align center
  7677. \begin_inset Graphics
  7678. filename graphics/methylvoom/unadj.dupcor/pval-histograms-PAGE2.pdf
  7679. lyxscale 33
  7680. width 30col%
  7681. groupId meth-pval-hist
  7682. \end_inset
  7683. \end_layout
  7684. \begin_layout Plain Layout
  7685. \series bold
  7686. \begin_inset Caption Standard
  7687. \begin_layout Plain Layout
  7688. ADNR vs.
  7689. TX, Analysis A
  7690. \end_layout
  7691. \end_inset
  7692. \end_layout
  7693. \end_inset
  7694. \begin_inset space \hfill{}
  7695. \end_inset
  7696. \begin_inset Float figure
  7697. wide false
  7698. sideways false
  7699. status collapsed
  7700. \begin_layout Plain Layout
  7701. \align center
  7702. \begin_inset Graphics
  7703. filename graphics/methylvoom/unadj.dupcor/pval-histograms-PAGE3.pdf
  7704. lyxscale 33
  7705. width 30col%
  7706. groupId meth-pval-hist
  7707. \end_inset
  7708. \end_layout
  7709. \begin_layout Plain Layout
  7710. \series bold
  7711. \begin_inset Caption Standard
  7712. \begin_layout Plain Layout
  7713. CAN vs.
  7714. TX, Analysis A
  7715. \end_layout
  7716. \end_inset
  7717. \end_layout
  7718. \end_inset
  7719. \end_layout
  7720. \begin_layout Plain Layout
  7721. \align center
  7722. \series bold
  7723. \begin_inset Float figure
  7724. wide false
  7725. sideways false
  7726. status collapsed
  7727. \begin_layout Plain Layout
  7728. \align center
  7729. \begin_inset Graphics
  7730. filename graphics/methylvoom/unadj.dupcor.sva.aw/pval-histograms-PAGE1.pdf
  7731. lyxscale 33
  7732. width 30col%
  7733. groupId meth-pval-hist
  7734. \end_inset
  7735. \end_layout
  7736. \begin_layout Plain Layout
  7737. \series bold
  7738. \begin_inset Caption Standard
  7739. \begin_layout Plain Layout
  7740. AR vs.
  7741. TX, Analysis B
  7742. \end_layout
  7743. \end_inset
  7744. \end_layout
  7745. \end_inset
  7746. \begin_inset space \hfill{}
  7747. \end_inset
  7748. \begin_inset Float figure
  7749. wide false
  7750. sideways false
  7751. status collapsed
  7752. \begin_layout Plain Layout
  7753. \align center
  7754. \begin_inset Graphics
  7755. filename graphics/methylvoom/unadj.dupcor.sva.aw/pval-histograms-PAGE2.pdf
  7756. lyxscale 33
  7757. width 30col%
  7758. groupId meth-pval-hist
  7759. \end_inset
  7760. \end_layout
  7761. \begin_layout Plain Layout
  7762. \series bold
  7763. \begin_inset Caption Standard
  7764. \begin_layout Plain Layout
  7765. ADNR vs.
  7766. TX, Analysis B
  7767. \end_layout
  7768. \end_inset
  7769. \end_layout
  7770. \end_inset
  7771. \begin_inset space \hfill{}
  7772. \end_inset
  7773. \begin_inset Float figure
  7774. wide false
  7775. sideways false
  7776. status collapsed
  7777. \begin_layout Plain Layout
  7778. \align center
  7779. \begin_inset Graphics
  7780. filename graphics/methylvoom/unadj.dupcor.sva.aw/pval-histograms-PAGE3.pdf
  7781. lyxscale 33
  7782. width 30col%
  7783. groupId meth-pval-hist
  7784. \end_inset
  7785. \end_layout
  7786. \begin_layout Plain Layout
  7787. \series bold
  7788. \begin_inset Caption Standard
  7789. \begin_layout Plain Layout
  7790. CAN vs.
  7791. TX, Analysis B
  7792. \end_layout
  7793. \end_inset
  7794. \end_layout
  7795. \end_inset
  7796. \end_layout
  7797. \begin_layout Plain Layout
  7798. \align center
  7799. \series bold
  7800. \begin_inset Float figure
  7801. wide false
  7802. sideways false
  7803. status collapsed
  7804. \begin_layout Plain Layout
  7805. \align center
  7806. \begin_inset Graphics
  7807. filename graphics/methylvoom/unadj.dupcor.sva.voomaw/pval-histograms-PAGE1.pdf
  7808. lyxscale 33
  7809. width 30col%
  7810. groupId meth-pval-hist
  7811. \end_inset
  7812. \end_layout
  7813. \begin_layout Plain Layout
  7814. \series bold
  7815. \begin_inset Caption Standard
  7816. \begin_layout Plain Layout
  7817. AR vs.
  7818. TX, Analysis C
  7819. \end_layout
  7820. \end_inset
  7821. \end_layout
  7822. \end_inset
  7823. \begin_inset space \hfill{}
  7824. \end_inset
  7825. \begin_inset Float figure
  7826. wide false
  7827. sideways false
  7828. status collapsed
  7829. \begin_layout Plain Layout
  7830. \align center
  7831. \begin_inset Graphics
  7832. filename graphics/methylvoom/unadj.dupcor.sva.voomaw/pval-histograms-PAGE2.pdf
  7833. lyxscale 33
  7834. width 30col%
  7835. groupId meth-pval-hist
  7836. \end_inset
  7837. \end_layout
  7838. \begin_layout Plain Layout
  7839. \series bold
  7840. \begin_inset Caption Standard
  7841. \begin_layout Plain Layout
  7842. ADNR vs.
  7843. TX, Analysis C
  7844. \end_layout
  7845. \end_inset
  7846. \end_layout
  7847. \end_inset
  7848. \begin_inset space \hfill{}
  7849. \end_inset
  7850. \begin_inset Float figure
  7851. wide false
  7852. sideways false
  7853. status collapsed
  7854. \begin_layout Plain Layout
  7855. \align center
  7856. \begin_inset Graphics
  7857. filename graphics/methylvoom/unadj.dupcor.sva.voomaw/pval-histograms-PAGE3.pdf
  7858. lyxscale 33
  7859. width 30col%
  7860. groupId meth-pval-hist
  7861. \end_inset
  7862. \end_layout
  7863. \begin_layout Plain Layout
  7864. \series bold
  7865. \begin_inset Caption Standard
  7866. \begin_layout Plain Layout
  7867. CAN vs.
  7868. TX, Analysis C
  7869. \end_layout
  7870. \end_inset
  7871. \end_layout
  7872. \end_inset
  7873. \end_layout
  7874. \begin_layout Plain Layout
  7875. \begin_inset Caption Standard
  7876. \begin_layout Plain Layout
  7877. \series bold
  7878. \begin_inset CommandInset label
  7879. LatexCommand label
  7880. name "fig:meth-p-value-histograms"
  7881. \end_inset
  7882. Probe p-value histograms for each contrast in each analysis.
  7883. \series default
  7884. For each differential methylation test of interest, the distribution of
  7885. p-values across all probes is plotted as a histogram.
  7886. The red solid line indicates the density that would be expected under the
  7887. null hypothesis for all probes (a
  7888. \begin_inset Formula $\mathrm{Uniform}(0,1)$
  7889. \end_inset
  7890. distribution), while the blue dotted line indicates the fraction of p-values
  7891. that actually follow the null hypothesis (
  7892. \begin_inset Formula $\hat{\pi}_{0}$
  7893. \end_inset
  7894. ) estimated using the method of averaging local FDR values
  7895. \begin_inset CommandInset citation
  7896. LatexCommand cite
  7897. key "Phipson2013Thesis"
  7898. literal "false"
  7899. \end_inset
  7900. .
  7901. the blue line is only shown in each plot if the estimate of
  7902. \begin_inset Formula $\hat{\pi}_{0}$
  7903. \end_inset
  7904. for that p-value distribution is different from 1.
  7905. \end_layout
  7906. \end_inset
  7907. \end_layout
  7908. \end_inset
  7909. \end_layout
  7910. \begin_layout Standard
  7911. Table
  7912. \begin_inset CommandInset ref
  7913. LatexCommand ref
  7914. reference "tab:methyl-num-signif"
  7915. plural "false"
  7916. caps "false"
  7917. noprefix "false"
  7918. \end_inset
  7919. shows the number of significantly differentially methylated probes reported
  7920. by each analysis for each comparison of interest at an FDR of 10%.
  7921. As expected, the more elaborate analyses, B and C, report more significant
  7922. probes than the more basic analysis A, consistent with the conclusions
  7923. above that the data contain hidden systematic variations that must be modeled.
  7924. Table
  7925. \begin_inset CommandInset ref
  7926. LatexCommand ref
  7927. reference "tab:methyl-est-nonnull"
  7928. plural "false"
  7929. caps "false"
  7930. noprefix "false"
  7931. \end_inset
  7932. shows the estimated number differentially methylated probes for each test
  7933. from each analysis.
  7934. This was computed by estimating the proportion of null hypotheses that
  7935. were true using the method of
  7936. \begin_inset CommandInset citation
  7937. LatexCommand cite
  7938. key "Phipson2013Thesis"
  7939. literal "false"
  7940. \end_inset
  7941. and subtracting that fraction from the total number of probes, yielding
  7942. an estimate of the number of null hypotheses that are false based on the
  7943. distribution of p-values across the entire dataset.
  7944. Note that this does not identify which null hypotheses should be rejected
  7945. (i.e.
  7946. which probes are significant); it only estimates the true number of such
  7947. probes.
  7948. Once again, analyses B and C result it much larger estimates for the number
  7949. of differentially methylated probes.
  7950. In this case, analysis C, the only analysis that includes voom, estimates
  7951. the largest number of differentially methylated probes for all 3 contrasts.
  7952. If the assumptions of all the methods employed hold, then this represents
  7953. a gain in statistical power over the simpler analysis A.
  7954. Figure
  7955. \begin_inset CommandInset ref
  7956. LatexCommand ref
  7957. reference "fig:meth-p-value-histograms"
  7958. plural "false"
  7959. caps "false"
  7960. noprefix "false"
  7961. \end_inset
  7962. shows the p-value distributions for each test, from which the numbers in
  7963. Table
  7964. \begin_inset CommandInset ref
  7965. LatexCommand ref
  7966. reference "tab:methyl-est-nonnull"
  7967. plural "false"
  7968. caps "false"
  7969. noprefix "false"
  7970. \end_inset
  7971. were generated.
  7972. The distributions for analysis A all have a dip in density near zero, which
  7973. is a strong sign of a poor model fit.
  7974. The histograms for analyses B and C are more well-behaved, with a uniform
  7975. component stretching all the way from 0 to 1 representing the probes for
  7976. which the null hypotheses is true (no differential methylation), and a
  7977. zero-biased component representing the probes for which the null hypothesis
  7978. is false (differentially methylated).
  7979. These histograms do not indicate any major issues with the model fit.
  7980. \end_layout
  7981. \begin_layout Standard
  7982. \begin_inset Flex TODO Note (inline)
  7983. status open
  7984. \begin_layout Plain Layout
  7985. If time allows, maybe generate the PCA plots before/after SVA effect subtraction
  7986. ?
  7987. \end_layout
  7988. \end_inset
  7989. \end_layout
  7990. \begin_layout Section
  7991. Discussion
  7992. \end_layout
  7993. \begin_layout Subsection
  7994. fRMA achieves clinically applicable normalization without sacrificing classifica
  7995. tion performance
  7996. \end_layout
  7997. \begin_layout Standard
  7998. As shown in Figure
  7999. \begin_inset CommandInset ref
  8000. LatexCommand ref
  8001. reference "fig:Classifier-probabilities-RMA"
  8002. plural "false"
  8003. caps "false"
  8004. noprefix "false"
  8005. \end_inset
  8006. , improper normalization, particularly separate normalization of training
  8007. and test samples, leads to unwanted biases in classification.
  8008. In a controlled experimental context, it is always possible to correct
  8009. this issue by normalizing all experimental samples together.
  8010. However, because it is not feasible to normalize all samples together in
  8011. a clinical context, a single-channel normalization is required is required.
  8012. \end_layout
  8013. \begin_layout Standard
  8014. The major concern in using a single-channel normalization is that non-single-cha
  8015. nnel methods can share information between arrays to improve the normalization,
  8016. and single-channel methods risk sacrificing the gains in normalization
  8017. accuracy that come from this information sharing.
  8018. In the case of RMA, this information sharing is accomplished through quantile
  8019. normalization and median polish steps.
  8020. The need for information sharing in quantile normalization can easily be
  8021. removed by learning a fixed set of quantiles from external data and normalizing
  8022. each array to these fixed quantiles, instead of the quantiles of the data
  8023. itself.
  8024. As long as the fixed quantiles are reasonable, the result will be similar
  8025. to standard RMA.
  8026. However, there is no analogous way to eliminate cross-array information
  8027. sharing in the median polish step, so fRMA replaces this with a weighted
  8028. average of probes on each array, with the weights learned from external
  8029. data.
  8030. This step of fRMA has the greatest potential to diverge from RMA un undesirable
  8031. ways.
  8032. \end_layout
  8033. \begin_layout Standard
  8034. However, when run on real data, fRMA performed at least as well as RMA in
  8035. both the internal validation and external validation tests.
  8036. This shows that fRMA can be used to normalize individual clinical samples
  8037. in a class prediction context without sacrificing the classifier performance
  8038. that would be obtained by using the more well-established RMA for normalization.
  8039. The other single-channel normalization method considered, SCAN, showed
  8040. some loss of AUC in the external validation test.
  8041. Based on these results, fRMA is the preferred normalization for clinical
  8042. samples in a class prediction context.
  8043. \end_layout
  8044. \begin_layout Subsection
  8045. Robust fRMA vectors can be generated for new array platforms
  8046. \end_layout
  8047. \begin_layout Standard
  8048. \begin_inset Flex TODO Note (inline)
  8049. status open
  8050. \begin_layout Plain Layout
  8051. Look up the exact numbers, do a find & replace for
  8052. \begin_inset Quotes eld
  8053. \end_inset
  8054. 850
  8055. \begin_inset Quotes erd
  8056. \end_inset
  8057. \end_layout
  8058. \end_inset
  8059. \end_layout
  8060. \begin_layout Standard
  8061. The published fRMA normalization vectors for the hgu133plus2 platform were
  8062. generated from a set of about 850 samples chosen from a wide range of tissues,
  8063. which the authors determined was sufficient to generate a robust set of
  8064. normalization vectors that could be applied across all tissues
  8065. \begin_inset CommandInset citation
  8066. LatexCommand cite
  8067. key "McCall2010"
  8068. literal "false"
  8069. \end_inset
  8070. .
  8071. Since we only had hthgu133pluspm for 2 tissues of interest, our needs were
  8072. more modest.
  8073. Even using only 130 samples in 26 batches of 5 samples each for kidney
  8074. biopsies, we were able to train a robust set of fRMA normalization vectors
  8075. that were not meaningfully affected by the random selection of 5 samples
  8076. from each batch.
  8077. As expected, the training process was just as robust for the blood samples
  8078. with 230 samples in 46 batches of 5 samples each.
  8079. Because these vectors were each generated using training samples from a
  8080. single tissue, they are not suitable for general use, unlike the vectors
  8081. provided with fRMA itself.
  8082. They are purpose-built for normalizing a specific type of sample on a specific
  8083. platform.
  8084. This is a mostly acceptable limitation in the context of developing a machine
  8085. learning classifier for diagnosing a disease based on samples of a specific
  8086. tissue.
  8087. \end_layout
  8088. \begin_layout Standard
  8089. \begin_inset Flex TODO Note (inline)
  8090. status open
  8091. \begin_layout Plain Layout
  8092. Talk about how these vectors can be used for any data from these tissues
  8093. on this platform even though they were custom made for this data set.
  8094. \end_layout
  8095. \end_inset
  8096. \end_layout
  8097. \begin_layout Standard
  8098. \begin_inset Flex TODO Note (inline)
  8099. status open
  8100. \begin_layout Plain Layout
  8101. How to bring up that these custom vectors were used in another project by
  8102. someone else that was never published?
  8103. \end_layout
  8104. \end_inset
  8105. \end_layout
  8106. \begin_layout Subsection
  8107. Methylation array data can be successfully analyzed using existing techniques,
  8108. but machine learning poses additional challenges
  8109. \end_layout
  8110. \begin_layout Standard
  8111. Both analysis strategies B and C both yield a reasonable analysis, with
  8112. a mean-variance trend that matches the expected behavior for the non-linear
  8113. M-value transformation (Figure
  8114. \begin_inset CommandInset ref
  8115. LatexCommand ref
  8116. reference "fig:meanvar-sva-aw"
  8117. plural "false"
  8118. caps "false"
  8119. noprefix "false"
  8120. \end_inset
  8121. ) and well-behaved p-value distributions (Figure
  8122. \begin_inset CommandInset ref
  8123. LatexCommand ref
  8124. reference "fig:meth-p-value-histograms"
  8125. plural "false"
  8126. caps "false"
  8127. noprefix "false"
  8128. \end_inset
  8129. ).
  8130. These two analyses also yield similar numbers of significant probes (Table
  8131. \begin_inset CommandInset ref
  8132. LatexCommand ref
  8133. reference "tab:methyl-num-signif"
  8134. plural "false"
  8135. caps "false"
  8136. noprefix "false"
  8137. \end_inset
  8138. ) and similar estimates of the number of differentially methylated probes
  8139. (Table
  8140. \begin_inset CommandInset ref
  8141. LatexCommand ref
  8142. reference "tab:methyl-est-nonnull"
  8143. plural "false"
  8144. caps "false"
  8145. noprefix "false"
  8146. \end_inset
  8147. ).
  8148. The main difference between these two analyses is the method used to account
  8149. for the mean-variance trend.
  8150. In analysis B, the trend is estimated and applied at the probe level: each
  8151. probe's estimated variance is squeezed toward the trend using an empirical
  8152. Bayes procedure (Figure
  8153. \begin_inset CommandInset ref
  8154. LatexCommand ref
  8155. reference "fig:meanvar-sva-aw"
  8156. plural "false"
  8157. caps "false"
  8158. noprefix "false"
  8159. \end_inset
  8160. ).
  8161. In analysis C, the trend is still estimated at the probe level, but instead
  8162. of estimating a single variance value shared across all observations for
  8163. a given probe, the voom method computes an initial estiamte of the variance
  8164. for each observation individually based on where its model-fitted M-value
  8165. falls on the trend line and then assigns inverse-variance weights to model
  8166. the difference in variance between observations.
  8167. An overall variance is still estimated for each probe using the same empirical
  8168. Bayes method, but now the residual trend is flat (Figure
  8169. \begin_inset CommandInset ref
  8170. LatexCommand ref
  8171. reference "fig:meanvar-sva-voomaw"
  8172. plural "false"
  8173. caps "false"
  8174. noprefix "false"
  8175. \end_inset
  8176. ), indicating that the mean-variance trend is adequately modeled by scaling
  8177. the estimated variance for each observation using the weights computed
  8178. by voom.
  8179. \end_layout
  8180. \begin_layout Standard
  8181. The difference between the standard empirical Bayes trended variance modeling
  8182. (analysis B) and voom (analysis C) is analogous to the difference between
  8183. a t-test with equal variance and a t-test with unequal variance, except
  8184. that the unequal group variances used in the latter test are estimated
  8185. based on the mean-variance trend from all the probes rather than the data
  8186. for the specific probe being tested, thus stabilizing the group variance
  8187. estimates by sharing information between probes.
  8188. Allowing voom to model the variance using observation weights in this manner
  8189. allows the linear model fit to concentrate statistical power where it will
  8190. do the most good.
  8191. For example, if a particular probe's M-values are always at the extreme
  8192. of the M-value range (e.g.
  8193. less than -4) for ADNR samples, but the M-values for that probe in TX and
  8194. CAN samples are within the flat region of the mean-variance trend (between
  8195. -3 and +3), voom is able to down-weight the contribution of the high-variance
  8196. M-values from the ADNR samples in order to gain more statistical power
  8197. while testing for differential methylation between TX and CAN.
  8198. In contrast, modeling the mean-variance trend only at the probe level would
  8199. combine the high-variance ADNR samples and lower-variance samples from
  8200. other conditions and estimate an intermediate variance for this probe.
  8201. In practice, analysis B shows that this approach is adequate, but the voom
  8202. approach in analysis C is at least as good on all model fit criteria and
  8203. yields a larger estimate for the number of differentially methylated genes,
  8204. \emph on
  8205. and
  8206. \emph default
  8207. it matches up better with the theoretical
  8208. \end_layout
  8209. \begin_layout Standard
  8210. The significant association of diebetes diagnosis with sample quality is
  8211. interesting.
  8212. The samples with Type 2 diabetes tended to have more variation, averaged
  8213. across all probes, than those with Type 1 diabetes.
  8214. This is consistent with the consensus that type 2 disbetes and the associated
  8215. metabolic syndrome represent a broad dysregulation of the body's endocrine
  8216. signalling related to metabolism [citation needed].
  8217. This dysregulation could easily manifest as a greater degree of variation
  8218. in the DNA methylation patterns of affected tissues.
  8219. In contrast, Type 1 disbetes has a more specific cause and effect, so a
  8220. less variable methylation signature is expected.
  8221. \end_layout
  8222. \begin_layout Standard
  8223. This preliminary anlaysis suggests that some degree of differential methylation
  8224. exists between TX and each of the three types of transplant disfunction
  8225. studied.
  8226. Hence, it may be feasible to train a classifier to diagnose transplant
  8227. disfunction from DNA methylation array data.
  8228. However, the major importance of both SVA and sample quality weighting
  8229. for proper modeling of this data poses significant challenges for any attempt
  8230. at a machine learning on data of similar quality.
  8231. While these are easily used in a modeling context with full sample information,
  8232. neither of these methods is directly applicable in a machine learning context,
  8233. where the diagnosis is not known ahead of time.
  8234. If a machine learning approach for methylation-based diagnosis is to be
  8235. pursued, it will either require machine-learning-friendly methods to address
  8236. the same systematic trends in the data that SVA and sample quality weighting
  8237. address, or it will require higher quality data with substantially less
  8238. systematic perturbation of the data.
  8239. \end_layout
  8240. \begin_layout Section
  8241. Future Directions
  8242. \end_layout
  8243. \begin_layout Standard
  8244. \begin_inset Flex TODO Note (inline)
  8245. status open
  8246. \begin_layout Plain Layout
  8247. Some work was already being done with the existing fRMA vectors.
  8248. Do I mention that here?
  8249. \end_layout
  8250. \end_inset
  8251. \end_layout
  8252. \begin_layout Subsection
  8253. Improving fRMA to allow training from batches of unequal size
  8254. \end_layout
  8255. \begin_layout Standard
  8256. Because the tools for building fRMA normalization vectors require equal-size
  8257. batches, many samples must be discarded from the training data.
  8258. This is undesirable for a few reasons.
  8259. First, more data is simply better, all other things being equal.
  8260. In this case,
  8261. \begin_inset Quotes eld
  8262. \end_inset
  8263. better
  8264. \begin_inset Quotes erd
  8265. \end_inset
  8266. means a more precise estimate of normalization parameters.
  8267. In addition, the samples to be discarded must be chosen arbitrarily, which
  8268. introduces an unnecessary element of randomness into the estimation process.
  8269. While the randomness can be made deterministic by setting a consistent
  8270. random seed, the need for equal size batches also introduces a need for
  8271. the analyst to decide on the appropriate trade-off between batch size and
  8272. the number of batches.
  8273. This introduces an unnecessary and undesirable
  8274. \begin_inset Quotes eld
  8275. \end_inset
  8276. researcher degree of freedom
  8277. \begin_inset Quotes erd
  8278. \end_inset
  8279. into the analysis, since the generated normalization vectors now depend
  8280. on the choice of batch size based on vague selection criteria and instinct,
  8281. which can unintentionally inproduce bias if the researcher chooses a batch
  8282. size based on what seems to yield the most favorable downstream results
  8283. \begin_inset CommandInset citation
  8284. LatexCommand cite
  8285. key "Simmons2011"
  8286. literal "false"
  8287. \end_inset
  8288. .
  8289. \end_layout
  8290. \begin_layout Standard
  8291. Fortunately, the requirement for equal-size batches is not inherent to the
  8292. fRMA algorithm but rather a limitation of the implementation in the frmaTools
  8293. package.
  8294. In personal communication, the package's author, Matthew McCall, has indicated
  8295. that with some work, it should be possible to improve the implementation
  8296. to work with batches of unequal sizes.
  8297. The current implementation ignores the batch size when calculating with-batch
  8298. and between-batch residual variances, since the batch size constant cancels
  8299. out later in the calculations as long as all batches are of equal size.
  8300. Hence, the calculations of these parameters would need to be modified to
  8301. remove this optimization and properly calculate the variances using the
  8302. full formula.
  8303. Once this modification is made, a new strategy would need to be developed
  8304. for assessing the stability of parameter estimates, since the random subsamplin
  8305. g step is eliminated, meaning that different subsamplings can no longer
  8306. be compared as in Figures
  8307. \begin_inset CommandInset ref
  8308. LatexCommand ref
  8309. reference "fig:frma-violin"
  8310. plural "false"
  8311. caps "false"
  8312. noprefix "false"
  8313. \end_inset
  8314. and
  8315. \begin_inset CommandInset ref
  8316. LatexCommand ref
  8317. reference "fig:Representative-MA-plots"
  8318. plural "false"
  8319. caps "false"
  8320. noprefix "false"
  8321. \end_inset
  8322. .
  8323. Bootstrap resampling is likely a good candidate here: sample many training
  8324. sets of equal size from the existing training set with replacement, estimate
  8325. parameters from each resampled training set, and compare the estimated
  8326. parameters between bootstraps in order to quantify the variability in each
  8327. parameter's estimation.
  8328. \end_layout
  8329. \begin_layout Chapter
  8330. Globin-blocking for more effective blood RNA-seq analysis in primate animal
  8331. model
  8332. \end_layout
  8333. \begin_layout Standard
  8334. \begin_inset Flex TODO Note (inline)
  8335. status open
  8336. \begin_layout Plain Layout
  8337. Choose between above and the paper title: Optimizing yield of deep RNA sequencin
  8338. g for gene expression profiling by globin reduction of peripheral blood
  8339. samples from cynomolgus monkeys (Macaca fascicularis).
  8340. \end_layout
  8341. \end_inset
  8342. \end_layout
  8343. \begin_layout Standard
  8344. \begin_inset Flex TODO Note (inline)
  8345. status open
  8346. \begin_layout Plain Layout
  8347. Chapter author list: https://tex.stackexchange.com/questions/156862/displaying-aut
  8348. hor-for-each-chapter-in-book Every chapter gets an author list, which may
  8349. or may not be part of a citation to a published/preprinted paper.
  8350. \end_layout
  8351. \end_inset
  8352. \end_layout
  8353. \begin_layout Standard
  8354. \begin_inset Flex TODO Note (inline)
  8355. status open
  8356. \begin_layout Plain Layout
  8357. Preprint then cite the paper
  8358. \end_layout
  8359. \end_inset
  8360. \end_layout
  8361. \begin_layout Section*
  8362. Abstract
  8363. \end_layout
  8364. \begin_layout Paragraph
  8365. Background
  8366. \end_layout
  8367. \begin_layout Standard
  8368. Primate blood contains high concentrations of globin messenger RNA.
  8369. Globin reduction is a standard technique used to improve the expression
  8370. results obtained by DNA microarrays on RNA from blood samples.
  8371. However, with whole transcriptome RNA-sequencing (RNA-seq) quickly replacing
  8372. microarrays for many applications, the impact of globin reduction for RNA-seq
  8373. has not been previously studied.
  8374. Moreover, no off-the-shelf kits are available for globin reduction in nonhuman
  8375. primates.
  8376. \end_layout
  8377. \begin_layout Paragraph
  8378. Results
  8379. \end_layout
  8380. \begin_layout Standard
  8381. Here we report a protocol for RNA-seq in primate blood samples that uses
  8382. complimentary oligonucleotides to block reverse transcription of the alpha
  8383. and beta globin genes.
  8384. In test samples from cynomolgus monkeys (Macaca fascicularis), this globin
  8385. blocking protocol approximately doubles the yield of informative (non-globin)
  8386. reads by greatly reducing the fraction of globin reads, while also improving
  8387. the consistency in sequencing depth between samples.
  8388. The increased yield enables detection of about 2000 more genes, significantly
  8389. increases the correlation in measured gene expression levels between samples,
  8390. and increases the sensitivity of differential gene expression tests.
  8391. \end_layout
  8392. \begin_layout Paragraph
  8393. Conclusions
  8394. \end_layout
  8395. \begin_layout Standard
  8396. These results show that globin blocking significantly improves the cost-effectiv
  8397. eness of mRNA sequencing in primate blood samples by doubling the yield
  8398. of useful reads, allowing detection of more genes, and improving the precision
  8399. of gene expression measurements.
  8400. Based on these results, a globin reducing or blocking protocol is recommended
  8401. for all RNA-seq studies of primate blood samples.
  8402. \end_layout
  8403. \begin_layout Section
  8404. Approach
  8405. \end_layout
  8406. \begin_layout Standard
  8407. \begin_inset Note Note
  8408. status open
  8409. \begin_layout Plain Layout
  8410. Consider putting some of this in the Intro chapter
  8411. \end_layout
  8412. \begin_layout Itemize
  8413. Cynomolgus monkeys as a model organism
  8414. \end_layout
  8415. \begin_deeper
  8416. \begin_layout Itemize
  8417. Highly related to humans
  8418. \end_layout
  8419. \begin_layout Itemize
  8420. Small size and short life cycle - good research animal
  8421. \end_layout
  8422. \begin_layout Itemize
  8423. Genomics resources still in development
  8424. \end_layout
  8425. \end_deeper
  8426. \begin_layout Itemize
  8427. Inadequacy of existing blood RNA-seq protocols
  8428. \end_layout
  8429. \begin_deeper
  8430. \begin_layout Itemize
  8431. Existing protocols use a separate globin pulldown step, slowing down processing
  8432. \end_layout
  8433. \end_deeper
  8434. \end_inset
  8435. \end_layout
  8436. \begin_layout Standard
  8437. Increasingly, researchers are turning to high-throughput mRNA sequencing
  8438. technologies (RNA-seq) in preference to expression microarrays for analysis
  8439. of gene expression
  8440. \begin_inset CommandInset citation
  8441. LatexCommand cite
  8442. key "Mutz2012"
  8443. literal "false"
  8444. \end_inset
  8445. .
  8446. The advantages are even greater for study of model organisms with no well-estab
  8447. lished array platforms available, such as the cynomolgus monkey (Macaca
  8448. fascicularis).
  8449. High fractions of globin mRNA are naturally present in mammalian peripheral
  8450. blood samples (up to 70% of total mRNA) and these are known to interfere
  8451. with the results of array-based expression profiling
  8452. \begin_inset CommandInset citation
  8453. LatexCommand cite
  8454. key "Winn2010"
  8455. literal "false"
  8456. \end_inset
  8457. .
  8458. The importance of globin reduction for RNA-seq of blood has only been evaluated
  8459. for a deepSAGE protocol on human samples
  8460. \begin_inset CommandInset citation
  8461. LatexCommand cite
  8462. key "Mastrokolias2012"
  8463. literal "false"
  8464. \end_inset
  8465. .
  8466. In the present report, we evaluated globin reduction using custom blocking
  8467. oligonucleotides for deep RNA-seq of peripheral blood samples from a nonhuman
  8468. primate, cynomolgus monkey, using the Illumina technology platform.
  8469. We demonstrate that globin reduction significantly improves the cost-effectiven
  8470. ess of RNA-seq in blood samples.
  8471. Thus, our protocol offers a significant advantage to any investigator planning
  8472. to use RNA-seq for gene expression profiling of nonhuman primate blood
  8473. samples.
  8474. Our method can be generally applied to any species by designing complementary
  8475. oligonucleotide blocking probes to the globin gene sequences of that species.
  8476. Indeed, any highly expressed but biologically uninformative transcripts
  8477. can also be blocked to further increase sequencing efficiency and value
  8478. \begin_inset CommandInset citation
  8479. LatexCommand cite
  8480. key "Arnaud2016"
  8481. literal "false"
  8482. \end_inset
  8483. .
  8484. \end_layout
  8485. \begin_layout Section
  8486. Methods
  8487. \end_layout
  8488. \begin_layout Subsection
  8489. Sample collection
  8490. \end_layout
  8491. \begin_layout Standard
  8492. All research reported here was done under IACUC-approved protocols at the
  8493. University of Miami and complied with all applicable federal and state
  8494. regulations and ethical principles for nonhuman primate research.
  8495. Blood draws occurred between 16 April 2012 and 18 June 2015.
  8496. The experimental system involved intrahepatic pancreatic islet transplantation
  8497. into Cynomolgus monkeys with induced diabetes mellitus with or without
  8498. concomitant infusion of mesenchymal stem cells.
  8499. Blood was collected at serial time points before and after transplantation
  8500. into PAXgene Blood RNA tubes (PreAnalytiX/Qiagen, Valencia, CA) at the
  8501. precise volume:volume ratio of 2.5 ml whole blood into 6.9 ml of PAX gene
  8502. additive.
  8503. \end_layout
  8504. \begin_layout Subsection
  8505. Globin Blocking
  8506. \end_layout
  8507. \begin_layout Standard
  8508. Four oligonucleotides were designed to hybridize to the 3’ end of the transcript
  8509. s for Cynomolgus HBA1, HBA2 and HBB, with two hybridization sites for HBB
  8510. and 2 sites for HBA (the chosen sites were identical in both HBA genes).
  8511. All oligos were purchased from Sigma and were entirely composed of 2’O-Me
  8512. bases with a C3 spacer positioned at the 3’ ends to prevent any polymerase
  8513. mediated primer extension.
  8514. \end_layout
  8515. \begin_layout Quote
  8516. HBA1/2 site 1: GCCCACUCAGACUUUAUUCAAAG-C3spacer
  8517. \end_layout
  8518. \begin_layout Quote
  8519. HBA1/2 site 2: GGUGCAAGGAGGGGAGGAG-C3spacer
  8520. \end_layout
  8521. \begin_layout Quote
  8522. HBB site 1: AAUGAAAAUAAAUGUUUUUUAUUAG-C3spacer
  8523. \end_layout
  8524. \begin_layout Quote
  8525. HBB site 2: CUCAAGGCCCUUCAUAAUAUCCC-C3spacer
  8526. \end_layout
  8527. \begin_layout Subsection
  8528. RNA-seq Library Preparation
  8529. \end_layout
  8530. \begin_layout Standard
  8531. Sequencing libraries were prepared with 200ng total RNA from each sample.
  8532. Polyadenylated mRNA was selected from 200 ng aliquots of cynomologus blood-deri
  8533. ved total RNA using Ambion Dynabeads Oligo(dT)25 beads (Invitrogen) following
  8534. manufacturer’s recommended protocol.
  8535. PolyA selected RNA was then combined with 8 pmol of HBA1/2 (site 1), 8
  8536. pmol of HBA1/2 (site 2), 12 pmol of HBB (site 1) and 12 pmol of HBB (site
  8537. 2) oligonucleotides.
  8538. In addition, 20 pmol of RT primer containing a portion of the Illumina
  8539. adapter sequence (B-oligo-dTV: GAGTTCCTTGGCACCCGAGAATTCCATTTTTTTTTTTTTTTTTTTV)
  8540. and 4 µL of 5X First Strand buffer (250 mM Tris-HCl pH 8.3, 375 mM KCl,
  8541. 15mM MgCl2) were added in a total volume of 15 µL.
  8542. The RNA was fragmented by heating this cocktail for 3 minutes at 95°C and
  8543. then placed on ice.
  8544. This was followed by the addition of 2 µL 0.1 M DTT, 1 µL RNaseOUT, 1 µL
  8545. 10mM dNTPs 10% biotin-16 aminoallyl-2’- dUTP and 10% biotin-16 aminoallyl-2’-
  8546. dCTP (TriLink Biotech, San Diego, CA), 1 µL Superscript II (200U/ µL, Thermo-Fi
  8547. sher).
  8548. A second “unblocked” library was prepared in the same way for each sample
  8549. but replacing the blocking oligos with an equivalent volume of water.
  8550. The reaction was carried out at 25°C for 15 minutes and 42°C for 40 minutes,
  8551. followed by incubation at 75°C for 10 minutes to inactivate the reverse
  8552. transcriptase.
  8553. \end_layout
  8554. \begin_layout Standard
  8555. The cDNA/RNA hybrid molecules were purified using 1.8X Ampure XP beads (Agencourt
  8556. ) following supplier’s recommended protocol.
  8557. The cDNA/RNA hybrid was eluted in 25 µL of 10 mM Tris-HCl pH 8.0, and then
  8558. bound to 25 µL of M280 Magnetic Streptavidin beads washed per recommended
  8559. protocol (Thermo-Fisher).
  8560. After 30 minutes of binding, beads were washed one time in 100 µL 0.1N NaOH
  8561. to denature and remove the bound RNA, followed by two 100 µL washes with
  8562. 1X TE buffer.
  8563. \end_layout
  8564. \begin_layout Standard
  8565. Subsequent attachment of the 5-prime Illumina A adapter was performed by
  8566. on-bead random primer extension of the following sequence (A-N8 primer:
  8567. TTCAGAGTTCTACAGTCCGACGATCNNNNNNNN).
  8568. Briefly, beads were resuspended in a 20 µL reaction containing 5 µM A-N8
  8569. primer, 40mM Tris-HCl pH 7.5, 20mM MgCl2, 50mM NaCl, 0.325U/µL Sequenase
  8570. 2.0 (Affymetrix, Santa Clara, CA), 0.0025U/µL inorganic pyrophosphatase (Affymetr
  8571. ix) and 300 µM each dNTP.
  8572. Reaction was incubated at 22°C for 30 minutes, then beads were washed 2
  8573. times with 1X TE buffer (200µL).
  8574. \end_layout
  8575. \begin_layout Standard
  8576. The magnetic streptavidin beads were resuspended in 34 µL nuclease-free
  8577. water and added directly to a PCR tube.
  8578. The two Illumina protocol-specified PCR primers were added at 0.53 µM (Illumina
  8579. TruSeq Universal Primer 1 and Illumina TruSeq barcoded PCR primer 2), along
  8580. with 40 µL 2X KAPA HiFi Hotstart ReadyMix (KAPA, Willmington MA) and thermocycl
  8581. ed as follows: starting with 98°C (2 min-hold); 15 cycles of 98°C, 20sec;
  8582. 60°C, 30sec; 72°C, 30sec; and finished with a 72°C (2 min-hold).
  8583. \end_layout
  8584. \begin_layout Standard
  8585. PCR products were purified with 1X Ampure Beads following manufacturer’s
  8586. recommended protocol.
  8587. Libraries were then analyzed using the Agilent TapeStation and quantitation
  8588. of desired size range was performed by “smear analysis”.
  8589. Samples were pooled in equimolar batches of 16 samples.
  8590. Pooled libraries were size selected on 2% agarose gels (E-Gel EX Agarose
  8591. Gels; Thermo-Fisher).
  8592. Products were cut between 250 and 350 bp (corresponding to insert sizes
  8593. of 130 to 230 bps).
  8594. Finished library pools were then sequenced on the Illumina NextSeq500 instrumen
  8595. t with 75 base read lengths.
  8596. \end_layout
  8597. \begin_layout Subsection
  8598. Read alignment and counting
  8599. \end_layout
  8600. \begin_layout Standard
  8601. Reads were aligned to the cynomolgus genome using STAR
  8602. \begin_inset CommandInset citation
  8603. LatexCommand cite
  8604. key "Dobin2013,Wilson2013"
  8605. literal "false"
  8606. \end_inset
  8607. .
  8608. Counts of uniquely mapped reads were obtained for every gene in each sample
  8609. with the “featureCounts” function from the Rsubread package, using each
  8610. of the three possibilities for the “strandSpecific” option: sense, antisense,
  8611. and unstranded
  8612. \begin_inset CommandInset citation
  8613. LatexCommand cite
  8614. key "Liao2014"
  8615. literal "false"
  8616. \end_inset
  8617. .
  8618. A few artifacts in the cynomolgus genome annotation complicated read counting.
  8619. First, no ortholog is annotated for alpha globin in the cynomolgus genome,
  8620. presumably because the human genome has two alpha globin genes with nearly
  8621. identical sequences, making the orthology relationship ambiguous.
  8622. However, two loci in the cynomolgus genome are as “hemoglobin subunit alpha-lik
  8623. e” (LOC102136192 and LOC102136846).
  8624. LOC102136192 is annotated as a pseudogene while LOC102136846 is annotated
  8625. as protein-coding.
  8626. Our globin reduction protocol was designed to include blocking of these
  8627. two genes.
  8628. Indeed, these two genes have almost the same read counts in each library
  8629. as the properly-annotated HBB gene and much larger counts than any other
  8630. gene in the unblocked libraries, giving confidence that reads derived from
  8631. the real alpha globin are mapping to both genes.
  8632. Thus, reads from both of these loci were counted as alpha globin reads
  8633. in all further analyses.
  8634. The second artifact is a small, uncharacterized non-coding RNA gene (LOC1021365
  8635. 91), which overlaps the HBA-like gene (LOC102136192) on the opposite strand.
  8636. If counting is not performed in stranded mode (or if a non-strand-specific
  8637. sequencing protocol is used), many reads mapping to the globin gene will
  8638. be discarded as ambiguous due to their overlap with this ncRNA gene, resulting
  8639. in significant undercounting of globin reads.
  8640. Therefore, stranded sense counts were used for all further analysis in
  8641. the present study to insure that we accurately accounted for globin transcript
  8642. reduction.
  8643. However, we note that stranded reads are not necessary for RNA-seq using
  8644. our protocol in standard practice.
  8645. \end_layout
  8646. \begin_layout Subsection
  8647. Normalization and Exploratory Data Analysis
  8648. \end_layout
  8649. \begin_layout Standard
  8650. Libraries were normalized by computing scaling factors using the edgeR package’s
  8651. Trimmed Mean of M-values method
  8652. \begin_inset CommandInset citation
  8653. LatexCommand cite
  8654. key "Robinson2010"
  8655. literal "false"
  8656. \end_inset
  8657. .
  8658. Log2 counts per million values (logCPM) were calculated using the cpm function
  8659. in edgeR for individual samples and aveLogCPM function for averages across
  8660. groups of samples, using those functions’ default prior count values to
  8661. avoid taking the logarithm of 0.
  8662. Genes were considered “present” if their average normalized logCPM values
  8663. across all libraries were at least -1.
  8664. Normalizing for gene length was unnecessary because the sequencing protocol
  8665. is 3’-biased and hence the expected read count for each gene is related
  8666. to the transcript’s copy number but not its length.
  8667. \end_layout
  8668. \begin_layout Standard
  8669. In order to assess the effect of blocking on reproducibility, Pearson and
  8670. Spearman correlation coefficients were computed between the logCPM values
  8671. for every pair of libraries within the globin-blocked (GB) and unblocked
  8672. (non-GB) groups, and edgeR's “estimateDisp” function was used to compute
  8673. negative binomial dispersions separately for the two groups
  8674. \begin_inset CommandInset citation
  8675. LatexCommand cite
  8676. key "Chen2014"
  8677. literal "false"
  8678. \end_inset
  8679. .
  8680. \end_layout
  8681. \begin_layout Subsection
  8682. Differential Expression Analysis
  8683. \end_layout
  8684. \begin_layout Standard
  8685. All tests for differential gene expression were performed using edgeR, by
  8686. first fitting a negative binomial generalized linear model to the counts
  8687. and normalization factors and then performing a quasi-likelihood F-test
  8688. with robust estimation of outlier gene dispersions
  8689. \begin_inset CommandInset citation
  8690. LatexCommand cite
  8691. key "Lund2012,Phipson2016"
  8692. literal "false"
  8693. \end_inset
  8694. .
  8695. To investigate the effects of globin blocking on each gene, an additive
  8696. model was fit to the full data with coefficients for globin blocking and
  8697. SampleID.
  8698. To test the effect of globin blocking on detection of differentially expressed
  8699. genes, the GB samples and non-GB samples were each analyzed independently
  8700. as follows: for each animal with both a pre-transplant and a post-transplant
  8701. time point in the data set, the pre-transplant sample and the earliest
  8702. post-transplant sample were selected, and all others were excluded, yielding
  8703. a pre-/post-transplant pair of samples for each animal (N=7 animals with
  8704. paired samples).
  8705. These samples were analyzed for pre-transplant vs.
  8706. post-transplant differential gene expression while controlling for inter-animal
  8707. variation using an additive model with coefficients for transplant and
  8708. animal ID.
  8709. In all analyses, p-values were adjusted using the Benjamini-Hochberg procedure
  8710. for FDR control
  8711. \begin_inset CommandInset citation
  8712. LatexCommand cite
  8713. key "Benjamini1995"
  8714. literal "false"
  8715. \end_inset
  8716. .
  8717. \end_layout
  8718. \begin_layout Standard
  8719. \begin_inset Note Note
  8720. status open
  8721. \begin_layout Itemize
  8722. New blood RNA-seq protocol to block reverse transcription of globin genes
  8723. \end_layout
  8724. \begin_layout Itemize
  8725. Blood RNA-seq time course after transplants with/without MSC infusion
  8726. \end_layout
  8727. \end_inset
  8728. \end_layout
  8729. \begin_layout Section
  8730. Results
  8731. \end_layout
  8732. \begin_layout Subsection
  8733. Globin blocking yields a larger and more consistent fraction of useful reads
  8734. \end_layout
  8735. \begin_layout Standard
  8736. \begin_inset ERT
  8737. status open
  8738. \begin_layout Plain Layout
  8739. \backslash
  8740. afterpage{
  8741. \end_layout
  8742. \begin_layout Plain Layout
  8743. \backslash
  8744. begin{landscape}
  8745. \end_layout
  8746. \end_inset
  8747. \end_layout
  8748. \begin_layout Standard
  8749. \begin_inset Float table
  8750. placement p
  8751. wide false
  8752. sideways false
  8753. status collapsed
  8754. \begin_layout Plain Layout
  8755. \align center
  8756. \begin_inset Tabular
  8757. <lyxtabular version="3" rows="4" columns="7">
  8758. <features tabularvalignment="middle">
  8759. <column alignment="center" valignment="top">
  8760. <column alignment="center" valignment="top">
  8761. <column alignment="center" valignment="top">
  8762. <column alignment="center" valignment="top">
  8763. <column alignment="center" valignment="top">
  8764. <column alignment="center" valignment="top">
  8765. <column alignment="center" valignment="top">
  8766. <row>
  8767. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  8768. \begin_inset Text
  8769. \begin_layout Plain Layout
  8770. \end_layout
  8771. \end_inset
  8772. </cell>
  8773. <cell multicolumn="1" alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  8774. \begin_inset Text
  8775. \begin_layout Plain Layout
  8776. \family roman
  8777. \series medium
  8778. \shape up
  8779. \size normal
  8780. \emph off
  8781. \bar no
  8782. \strikeout off
  8783. \xout off
  8784. \uuline off
  8785. \uwave off
  8786. \noun off
  8787. \color none
  8788. Percent of Total Reads
  8789. \end_layout
  8790. \end_inset
  8791. </cell>
  8792. <cell multicolumn="2" alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  8793. \begin_inset Text
  8794. \begin_layout Plain Layout
  8795. \end_layout
  8796. \end_inset
  8797. </cell>
  8798. <cell multicolumn="2" alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  8799. \begin_inset Text
  8800. \begin_layout Plain Layout
  8801. \end_layout
  8802. \end_inset
  8803. </cell>
  8804. <cell multicolumn="2" alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  8805. \begin_inset Text
  8806. \begin_layout Plain Layout
  8807. \end_layout
  8808. \end_inset
  8809. </cell>
  8810. <cell multicolumn="1" alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  8811. \begin_inset Text
  8812. \begin_layout Plain Layout
  8813. \family roman
  8814. \series medium
  8815. \shape up
  8816. \size normal
  8817. \emph off
  8818. \bar no
  8819. \strikeout off
  8820. \xout off
  8821. \uuline off
  8822. \uwave off
  8823. \noun off
  8824. \color none
  8825. Percent of Genic Reads
  8826. \end_layout
  8827. \end_inset
  8828. </cell>
  8829. <cell multicolumn="2" alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  8830. \begin_inset Text
  8831. \begin_layout Plain Layout
  8832. \end_layout
  8833. \end_inset
  8834. </cell>
  8835. </row>
  8836. <row>
  8837. <cell alignment="center" valignment="top" bottomline="true" leftline="true" usebox="none">
  8838. \begin_inset Text
  8839. \begin_layout Plain Layout
  8840. GB
  8841. \end_layout
  8842. \end_inset
  8843. </cell>
  8844. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  8845. \begin_inset Text
  8846. \begin_layout Plain Layout
  8847. \family roman
  8848. \series medium
  8849. \shape up
  8850. \size normal
  8851. \emph off
  8852. \bar no
  8853. \strikeout off
  8854. \xout off
  8855. \uuline off
  8856. \uwave off
  8857. \noun off
  8858. \color none
  8859. Non-globin Reads
  8860. \end_layout
  8861. \end_inset
  8862. </cell>
  8863. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  8864. \begin_inset Text
  8865. \begin_layout Plain Layout
  8866. \family roman
  8867. \series medium
  8868. \shape up
  8869. \size normal
  8870. \emph off
  8871. \bar no
  8872. \strikeout off
  8873. \xout off
  8874. \uuline off
  8875. \uwave off
  8876. \noun off
  8877. \color none
  8878. Globin Reads
  8879. \end_layout
  8880. \end_inset
  8881. </cell>
  8882. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  8883. \begin_inset Text
  8884. \begin_layout Plain Layout
  8885. \family roman
  8886. \series medium
  8887. \shape up
  8888. \size normal
  8889. \emph off
  8890. \bar no
  8891. \strikeout off
  8892. \xout off
  8893. \uuline off
  8894. \uwave off
  8895. \noun off
  8896. \color none
  8897. All Genic Reads
  8898. \end_layout
  8899. \end_inset
  8900. </cell>
  8901. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  8902. \begin_inset Text
  8903. \begin_layout Plain Layout
  8904. \family roman
  8905. \series medium
  8906. \shape up
  8907. \size normal
  8908. \emph off
  8909. \bar no
  8910. \strikeout off
  8911. \xout off
  8912. \uuline off
  8913. \uwave off
  8914. \noun off
  8915. \color none
  8916. All Aligned Reads
  8917. \end_layout
  8918. \end_inset
  8919. </cell>
  8920. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  8921. \begin_inset Text
  8922. \begin_layout Plain Layout
  8923. \family roman
  8924. \series medium
  8925. \shape up
  8926. \size normal
  8927. \emph off
  8928. \bar no
  8929. \strikeout off
  8930. \xout off
  8931. \uuline off
  8932. \uwave off
  8933. \noun off
  8934. \color none
  8935. Non-globin Reads
  8936. \end_layout
  8937. \end_inset
  8938. </cell>
  8939. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  8940. \begin_inset Text
  8941. \begin_layout Plain Layout
  8942. \family roman
  8943. \series medium
  8944. \shape up
  8945. \size normal
  8946. \emph off
  8947. \bar no
  8948. \strikeout off
  8949. \xout off
  8950. \uuline off
  8951. \uwave off
  8952. \noun off
  8953. \color none
  8954. Globin Reads
  8955. \end_layout
  8956. \end_inset
  8957. </cell>
  8958. </row>
  8959. <row>
  8960. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  8961. \begin_inset Text
  8962. \begin_layout Plain Layout
  8963. \family roman
  8964. \series medium
  8965. \shape up
  8966. \size normal
  8967. \emph off
  8968. \bar no
  8969. \strikeout off
  8970. \xout off
  8971. \uuline off
  8972. \uwave off
  8973. \noun off
  8974. \color none
  8975. Yes
  8976. \end_layout
  8977. \end_inset
  8978. </cell>
  8979. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  8980. \begin_inset Text
  8981. \begin_layout Plain Layout
  8982. \family roman
  8983. \series medium
  8984. \shape up
  8985. \size normal
  8986. \emph off
  8987. \bar no
  8988. \strikeout off
  8989. \xout off
  8990. \uuline off
  8991. \uwave off
  8992. \noun off
  8993. \color none
  8994. 50.4% ± 6.82
  8995. \end_layout
  8996. \end_inset
  8997. </cell>
  8998. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  8999. \begin_inset Text
  9000. \begin_layout Plain Layout
  9001. \family roman
  9002. \series medium
  9003. \shape up
  9004. \size normal
  9005. \emph off
  9006. \bar no
  9007. \strikeout off
  9008. \xout off
  9009. \uuline off
  9010. \uwave off
  9011. \noun off
  9012. \color none
  9013. 3.48% ± 2.94
  9014. \end_layout
  9015. \end_inset
  9016. </cell>
  9017. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  9018. \begin_inset Text
  9019. \begin_layout Plain Layout
  9020. \family roman
  9021. \series medium
  9022. \shape up
  9023. \size normal
  9024. \emph off
  9025. \bar no
  9026. \strikeout off
  9027. \xout off
  9028. \uuline off
  9029. \uwave off
  9030. \noun off
  9031. \color none
  9032. 53.9% ± 6.81
  9033. \end_layout
  9034. \end_inset
  9035. </cell>
  9036. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  9037. \begin_inset Text
  9038. \begin_layout Plain Layout
  9039. \family roman
  9040. \series medium
  9041. \shape up
  9042. \size normal
  9043. \emph off
  9044. \bar no
  9045. \strikeout off
  9046. \xout off
  9047. \uuline off
  9048. \uwave off
  9049. \noun off
  9050. \color none
  9051. 89.7% ± 2.40
  9052. \end_layout
  9053. \end_inset
  9054. </cell>
  9055. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  9056. \begin_inset Text
  9057. \begin_layout Plain Layout
  9058. \family roman
  9059. \series medium
  9060. \shape up
  9061. \size normal
  9062. \emph off
  9063. \bar no
  9064. \strikeout off
  9065. \xout off
  9066. \uuline off
  9067. \uwave off
  9068. \noun off
  9069. \color none
  9070. 93.5% ± 5.25
  9071. \end_layout
  9072. \end_inset
  9073. </cell>
  9074. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  9075. \begin_inset Text
  9076. \begin_layout Plain Layout
  9077. \family roman
  9078. \series medium
  9079. \shape up
  9080. \size normal
  9081. \emph off
  9082. \bar no
  9083. \strikeout off
  9084. \xout off
  9085. \uuline off
  9086. \uwave off
  9087. \noun off
  9088. \color none
  9089. 6.49% ± 5.25
  9090. \end_layout
  9091. \end_inset
  9092. </cell>
  9093. </row>
  9094. <row>
  9095. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  9096. \begin_inset Text
  9097. \begin_layout Plain Layout
  9098. \family roman
  9099. \series medium
  9100. \shape up
  9101. \size normal
  9102. \emph off
  9103. \bar no
  9104. \strikeout off
  9105. \xout off
  9106. \uuline off
  9107. \uwave off
  9108. \noun off
  9109. \color none
  9110. No
  9111. \end_layout
  9112. \end_inset
  9113. </cell>
  9114. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  9115. \begin_inset Text
  9116. \begin_layout Plain Layout
  9117. \family roman
  9118. \series medium
  9119. \shape up
  9120. \size normal
  9121. \emph off
  9122. \bar no
  9123. \strikeout off
  9124. \xout off
  9125. \uuline off
  9126. \uwave off
  9127. \noun off
  9128. \color none
  9129. 26.3% ± 8.95
  9130. \end_layout
  9131. \end_inset
  9132. </cell>
  9133. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  9134. \begin_inset Text
  9135. \begin_layout Plain Layout
  9136. \family roman
  9137. \series medium
  9138. \shape up
  9139. \size normal
  9140. \emph off
  9141. \bar no
  9142. \strikeout off
  9143. \xout off
  9144. \uuline off
  9145. \uwave off
  9146. \noun off
  9147. \color none
  9148. 44.6% ± 16.6
  9149. \end_layout
  9150. \end_inset
  9151. </cell>
  9152. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  9153. \begin_inset Text
  9154. \begin_layout Plain Layout
  9155. \family roman
  9156. \series medium
  9157. \shape up
  9158. \size normal
  9159. \emph off
  9160. \bar no
  9161. \strikeout off
  9162. \xout off
  9163. \uuline off
  9164. \uwave off
  9165. \noun off
  9166. \color none
  9167. 70.1% ± 9.38
  9168. \end_layout
  9169. \end_inset
  9170. </cell>
  9171. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  9172. \begin_inset Text
  9173. \begin_layout Plain Layout
  9174. \family roman
  9175. \series medium
  9176. \shape up
  9177. \size normal
  9178. \emph off
  9179. \bar no
  9180. \strikeout off
  9181. \xout off
  9182. \uuline off
  9183. \uwave off
  9184. \noun off
  9185. \color none
  9186. 90.7% ± 5.16
  9187. \end_layout
  9188. \end_inset
  9189. </cell>
  9190. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  9191. \begin_inset Text
  9192. \begin_layout Plain Layout
  9193. \family roman
  9194. \series medium
  9195. \shape up
  9196. \size normal
  9197. \emph off
  9198. \bar no
  9199. \strikeout off
  9200. \xout off
  9201. \uuline off
  9202. \uwave off
  9203. \noun off
  9204. \color none
  9205. 38.8% ± 17.1
  9206. \end_layout
  9207. \end_inset
  9208. </cell>
  9209. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  9210. \begin_inset Text
  9211. \begin_layout Plain Layout
  9212. \family roman
  9213. \series medium
  9214. \shape up
  9215. \size normal
  9216. \emph off
  9217. \bar no
  9218. \strikeout off
  9219. \xout off
  9220. \uuline off
  9221. \uwave off
  9222. \noun off
  9223. \color none
  9224. 61.2% ± 17.1
  9225. \end_layout
  9226. \end_inset
  9227. </cell>
  9228. </row>
  9229. </lyxtabular>
  9230. \end_inset
  9231. \end_layout
  9232. \begin_layout Plain Layout
  9233. \begin_inset Caption Standard
  9234. \begin_layout Plain Layout
  9235. \series bold
  9236. \begin_inset Argument 1
  9237. status collapsed
  9238. \begin_layout Plain Layout
  9239. Fractions of reads mapping to genomic features in GB and non-GB samples.
  9240. \end_layout
  9241. \end_inset
  9242. \begin_inset CommandInset label
  9243. LatexCommand label
  9244. name "tab:Fractions-of-reads"
  9245. \end_inset
  9246. Fractions of reads mapping to genomic features in GB and non-GB samples.
  9247. \series default
  9248. All values are given as mean ± standard deviation.
  9249. \end_layout
  9250. \end_inset
  9251. \end_layout
  9252. \end_inset
  9253. \end_layout
  9254. \begin_layout Standard
  9255. \begin_inset ERT
  9256. status open
  9257. \begin_layout Plain Layout
  9258. \backslash
  9259. end{landscape}
  9260. \end_layout
  9261. \begin_layout Plain Layout
  9262. }
  9263. \end_layout
  9264. \end_inset
  9265. \end_layout
  9266. \begin_layout Standard
  9267. The objective of the present study was to validate a new protocol for deep
  9268. RNA-seq of whole blood drawn into PaxGene tubes from cynomolgus monkeys
  9269. undergoing islet transplantation, with particular focus on minimizing the
  9270. loss of useful sequencing space to uninformative globin reads.
  9271. The details of the analysis with respect to transplant outcomes and the
  9272. impact of mesenchymal stem cell treatment will be reported in a separate
  9273. manuscript (in preparation).
  9274. To focus on the efficacy of our globin blocking protocol, 37 blood samples,
  9275. 16 from pre-transplant and 21 from post-transplant time points, were each
  9276. prepped once with and once without globin blocking oligos, and were then
  9277. sequenced on an Illumina NextSeq500 instrument.
  9278. The number of reads aligning to each gene in the cynomolgus genome was
  9279. counted.
  9280. Table 1 summarizes the distribution of read fractions among the GB and
  9281. non-GB libraries.
  9282. In the libraries with no globin blocking, globin reads made up an average
  9283. of 44.6% of total input reads, while reads assigned to all other genes made
  9284. up an average of 26.3%.
  9285. The remaining reads either aligned to intergenic regions (that include
  9286. long non-coding RNAs) or did not align with any annotated transcripts in
  9287. the current build of the cynomolgus genome.
  9288. In the GB libraries, globin reads made up only 3.48% and reads assigned
  9289. to all other genes increased to 50.4%.
  9290. Thus, globin blocking resulted in a 92.2% reduction in globin reads and
  9291. a 91.6% increase in yield of useful non-globin reads.
  9292. \end_layout
  9293. \begin_layout Standard
  9294. This reduction is not quite as efficient as the previous analysis showed
  9295. for human samples by DeepSAGE (<0.4% globin reads after globin reduction)
  9296. \begin_inset CommandInset citation
  9297. LatexCommand cite
  9298. key "Mastrokolias2012"
  9299. literal "false"
  9300. \end_inset
  9301. .
  9302. Nonetheless, this degree of globin reduction is sufficient to nearly double
  9303. the yield of useful reads.
  9304. Thus, globin blocking cuts the required sequencing effort (and costs) to
  9305. achieve a target coverage depth by almost 50%.
  9306. Consistent with this near doubling of yield, the average difference in
  9307. un-normalized logCPM across all genes between the GB libraries and non-GB
  9308. libraries is approximately 1 (mean = 1.01, median = 1.08), an overall 2-fold
  9309. increase.
  9310. Un-normalized values are used here because the TMM normalization correctly
  9311. identifies this 2-fold difference as biologically irrelevant and removes
  9312. it.
  9313. \end_layout
  9314. \begin_layout Standard
  9315. \begin_inset Float figure
  9316. wide false
  9317. sideways false
  9318. status collapsed
  9319. \begin_layout Plain Layout
  9320. \align center
  9321. \begin_inset Graphics
  9322. filename graphics/Globin Paper/figure1 - globin-fractions.pdf
  9323. lyxscale 50
  9324. width 75col%
  9325. \end_inset
  9326. \end_layout
  9327. \begin_layout Plain Layout
  9328. \begin_inset Caption Standard
  9329. \begin_layout Plain Layout
  9330. \series bold
  9331. \begin_inset Argument 1
  9332. status collapsed
  9333. \begin_layout Plain Layout
  9334. Fraction of genic reads in each sample aligned to non-globin genes, with
  9335. and without globin blocking (GB).
  9336. \end_layout
  9337. \end_inset
  9338. \begin_inset CommandInset label
  9339. LatexCommand label
  9340. name "fig:Fraction-of-genic-reads"
  9341. \end_inset
  9342. Fraction of genic reads in each sample aligned to non-globin genes, with
  9343. and without globin blocking (GB).
  9344. \series default
  9345. All reads in each sequencing library were aligned to the cyno genome, and
  9346. the number of reads uniquely aligning to each gene was counted.
  9347. For each sample, counts were summed separately for all globin genes and
  9348. for the remainder of the genes (non-globin genes), and the fraction of
  9349. genic reads aligned to non-globin genes was computed.
  9350. Each point represents an individual sample.
  9351. Gray + signs indicate the means for globin-blocked libraries and unblocked
  9352. libraries.
  9353. The overall distribution for each group is represented as a notched box
  9354. plots.
  9355. Points are randomly spread vertically to avoid excessive overlapping.
  9356. \end_layout
  9357. \end_inset
  9358. \end_layout
  9359. \end_inset
  9360. \end_layout
  9361. \begin_layout Standard
  9362. Another important aspect is that the standard deviations in Table
  9363. \begin_inset CommandInset ref
  9364. LatexCommand ref
  9365. reference "tab:Fractions-of-reads"
  9366. plural "false"
  9367. caps "false"
  9368. noprefix "false"
  9369. \end_inset
  9370. are uniformly smaller in the GB samples than the non-GB ones, indicating
  9371. much greater consistency of yield.
  9372. This is best seen in the percentage of non-globin reads as a fraction of
  9373. total reads aligned to annotated genes (genic reads).
  9374. For the non-GB samples, this measure ranges from 10.9% to 80.9%, while for
  9375. the GB samples it ranges from 81.9% to 99.9% (Figure
  9376. \begin_inset CommandInset ref
  9377. LatexCommand ref
  9378. reference "fig:Fraction-of-genic-reads"
  9379. plural "false"
  9380. caps "false"
  9381. noprefix "false"
  9382. \end_inset
  9383. ).
  9384. This means that for applications where it is critical that each sample
  9385. achieve a specified minimum coverage in order to provide useful information,
  9386. it would be necessary to budget up to 10 times the sequencing depth per
  9387. sample without globin blocking, even though the average yield improvement
  9388. for globin blocking is only 2-fold, because every sample has a chance of
  9389. being 90% globin and 10% useful reads.
  9390. Hence, the more consistent behavior of GB samples makes planning an experiment
  9391. easier and more efficient because it eliminates the need to over-sequence
  9392. every sample in order to guard against the worst case of a high-globin
  9393. fraction.
  9394. \end_layout
  9395. \begin_layout Subsection
  9396. Globin blocking lowers the noise floor and allows detection of about 2000
  9397. more low-expression genes
  9398. \end_layout
  9399. \begin_layout Standard
  9400. \begin_inset Flex TODO Note (inline)
  9401. status open
  9402. \begin_layout Plain Layout
  9403. Remove redundant titles from figures
  9404. \end_layout
  9405. \end_inset
  9406. \end_layout
  9407. \begin_layout Standard
  9408. \begin_inset Float figure
  9409. wide false
  9410. sideways false
  9411. status collapsed
  9412. \begin_layout Plain Layout
  9413. \align center
  9414. \begin_inset Graphics
  9415. filename graphics/Globin Paper/figure2 - aveLogCPM-colored.pdf
  9416. lyxscale 50
  9417. height 60theight%
  9418. \end_inset
  9419. \end_layout
  9420. \begin_layout Plain Layout
  9421. \begin_inset Caption Standard
  9422. \begin_layout Plain Layout
  9423. \series bold
  9424. \begin_inset Argument 1
  9425. status collapsed
  9426. \begin_layout Plain Layout
  9427. Distributions of average group gene abundances when normalized separately
  9428. or together.
  9429. \end_layout
  9430. \end_inset
  9431. \begin_inset CommandInset label
  9432. LatexCommand label
  9433. name "fig:logcpm-dists"
  9434. \end_inset
  9435. Distributions of average group gene abundances when normalized separately
  9436. or together.
  9437. \series default
  9438. All reads in each sequencing library were aligned to the cyno genome, and
  9439. the number of reads uniquely aligning to each gene was counted.
  9440. Genes with zero counts in all libraries were discarded.
  9441. Libraries were normalized using the TMM method.
  9442. Libraries were split into globin-blocked (GB) and non-GB groups and the
  9443. average abundance for each gene in both groups, measured in log2 counts
  9444. per million reads counted, was computed using the aveLogCPM function.
  9445. The distribution of average gene logCPM values was plotted for both groups
  9446. using a kernel density plot to approximate a continuous distribution.
  9447. The logCPM GB distributions are marked in red, non-GB in blue.
  9448. The black vertical line denotes the chosen detection threshold of -1.
  9449. Top panel: Libraries were split into GB and non-GB groups first and normalized
  9450. separately.
  9451. Bottom panel: Libraries were all normalized together first and then split
  9452. into groups.
  9453. \end_layout
  9454. \end_inset
  9455. \end_layout
  9456. \begin_layout Plain Layout
  9457. \end_layout
  9458. \end_inset
  9459. \end_layout
  9460. \begin_layout Standard
  9461. Since globin blocking yields more usable sequencing depth, it should also
  9462. allow detection of more genes at any given threshold.
  9463. When we looked at the distribution of average normalized logCPM values
  9464. across all libraries for genes with at least one read assigned to them,
  9465. we observed the expected bimodal distribution, with a high-abundance "signal"
  9466. peak representing detected genes and a low-abundance "noise" peak representing
  9467. genes whose read count did not rise above the noise floor (Figure
  9468. \begin_inset CommandInset ref
  9469. LatexCommand ref
  9470. reference "fig:logcpm-dists"
  9471. plural "false"
  9472. caps "false"
  9473. noprefix "false"
  9474. \end_inset
  9475. ).
  9476. Consistent with the 2-fold increase in raw counts assigned to non-globin
  9477. genes, the signal peak for GB samples is shifted to the right relative
  9478. to the non-GB signal peak.
  9479. When all the samples are normalized together, this difference is normalized
  9480. out, lining up the signal peaks, and this reveals that, as expected, the
  9481. noise floor for the GB samples is about 2-fold lower.
  9482. This greater separation between signal and noise peaks in the GB samples
  9483. means that low-expression genes should be more easily detected and more
  9484. precisely quantified than in the non-GB samples.
  9485. \end_layout
  9486. \begin_layout Standard
  9487. \begin_inset Float figure
  9488. wide false
  9489. sideways false
  9490. status collapsed
  9491. \begin_layout Plain Layout
  9492. \align center
  9493. \begin_inset Graphics
  9494. filename graphics/Globin Paper/figure3 - detection.pdf
  9495. lyxscale 50
  9496. width 70col%
  9497. \end_inset
  9498. \end_layout
  9499. \begin_layout Plain Layout
  9500. \begin_inset Caption Standard
  9501. \begin_layout Plain Layout
  9502. \series bold
  9503. \begin_inset Argument 1
  9504. status collapsed
  9505. \begin_layout Plain Layout
  9506. Gene detections as a function of abundance thresholds in globin-blocked
  9507. (GB) and non-GB samples.
  9508. \end_layout
  9509. \end_inset
  9510. \begin_inset CommandInset label
  9511. LatexCommand label
  9512. name "fig:Gene-detections"
  9513. \end_inset
  9514. Gene detections as a function of abundance thresholds in globin-blocked
  9515. (GB) and non-GB samples.
  9516. \series default
  9517. Average abundance (logCPM,
  9518. \begin_inset Formula $\log_{2}$
  9519. \end_inset
  9520. counts per million reads counted) was computed by separate group normalization
  9521. as described in Figure
  9522. \begin_inset CommandInset ref
  9523. LatexCommand ref
  9524. reference "fig:logcpm-dists"
  9525. plural "false"
  9526. caps "false"
  9527. noprefix "false"
  9528. \end_inset
  9529. for both the GB and non-GB groups, as well as for all samples considered
  9530. as one large group.
  9531. For each every integer threshold from -2 to 3, the number of genes detected
  9532. at or above that logCPM threshold was plotted for each group.
  9533. \end_layout
  9534. \end_inset
  9535. \end_layout
  9536. \begin_layout Plain Layout
  9537. \end_layout
  9538. \end_inset
  9539. \end_layout
  9540. \begin_layout Standard
  9541. Based on these distributions, we selected a detection threshold of -1, which
  9542. is approximately the leftmost edge of the trough between the signal and
  9543. noise peaks.
  9544. This represents the most liberal possible detection threshold that doesn't
  9545. call substantial numbers of noise genes as detected.
  9546. Among the full dataset, 13429 genes were detected at this threshold, and
  9547. 22276 were not.
  9548. When considering the GB libraries and non-GB libraries separately and re-comput
  9549. ing normalization factors independently within each group, 14535 genes were
  9550. detected in the GB libraries while only 12460 were detected in the non-GB
  9551. libraries.
  9552. Thus, GB allowed the detection of 2000 extra genes that were buried under
  9553. the noise floor without GB.
  9554. This pattern of at least 2000 additional genes detected with GB was also
  9555. consistent across a wide range of possible detection thresholds, from -2
  9556. to 3 (see Figure
  9557. \begin_inset CommandInset ref
  9558. LatexCommand ref
  9559. reference "fig:Gene-detections"
  9560. plural "false"
  9561. caps "false"
  9562. noprefix "false"
  9563. \end_inset
  9564. ).
  9565. \end_layout
  9566. \begin_layout Subsection
  9567. Globin blocking does not add significant additional noise or decrease sample
  9568. quality
  9569. \end_layout
  9570. \begin_layout Standard
  9571. One potential worry is that the globin blocking protocol could perturb the
  9572. levels of non-globin genes.
  9573. There are two kinds of possible perturbations: systematic and random.
  9574. The former is not a major concern for detection of differential expression,
  9575. since a 2-fold change in every sample has no effect on the relative fold
  9576. change between samples.
  9577. In contrast, random perturbations would increase the noise and obscure
  9578. the signal in the dataset, reducing the capacity to detect differential
  9579. expression.
  9580. \end_layout
  9581. \begin_layout Standard
  9582. \begin_inset Float figure
  9583. wide false
  9584. sideways false
  9585. status collapsed
  9586. \begin_layout Plain Layout
  9587. \align center
  9588. \begin_inset Graphics
  9589. filename graphics/Globin Paper/figure4 - maplot-colored.pdf
  9590. lyxscale 50
  9591. width 60col%
  9592. groupId colwidth
  9593. \end_inset
  9594. \end_layout
  9595. \begin_layout Plain Layout
  9596. \begin_inset Caption Standard
  9597. \begin_layout Plain Layout
  9598. \begin_inset Argument 1
  9599. status collapsed
  9600. \begin_layout Plain Layout
  9601. MA plot showing effects of globin blocking on each gene's abundance.
  9602. \end_layout
  9603. \end_inset
  9604. \begin_inset CommandInset label
  9605. LatexCommand label
  9606. name "fig:MA-plot"
  9607. \end_inset
  9608. \series bold
  9609. MA plot showing effects of globin blocking on each gene's abundance.
  9610. \series default
  9611. All libraries were normalized together as described in Figure
  9612. \begin_inset CommandInset ref
  9613. LatexCommand ref
  9614. reference "fig:logcpm-dists"
  9615. plural "false"
  9616. caps "false"
  9617. noprefix "false"
  9618. \end_inset
  9619. , and genes with an average logCPM below -1 were filtered out.
  9620. Each remaining gene was tested for differential abundance with respect
  9621. to globin blocking (GB) using edgeR’s quasi-likelihod F-test, fitting a
  9622. negative binomial generalized linear model to table of read counts in each
  9623. library.
  9624. For each gene, edgeR reported average abundance (logCPM),
  9625. \begin_inset Formula $\log_{2}$
  9626. \end_inset
  9627. fold change (logFC), p-value, and Benjamini-Hochberg adjusted false discovery
  9628. rate (FDR).
  9629. Each gene's logFC was plotted against its logCPM, colored by FDR.
  9630. Red points are significant at ≤10% FDR, and blue are not significant at
  9631. that threshold.
  9632. The alpha and beta globin genes targeted for blocking are marked with large
  9633. triangles, while all other genes are represented as small points.
  9634. \end_layout
  9635. \end_inset
  9636. \end_layout
  9637. \begin_layout Plain Layout
  9638. \end_layout
  9639. \end_inset
  9640. \end_layout
  9641. \begin_layout Standard
  9642. \begin_inset Flex TODO Note (inline)
  9643. status open
  9644. \begin_layout Plain Layout
  9645. Standardize on
  9646. \begin_inset Quotes eld
  9647. \end_inset
  9648. log2
  9649. \begin_inset Quotes erd
  9650. \end_inset
  9651. notation
  9652. \end_layout
  9653. \end_inset
  9654. \end_layout
  9655. \begin_layout Standard
  9656. The data do indeed show small systematic perturbations in gene levels (Figure
  9657. \begin_inset CommandInset ref
  9658. LatexCommand ref
  9659. reference "fig:MA-plot"
  9660. plural "false"
  9661. caps "false"
  9662. noprefix "false"
  9663. \end_inset
  9664. ).
  9665. Other than the 3 designated alpha and beta globin genes, two other genes
  9666. stand out as having especially large negative log fold changes: HBD and
  9667. LOC1021365.
  9668. HBD, delta globin, is most likely targeted by the blocking oligos due to
  9669. high sequence homology with the other globin genes.
  9670. LOC1021365 is the aforementioned ncRNA that is reverse-complementary to
  9671. one of the alpha-like genes and that would be expected to be removed during
  9672. the globin blocking step.
  9673. All other genes appear in a cluster centered vertically at 0, and the vast
  9674. majority of genes in this cluster show an absolute log2(FC) of 0.5 or less.
  9675. Nevertheless, many of these small perturbations are still statistically
  9676. significant, indicating that the globin blocking oligos likely cause very
  9677. small but non-zero systematic perturbations in measured gene expression
  9678. levels.
  9679. \end_layout
  9680. \begin_layout Standard
  9681. \begin_inset Float figure
  9682. wide false
  9683. sideways false
  9684. status collapsed
  9685. \begin_layout Plain Layout
  9686. \align center
  9687. \begin_inset Graphics
  9688. filename graphics/Globin Paper/figure5 - corrplot.pdf
  9689. lyxscale 50
  9690. width 70col%
  9691. \end_inset
  9692. \end_layout
  9693. \begin_layout Plain Layout
  9694. \begin_inset Caption Standard
  9695. \begin_layout Plain Layout
  9696. \series bold
  9697. \begin_inset Argument 1
  9698. status collapsed
  9699. \begin_layout Plain Layout
  9700. Comparison of inter-sample gene abundance correlations with and without
  9701. globin blocking.
  9702. \end_layout
  9703. \end_inset
  9704. \begin_inset CommandInset label
  9705. LatexCommand label
  9706. name "fig:gene-abundance-correlations"
  9707. \end_inset
  9708. Comparison of inter-sample gene abundance correlations with and without
  9709. globin blocking (GB).
  9710. \series default
  9711. All libraries were normalized together as described in Figure 2, and genes
  9712. with an average abundance (logCPM, log2 counts per million reads counted)
  9713. less than -1 were filtered out.
  9714. Each gene’s logCPM was computed in each library using the edgeR cpm function.
  9715. For each pair of biological samples, the Pearson correlation between those
  9716. samples' GB libraries was plotted against the correlation between the same
  9717. samples’ non-GB libraries.
  9718. Each point represents an unique pair of samples.
  9719. The solid gray line shows a quantile-quantile plot of distribution of GB
  9720. correlations vs.
  9721. that of non-GB correlations.
  9722. The thin dashed line is the identity line, provided for reference.
  9723. \end_layout
  9724. \end_inset
  9725. \end_layout
  9726. \begin_layout Plain Layout
  9727. \end_layout
  9728. \end_inset
  9729. \end_layout
  9730. \begin_layout Standard
  9731. To evaluate the possibility of globin blocking causing random perturbations
  9732. and reducing sample quality, we computed the Pearson correlation between
  9733. logCPM values for every pair of samples with and without GB and plotted
  9734. them against each other (Figure
  9735. \begin_inset CommandInset ref
  9736. LatexCommand ref
  9737. reference "fig:gene-abundance-correlations"
  9738. plural "false"
  9739. caps "false"
  9740. noprefix "false"
  9741. \end_inset
  9742. ).
  9743. The plot indicated that the GB libraries have higher sample-to-sample correlati
  9744. ons than the non-GB libraries.
  9745. Parametric and nonparametric tests for differences between the correlations
  9746. with and without GB both confirmed that this difference was highly significant
  9747. (2-sided paired t-test: t = 37.2, df = 665, P ≪ 2.2e-16; 2-sided Wilcoxon
  9748. sign-rank test: V = 2195, P ≪ 2.2e-16).
  9749. Performing the same tests on the Spearman correlations gave the same conclusion
  9750. (t-test: t = 26.8, df = 665, P ≪ 2.2e-16; sign-rank test: V = 8781, P ≪ 2.2e-16).
  9751. The edgeR package was used to compute the overall biological coefficient
  9752. of variation (BCV) for GB and non-GB libraries, and found that globin blocking
  9753. resulted in a negligible increase in the BCV (0.417 with GB vs.
  9754. 0.400 without).
  9755. The near equality of the BCVs for both sets indicates that the higher correlati
  9756. ons in the GB libraries are most likely a result of the increased yield
  9757. of useful reads, which reduces the contribution of Poisson counting uncertainty
  9758. to the overall variance of the logCPM values
  9759. \begin_inset CommandInset citation
  9760. LatexCommand cite
  9761. key "McCarthy2012"
  9762. literal "false"
  9763. \end_inset
  9764. .
  9765. This improves the precision of expression measurements and more than offsets
  9766. the negligible increase in BCV.
  9767. \end_layout
  9768. \begin_layout Subsection
  9769. More differentially expressed genes are detected with globin blocking
  9770. \end_layout
  9771. \begin_layout Standard
  9772. \begin_inset Float table
  9773. wide false
  9774. sideways false
  9775. status collapsed
  9776. \begin_layout Plain Layout
  9777. \align center
  9778. \begin_inset Tabular
  9779. <lyxtabular version="3" rows="5" columns="5">
  9780. <features tabularvalignment="middle">
  9781. <column alignment="center" valignment="top">
  9782. <column alignment="center" valignment="top">
  9783. <column alignment="center" valignment="top">
  9784. <column alignment="center" valignment="top">
  9785. <column alignment="center" valignment="top">
  9786. <row>
  9787. <cell alignment="center" valignment="top" usebox="none">
  9788. \begin_inset Text
  9789. \begin_layout Plain Layout
  9790. \end_layout
  9791. \end_inset
  9792. </cell>
  9793. <cell alignment="center" valignment="top" usebox="none">
  9794. \begin_inset Text
  9795. \begin_layout Plain Layout
  9796. \end_layout
  9797. \end_inset
  9798. </cell>
  9799. <cell multicolumn="1" alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  9800. \begin_inset Text
  9801. \begin_layout Plain Layout
  9802. \series bold
  9803. No Globin Blocking
  9804. \end_layout
  9805. \end_inset
  9806. </cell>
  9807. <cell multicolumn="2" alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  9808. \begin_inset Text
  9809. \begin_layout Plain Layout
  9810. \end_layout
  9811. \end_inset
  9812. </cell>
  9813. <cell multicolumn="2" alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  9814. \begin_inset Text
  9815. \begin_layout Plain Layout
  9816. \end_layout
  9817. \end_inset
  9818. </cell>
  9819. </row>
  9820. <row>
  9821. <cell alignment="center" valignment="top" usebox="none">
  9822. \begin_inset Text
  9823. \begin_layout Plain Layout
  9824. \end_layout
  9825. \end_inset
  9826. </cell>
  9827. <cell alignment="center" valignment="top" usebox="none">
  9828. \begin_inset Text
  9829. \begin_layout Plain Layout
  9830. \end_layout
  9831. \end_inset
  9832. </cell>
  9833. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  9834. \begin_inset Text
  9835. \begin_layout Plain Layout
  9836. \series bold
  9837. Up
  9838. \end_layout
  9839. \end_inset
  9840. </cell>
  9841. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  9842. \begin_inset Text
  9843. \begin_layout Plain Layout
  9844. \series bold
  9845. NS
  9846. \end_layout
  9847. \end_inset
  9848. </cell>
  9849. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  9850. \begin_inset Text
  9851. \begin_layout Plain Layout
  9852. \series bold
  9853. Down
  9854. \end_layout
  9855. \end_inset
  9856. </cell>
  9857. </row>
  9858. <row>
  9859. <cell multirow="3" alignment="center" valignment="middle" topline="true" bottomline="true" leftline="true" usebox="none">
  9860. \begin_inset Text
  9861. \begin_layout Plain Layout
  9862. \series bold
  9863. Globin-Blocking
  9864. \end_layout
  9865. \end_inset
  9866. </cell>
  9867. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  9868. \begin_inset Text
  9869. \begin_layout Plain Layout
  9870. \series bold
  9871. Up
  9872. \end_layout
  9873. \end_inset
  9874. </cell>
  9875. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  9876. \begin_inset Text
  9877. \begin_layout Plain Layout
  9878. \family roman
  9879. \series medium
  9880. \shape up
  9881. \size normal
  9882. \emph off
  9883. \bar no
  9884. \strikeout off
  9885. \xout off
  9886. \uuline off
  9887. \uwave off
  9888. \noun off
  9889. \color none
  9890. 231
  9891. \end_layout
  9892. \end_inset
  9893. </cell>
  9894. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  9895. \begin_inset Text
  9896. \begin_layout Plain Layout
  9897. \family roman
  9898. \series medium
  9899. \shape up
  9900. \size normal
  9901. \emph off
  9902. \bar no
  9903. \strikeout off
  9904. \xout off
  9905. \uuline off
  9906. \uwave off
  9907. \noun off
  9908. \color none
  9909. 515
  9910. \end_layout
  9911. \end_inset
  9912. </cell>
  9913. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  9914. \begin_inset Text
  9915. \begin_layout Plain Layout
  9916. \family roman
  9917. \series medium
  9918. \shape up
  9919. \size normal
  9920. \emph off
  9921. \bar no
  9922. \strikeout off
  9923. \xout off
  9924. \uuline off
  9925. \uwave off
  9926. \noun off
  9927. \color none
  9928. 2
  9929. \end_layout
  9930. \end_inset
  9931. </cell>
  9932. </row>
  9933. <row>
  9934. <cell multirow="4" alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  9935. \begin_inset Text
  9936. \begin_layout Plain Layout
  9937. \end_layout
  9938. \end_inset
  9939. </cell>
  9940. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  9941. \begin_inset Text
  9942. \begin_layout Plain Layout
  9943. \series bold
  9944. NS
  9945. \end_layout
  9946. \end_inset
  9947. </cell>
  9948. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  9949. \begin_inset Text
  9950. \begin_layout Plain Layout
  9951. \family roman
  9952. \series medium
  9953. \shape up
  9954. \size normal
  9955. \emph off
  9956. \bar no
  9957. \strikeout off
  9958. \xout off
  9959. \uuline off
  9960. \uwave off
  9961. \noun off
  9962. \color none
  9963. 160
  9964. \end_layout
  9965. \end_inset
  9966. </cell>
  9967. <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
  9968. \begin_inset Text
  9969. \begin_layout Plain Layout
  9970. \family roman
  9971. \series medium
  9972. \shape up
  9973. \size normal
  9974. \emph off
  9975. \bar no
  9976. \strikeout off
  9977. \xout off
  9978. \uuline off
  9979. \uwave off
  9980. \noun off
  9981. \color none
  9982. 11235
  9983. \end_layout
  9984. \end_inset
  9985. </cell>
  9986. <cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
  9987. \begin_inset Text
  9988. \begin_layout Plain Layout
  9989. \family roman
  9990. \series medium
  9991. \shape up
  9992. \size normal
  9993. \emph off
  9994. \bar no
  9995. \strikeout off
  9996. \xout off
  9997. \uuline off
  9998. \uwave off
  9999. \noun off
  10000. \color none
  10001. 136
  10002. \end_layout
  10003. \end_inset
  10004. </cell>
  10005. </row>
  10006. <row>
  10007. <cell multirow="4" alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  10008. \begin_inset Text
  10009. \begin_layout Plain Layout
  10010. \end_layout
  10011. \end_inset
  10012. </cell>
  10013. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  10014. \begin_inset Text
  10015. \begin_layout Plain Layout
  10016. \series bold
  10017. Down
  10018. \end_layout
  10019. \end_inset
  10020. </cell>
  10021. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  10022. \begin_inset Text
  10023. \begin_layout Plain Layout
  10024. \family roman
  10025. \series medium
  10026. \shape up
  10027. \size normal
  10028. \emph off
  10029. \bar no
  10030. \strikeout off
  10031. \xout off
  10032. \uuline off
  10033. \uwave off
  10034. \noun off
  10035. \color none
  10036. 0
  10037. \end_layout
  10038. \end_inset
  10039. </cell>
  10040. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
  10041. \begin_inset Text
  10042. \begin_layout Plain Layout
  10043. \family roman
  10044. \series medium
  10045. \shape up
  10046. \size normal
  10047. \emph off
  10048. \bar no
  10049. \strikeout off
  10050. \xout off
  10051. \uuline off
  10052. \uwave off
  10053. \noun off
  10054. \color none
  10055. 548
  10056. \end_layout
  10057. \end_inset
  10058. </cell>
  10059. <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
  10060. \begin_inset Text
  10061. \begin_layout Plain Layout
  10062. \family roman
  10063. \series medium
  10064. \shape up
  10065. \size normal
  10066. \emph off
  10067. \bar no
  10068. \strikeout off
  10069. \xout off
  10070. \uuline off
  10071. \uwave off
  10072. \noun off
  10073. \color none
  10074. 127
  10075. \end_layout
  10076. \end_inset
  10077. </cell>
  10078. </row>
  10079. </lyxtabular>
  10080. \end_inset
  10081. \end_layout
  10082. \begin_layout Plain Layout
  10083. \begin_inset Caption Standard
  10084. \begin_layout Plain Layout
  10085. \series bold
  10086. \begin_inset Argument 1
  10087. status open
  10088. \begin_layout Plain Layout
  10089. Comparison of significantly differentially expressed genes with and without
  10090. globin blocking.
  10091. \end_layout
  10092. \end_inset
  10093. \begin_inset CommandInset label
  10094. LatexCommand label
  10095. name "tab:Comparison-of-significant"
  10096. \end_inset
  10097. Comparison of significantly differentially expressed genes with and without
  10098. globin blocking.
  10099. \series default
  10100. Up, Down: Genes significantly up/down-regulated in post-transplant samples
  10101. relative to pre-transplant samples, with a false discovery rate of 10%
  10102. or less.
  10103. NS: Non-significant genes (false discovery rate greater than 10%).
  10104. \end_layout
  10105. \end_inset
  10106. \end_layout
  10107. \begin_layout Plain Layout
  10108. \end_layout
  10109. \end_inset
  10110. \end_layout
  10111. \begin_layout Standard
  10112. To compare performance on differential gene expression tests, we took subsets
  10113. of both the GB and non-GB libraries with exactly one pre-transplant and
  10114. one post-transplant sample for each animal that had paired samples available
  10115. for analysis (N=7 animals, N=14 samples in each subset).
  10116. The same test for pre- vs.
  10117. post-transplant differential gene expression was performed on the same
  10118. 7 pairs of samples from GB libraries and non-GB libraries, in each case
  10119. using an FDR of 10% as the threshold of significance.
  10120. Out of 12954 genes that passed the detection threshold in both subsets,
  10121. 358 were called significantly differentially expressed in the same direction
  10122. in both sets; 1063 were differentially expressed in the GB set only; 296
  10123. were differentially expressed in the non-GB set only; 2 genes were called
  10124. significantly up in the GB set but significantly down in the non-GB set;
  10125. and the remaining 11235 were not called differentially expressed in either
  10126. set.
  10127. These data are summarized in Table
  10128. \begin_inset CommandInset ref
  10129. LatexCommand ref
  10130. reference "tab:Comparison-of-significant"
  10131. plural "false"
  10132. caps "false"
  10133. noprefix "false"
  10134. \end_inset
  10135. .
  10136. The differences in BCV calculated by EdgeR for these subsets of samples
  10137. were negligible (BCV = 0.302 for GB and 0.297 for non-GB).
  10138. \end_layout
  10139. \begin_layout Standard
  10140. The key point is that the GB data results in substantially more differentially
  10141. expressed calls than the non-GB data.
  10142. Since there is no gold standard for this dataset, it is impossible to be
  10143. certain whether this is due to under-calling of differential expression
  10144. in the non-GB samples or over-calling in the GB samples.
  10145. However, given that both datasets are derived from the same biological
  10146. samples and have nearly equal BCVs, it is more likely that the larger number
  10147. of DE calls in the GB samples are genuine detections that were enabled
  10148. by the higher sequencing depth and measurement precision of the GB samples.
  10149. Note that the same set of genes was considered in both subsets, so the
  10150. larger number of differentially expressed gene calls in the GB data set
  10151. reflects a greater sensitivity to detect significant differential gene
  10152. expression and not simply the larger total number of detected genes in
  10153. GB samples described earlier.
  10154. \end_layout
  10155. \begin_layout Section
  10156. Discussion
  10157. \end_layout
  10158. \begin_layout Standard
  10159. The original experience with whole blood gene expression profiling on DNA
  10160. microarrays demonstrated that the high concentration of globin transcripts
  10161. reduced the sensitivity to detect genes with relatively low expression
  10162. levels, in effect, significantly reducing the sensitivity.
  10163. To address this limitation, commercial protocols for globin reduction were
  10164. developed based on strategies to block globin transcript amplification
  10165. during labeling or physically removing globin transcripts by affinity bead
  10166. methods
  10167. \begin_inset CommandInset citation
  10168. LatexCommand cite
  10169. key "Winn2010"
  10170. literal "false"
  10171. \end_inset
  10172. .
  10173. More recently, using the latest generation of labeling protocols and arrays,
  10174. it was determined that globin reduction was no longer necessary to obtain
  10175. sufficient sensitivity to detect differential transcript expression
  10176. \begin_inset CommandInset citation
  10177. LatexCommand cite
  10178. key "NuGEN2010"
  10179. literal "false"
  10180. \end_inset
  10181. .
  10182. However, we are not aware of any publications using these currently available
  10183. protocols the with latest generation of microarrays that actually compare
  10184. the detection sensitivity with and without globin reduction.
  10185. However, in practice this has now been adopted generally primarily driven
  10186. by concerns for cost control.
  10187. The main objective of our work was to directly test the impact of globin
  10188. gene transcripts and a new globin blocking protocol for application to
  10189. the newest generation of differential gene expression profiling determined
  10190. using next generation sequencing.
  10191. \end_layout
  10192. \begin_layout Standard
  10193. The challenge of doing global gene expression profiling in cynomolgus monkeys
  10194. is that the current available arrays were never designed to comprehensively
  10195. cover this genome and have not been updated since the first assemblies
  10196. of the cynomolgus genome were published.
  10197. Therefore, we determined that the best strategy for peripheral blood profiling
  10198. was to do deep RNA-seq and inform the workflow using the latest available
  10199. genome assembly and annotation
  10200. \begin_inset CommandInset citation
  10201. LatexCommand cite
  10202. key "Wilson2013"
  10203. literal "false"
  10204. \end_inset
  10205. .
  10206. However, it was not immediately clear whether globin reduction was necessary
  10207. for RNA-seq or how much improvement in efficiency or sensitivity to detect
  10208. differential gene expression would be achieved for the added cost and work.
  10209. \end_layout
  10210. \begin_layout Standard
  10211. We only found one report that demonstrated that globin reduction significantly
  10212. improved the effective read yields for sequencing of human peripheral blood
  10213. cell RNA using a DeepSAGE protocol
  10214. \begin_inset CommandInset citation
  10215. LatexCommand cite
  10216. key "Mastrokolias2012"
  10217. literal "false"
  10218. \end_inset
  10219. .
  10220. The approach to DeepSAGE involves two different restriction enzymes that
  10221. purify and then tag small fragments of transcripts at specific locations
  10222. and thus, significantly reduces the complexity of the transcriptome.
  10223. Therefore, we could not determine how DeepSAGE results would translate
  10224. to the common strategy in the field for assaying the entire transcript
  10225. population by whole-transcriptome 3’-end RNA-seq.
  10226. Furthermore, if globin reduction is necessary, we also needed a globin
  10227. reduction method specific to cynomolgus globin sequences that would work
  10228. an organism for which no kit is available off the shelf.
  10229. \end_layout
  10230. \begin_layout Standard
  10231. As mentioned above, the addition of globin blocking oligos has a very small
  10232. impact on measured expression levels of gene expression.
  10233. However, this is a non-issue for the purposes of differential expression
  10234. testing, since a systematic change in a gene in all samples does not affect
  10235. relative expression levels between samples.
  10236. However, we must acknowledge that simple comparisons of gene expression
  10237. data obtained by GB and non-GB protocols are not possible without additional
  10238. normalization.
  10239. \end_layout
  10240. \begin_layout Standard
  10241. More importantly, globin blocking not only nearly doubles the yield of usable
  10242. reads, it also increases inter-sample correlation and sensitivity to detect
  10243. differential gene expression relative to the same set of samples profiled
  10244. without blocking.
  10245. In addition, globin blocking does not add a significant amount of random
  10246. noise to the data.
  10247. Globin blocking thus represents a cost-effective way to squeeze more data
  10248. and statistical power out of the same blood samples and the same amount
  10249. of sequencing.
  10250. In conclusion, globin reduction greatly increases the yield of useful RNA-seq
  10251. reads mapping to the rest of the genome, with minimal perturbations in
  10252. the relative levels of non-globin genes.
  10253. Based on these results, globin transcript reduction using sequence-specific,
  10254. complementary blocking oligonucleotides is recommended for all deep RNA-seq
  10255. of cynomolgus and other nonhuman primate blood samples.
  10256. \end_layout
  10257. \begin_layout Chapter
  10258. Future Directions
  10259. \end_layout
  10260. \begin_layout Standard
  10261. \begin_inset Flex TODO Note (inline)
  10262. status open
  10263. \begin_layout Plain Layout
  10264. Consider putting each chapter's future directions with that chapter instead
  10265. of in a separate one.
  10266. Check instructions to see if this is allowed/appropriate.
  10267. \end_layout
  10268. \end_inset
  10269. \end_layout
  10270. \begin_layout Section*
  10271. Ch2
  10272. \end_layout
  10273. \begin_layout Standard
  10274. The analysis of RNA-seq and ChIP-seq in CD4 T-cells in Chapter 2 is in many
  10275. ways a preliminary study that suggests a multitude of new avenues of investigat
  10276. ion.
  10277. Here we consider a selection of such avenues.
  10278. \end_layout
  10279. \begin_layout Subsection*
  10280. Improving on the effective promoter radius
  10281. \end_layout
  10282. \begin_layout Standard
  10283. This study introduced the concept of an
  10284. \begin_inset Quotes eld
  10285. \end_inset
  10286. effective promoter radius
  10287. \begin_inset Quotes erd
  10288. \end_inset
  10289. specific to each histone mark based on distince from the TSS within which
  10290. an excess of peaks was called for that mark.
  10291. This concept was then used to guide further analyses throughout the study.
  10292. However, while the effective promoter radius was useful in those analyses,
  10293. it is both limited in theory and shown in practice to be a possible oversimplif
  10294. ication.
  10295. First, the effective promoter radii used in this study were chosen based
  10296. on manual inspection of the TSS-to-peak distance distributions in Figure
  10297. \begin_inset CommandInset ref
  10298. LatexCommand ref
  10299. reference "fig:near-promoter-peak-enrich"
  10300. plural "false"
  10301. caps "false"
  10302. noprefix "false"
  10303. \end_inset
  10304. , selecting round numbers of analyst convenience (Table
  10305. \begin_inset CommandInset ref
  10306. LatexCommand ref
  10307. reference "tab:effective-promoter-radius"
  10308. plural "false"
  10309. caps "false"
  10310. noprefix "false"
  10311. \end_inset
  10312. ).
  10313. It would be better to define an algorithm that selects a more precise radius
  10314. based on the features of the graph.
  10315. One possible way to do this would be to randomly rearrange the called peaks
  10316. throughout the genome many (while preserving the distribution of peak widths)
  10317. and re-generate the same plot as in Figure
  10318. \begin_inset CommandInset ref
  10319. LatexCommand ref
  10320. reference "fig:near-promoter-peak-enrich"
  10321. plural "false"
  10322. caps "false"
  10323. noprefix "false"
  10324. \end_inset
  10325. .
  10326. This would yield a better
  10327. \begin_inset Quotes eld
  10328. \end_inset
  10329. background
  10330. \begin_inset Quotes erd
  10331. \end_inset
  10332. distribution that demonstrates the degree of near-TSS enrichment that would
  10333. be expected by random chance.
  10334. The effective promoter radius could be defined as the point where the true
  10335. distribution diverges from the randomized background distribution.
  10336. \end_layout
  10337. \begin_layout Standard
  10338. Furthermore, the above definition of effective promoter radius has the significa
  10339. nt limitation of being based on the peak calling method.
  10340. It is thus very sensitive to the choice of peak caller and significance
  10341. threshold for calling peaks, as well as the degree of saturation in the
  10342. sequencing.
  10343. Calling peaks from ChIP-seq samples with insufficient coverage depth, with
  10344. the wrong peak caller, or with a different significance threshold could
  10345. give a drastically different number of called peaks, and hence a drastically
  10346. different distribution of peak-to-TSS distances.
  10347. To address this, it is desirable to develop a better method of determining
  10348. the effective promoter radius that relies only on the distribution of read
  10349. coverage around the TSS, independent of the peak calling.
  10350. Furthermore, as demonstrated by the upstream-downstream asymmetries observed
  10351. in Figures
  10352. \begin_inset CommandInset ref
  10353. LatexCommand ref
  10354. reference "fig:H3K4me2-neighborhood"
  10355. plural "false"
  10356. caps "false"
  10357. noprefix "false"
  10358. \end_inset
  10359. ,
  10360. \begin_inset CommandInset ref
  10361. LatexCommand ref
  10362. reference "fig:H3K4me3-neighborhood"
  10363. plural "false"
  10364. caps "false"
  10365. noprefix "false"
  10366. \end_inset
  10367. , and
  10368. \begin_inset CommandInset ref
  10369. LatexCommand ref
  10370. reference "fig:H3K27me3-neighborhood"
  10371. plural "false"
  10372. caps "false"
  10373. noprefix "false"
  10374. \end_inset
  10375. , this definition should determine a different radius for the upstream and
  10376. downstream directions.
  10377. At this point, it may be better to rename this concept
  10378. \begin_inset Quotes eld
  10379. \end_inset
  10380. effective promoter extent
  10381. \begin_inset Quotes erd
  10382. \end_inset
  10383. and avoid the word
  10384. \begin_inset Quotes eld
  10385. \end_inset
  10386. radius
  10387. \begin_inset Quotes erd
  10388. \end_inset
  10389. , since a radius implies a symmetry about the TSS that is not supported
  10390. by the data.
  10391. \end_layout
  10392. \begin_layout Standard
  10393. Beyond improving the definition of effective promoter extent, functional
  10394. validation is necessary to show that this measure of near-TSS enrichment
  10395. has biological meaning.
  10396. Figures
  10397. \begin_inset CommandInset ref
  10398. LatexCommand ref
  10399. reference "fig:H3K4me2-neighborhood"
  10400. plural "false"
  10401. caps "false"
  10402. noprefix "false"
  10403. \end_inset
  10404. and
  10405. \begin_inset CommandInset ref
  10406. LatexCommand ref
  10407. reference "fig:H3K4me3-neighborhood"
  10408. plural "false"
  10409. caps "false"
  10410. noprefix "false"
  10411. \end_inset
  10412. already provide a very limited functional validation of the chosen promoter
  10413. extents for H3K4me2 and H3K4me3 by showing that spikes in coverage within
  10414. this region are most strongly correlated with elevated gene expression.
  10415. However, there are other ways to show functional relevance of the promoter
  10416. extent.
  10417. For example, correlations could be computed between read counts in peaks
  10418. nearby gene promoters and the expression level of those genes, and these
  10419. correlations could be plotted against the distance of the peak upstream
  10420. or downstream of the gene's TSS.
  10421. If the promoter extent truly defines a
  10422. \begin_inset Quotes eld
  10423. \end_inset
  10424. sphere of influence
  10425. \begin_inset Quotes erd
  10426. \end_inset
  10427. within which a histone mark is involved with the regulation of a gene,
  10428. then the correlations for peaks within this extent should be significantly
  10429. higher than those further upstream or downstream.
  10430. Peaks within these extents may also be more likely to show differential
  10431. modification than those outside genic regions of the genome.
  10432. \end_layout
  10433. \begin_layout Subsection*
  10434. Post-activation convergence of naive & memory cells
  10435. \end_layout
  10436. \begin_layout Standard
  10437. In this study, a convergence between naive and memory cells was observed
  10438. in both the pattern of gene expression and in epigenetic state of the 3
  10439. histone marks studied.
  10440. \end_layout
  10441. \begin_layout Itemize
  10442. N-to-M convergence deserves further study of some kind
  10443. \end_layout
  10444. \begin_deeper
  10445. \begin_layout Itemize
  10446. maybe serial activation & rest cycles for naive and memory, showing a cyclical
  10447. pattern returning to the same state again and again after the first activation
  10448. \end_layout
  10449. \end_deeper
  10450. \begin_layout Itemize
  10451. Study other epigenetic marks in more contexts, including looking for similar
  10452. convergence patterns.
  10453. Use MOFA to identify coordinated patterns.
  10454. \end_layout
  10455. \begin_deeper
  10456. \begin_layout Itemize
  10457. DNA methylation, histone marks, chromatin accessibility & conformation in
  10458. CD4 T-cells
  10459. \end_layout
  10460. \begin_layout Itemize
  10461. Also look at other types of lymphocytes: CD8 T-cells, B-cells, NK cells
  10462. \end_layout
  10463. \end_deeper
  10464. \begin_layout Subsection*
  10465. Promoter positional coverage: follow up on hints of interesting patterns
  10466. \end_layout
  10467. \begin_layout Itemize
  10468. Also find better normalizations: maybe borrow from MACS/SICER background
  10469. correction methods?
  10470. \end_layout
  10471. \begin_layout Itemize
  10472. For H3K4, define polar coordinates based on PC1 & 2: R = peak size, Theta
  10473. = peak position.
  10474. Then correlate with expression.
  10475. \end_layout
  10476. \begin_layout Itemize
  10477. Current analysis only at Day 0.
  10478. Need to study across time points.
  10479. \end_layout
  10480. \begin_layout Subsection*
  10481. H3K4me correlation
  10482. \end_layout
  10483. \begin_layout Standard
  10484. The high correlation between coverage depth observed between H3K4me2 and
  10485. H3K4me3 is both expected and unexpected.
  10486. Since both marks are associated with elevated gene transcription, a positive
  10487. correlation between them is not surprising.
  10488. However, these two marks represent different post-translational modifications
  10489. of the
  10490. \emph on
  10491. same
  10492. \emph default
  10493. lysine residue on the histone H3 polypeptide, which means that they cannot
  10494. both be present on the same H3 subunit.
  10495. Thus, the high correlation between them has several potential explanations.
  10496. One possible reason is cell population heterogeneity: perhaps some genomic
  10497. loci are frequently marked with H3K4me2 in some cells, while in other cells
  10498. the same loci are marked with H3K4me3.
  10499. Another possibility is allele-specific modifications: the loci are marked
  10500. in each diploid cell with H3K4me2 on one allele and H3K4me3 on the other
  10501. allele.
  10502. Lastly, since each histone octamer contains 2 H3 subunits, it is possible
  10503. that having one H3K4me2 mark and one H3K4me3 mark on a given histone octamer
  10504. represents a distinct epigenetic state with a different function than either
  10505. double H3K4me2 or double H3K4me3.
  10506. \end_layout
  10507. \begin_layout Standard
  10508. These three hypotheses could be disentangled by single-cell ChIP-seq.
  10509. If the correlation between these two histone marks persists even within
  10510. the reads for each individual cell, then cell population heterogeneity
  10511. cannot explain the correlation.
  10512. Allele-specific modification can be tested for by looking at the correlation
  10513. between read coverage of the two histone marks at heterozygous loci.
  10514. If the correlation between read counts for opposite loci is low, then this
  10515. is consistent with allele-specific modification.
  10516. Finally if the modifications do not separate by either cell or allele,
  10517. the colocation of these two marks is most likely occurring at the level
  10518. of individual histones, with the heterogenously modified histone representing
  10519. a distinct state.
  10520. \end_layout
  10521. \begin_layout Standard
  10522. However, another experiment would be required to show direct evidence of
  10523. such a heterogeneously modified state.
  10524. Specifically a
  10525. \begin_inset Quotes eld
  10526. \end_inset
  10527. double ChIP
  10528. \begin_inset Quotes erd
  10529. \end_inset
  10530. experiment would need to be performed, where the input DNA is first subjected
  10531. to an immunoprecipitation pulldown from the anti-H3K4me2 antibody, and
  10532. then the enriched material is collected, with proteins still bound, and
  10533. immunoprecipitated
  10534. \emph on
  10535. again
  10536. \emph default
  10537. using the anti-H3K4me3 antibody.
  10538. If this yields significant numbers of non-artifactual reads in the same
  10539. regions as the individual pulldowns of the two marks, this is strong evidence
  10540. that the two marks are occurring on opposite H3 subunits of the same histones.
  10541. \end_layout
  10542. \begin_layout Standard
  10543. \begin_inset Flex TODO Note (inline)
  10544. status open
  10545. \begin_layout Plain Layout
  10546. Try to see if double ChIP-seq is actually feasible, and if not, come up
  10547. with some other idea for directly detecting the mixed mod state.
  10548. Oh! Actually ChIP-seq isn't required, only double ChIP followed by quantificati
  10549. on.
  10550. That's one possible angle.
  10551. \end_layout
  10552. \end_inset
  10553. \end_layout
  10554. \begin_layout Section*
  10555. Ch3
  10556. \end_layout
  10557. \begin_layout Itemize
  10558. Use CV or bootstrap to better evaluate classifiers
  10559. \end_layout
  10560. \begin_layout Itemize
  10561. fRMAtools could be adapted to not require equal-sized groups
  10562. \end_layout
  10563. \begin_layout Section*
  10564. Ch4
  10565. \end_layout
  10566. \begin_layout Standard
  10567. \begin_inset Flex TODO Note (inline)
  10568. status open
  10569. \begin_layout Plain Layout
  10570. I've already done a good bit of work outside just this globin blocking thing,
  10571. so I'm not sure what to put for future directions.
  10572. Does it inculde the other stuff I've done but not published?
  10573. \end_layout
  10574. \end_inset
  10575. \end_layout
  10576. \begin_layout Standard
  10577. \begin_inset ERT
  10578. status collapsed
  10579. \begin_layout Plain Layout
  10580. % Call it "References" instead of "Bibliography"
  10581. \end_layout
  10582. \begin_layout Plain Layout
  10583. \backslash
  10584. renewcommand{
  10585. \backslash
  10586. bibname}{References}
  10587. \end_layout
  10588. \end_inset
  10589. \end_layout
  10590. \begin_layout Standard
  10591. \begin_inset CommandInset bibtex
  10592. LatexCommand bibtex
  10593. btprint "btPrintCited"
  10594. bibfiles "code-refs,refs-PROCESSED"
  10595. options "bibtotoc,unsrt"
  10596. \end_inset
  10597. \end_layout
  10598. \begin_layout Standard
  10599. \begin_inset Flex TODO Note (inline)
  10600. status open
  10601. \begin_layout Plain Layout
  10602. Check bib entry formatting & sort order
  10603. \end_layout
  10604. \end_inset
  10605. \end_layout
  10606. \begin_layout Standard
  10607. \begin_inset Flex TODO Note (inline)
  10608. status open
  10609. \begin_layout Plain Layout
  10610. Check in-text citation format.
  10611. Probably don't just want [1], [2], etc.
  10612. \end_layout
  10613. \end_inset
  10614. \end_layout
  10615. \end_body
  10616. \end_document