improved regulatory element prediction based on tissue-specific local epigenomic signatures 2017 yupeng he资料.pdf

该【improved regulatory element prediction based on tissue-specific local epigenomic signatures 2017 yupeng he资料 】是由【玉柱儿】上传分享，beplayapp体育下载一共【8】页，该beplayapp体育下载可以免费在线阅读，需要了解更多关于【improved regulatory element prediction based on tissue-specific local epigenomic signatures 2017 yupeng he资料 】的内容，可以使用beplayapp体育下载的站内搜索功能，选择自己适合的beplayapp体育下载，以下文字是截取该文章内的部分文字，如需要获得完整电子版，请下载此beplayapp体育下载到您的设备，方便您编辑和打印。ImprovedregulatoryelementpredictionbasedonPNASPLUStissue-specificlocalepigenomicsignaturesYupengHea,b,,,,,AhYoungLeec,YinShene,f,AxelViseld,g,h,,g,BingRenc,i,,j,1aGenomicAnalysisLaboratory,TheSalkInstituteforBiologicalStudies,LaJolla,CA92037;bBioinformaticsProgram,UniversityofCalifornia,SanDiego,LaJolla,CA92093;cLudwigInstituteforCancerResearch,UniversityofCalifornia,SanDiego,LaJolla,CA92093;dLawrenceBerkeleyNationalLaboratory,Berkeley,CA94720;ics,UniversityofCalifornia,SanFrancisco,CA94143;fDepartmentofNeurology,UniversityofCalifornia,SanFrancisco,CA94143;gUSDepartmentofEnergyJointGenomeInstitute,WalnutCreek,CA94598;hSchoolofNaturalSciences,UniversityofCalifornia,Merced,CA95343;iDepartmentofCellularandMolecularMedicine,UniversityofCalifornia,SanDiego,LaJolla,CA92093;andjHowardHughesMedicalInstitute,TheSalkInstituteforBiologicalStudies,LaJolla,,January9,2017(sentforreviewNovember7,2016;)AccurateenhanceridentificationiscriticalforunderstandingtheSecond,existingmethodsoftenperformworsewhentestedonspatiotemporaltranscriptionalregulationduringdevelopmentascellsandtissuesotherthanthecell/tissuetypesusedfortrainingofwellasthefunctionalimpactofdisease-,existingmethodsconsideronlyonecell/tis-putationalmethodshavebeendevelopedtopredictsuetypeatatime,andthusneglectpotentiallyusefulinformationthegenomiclocationsofactiveenhancersbasedonhistoneaboutthevariationbetweencell/,uracyandresolutionofthesemethodsToaddresstheselimitations,,wepresentanalgorithm,regulatoryelementpredictionbasedontissue-icmarkspredictionbasedontissue-icmarks(REPTILE),(REPTILE),analgorithmtopredictenhancersbyintegratingwhichintegrateshistonemodificationandwhole-genomecytosinewhole-genome,base-resolutioncell/tissue-specificDNAmethyl-(mC)isatypeofchemicalmodificationthatplaysparedwithexistingcriticalrolesingeneregulation,transposonrepression,andtheBIOPHYSICSANDmethods,REPTILEshowsconsistentlysuperiorperformanceacross–determinationofcellidentity(1417).PUTATIONALBIOLOGYdiversecellandtissuetypes,andtheenhancerlocationsaresignifi-ursinbothCGandnon-CGcontexts(18–22),byincorporatingbase-resolutionquantifiedatnucleotideresolutionusingwhole-genomebisulfitemethylationdata,REPTILEgreatlyimprovesuponcurrentmethodssequencing(WGBS)(18).Inthisstudy,:///yupenghe/REPTILE/.prevalentformofcytosinemethylation(mCG).Transcriptionfac-torbindingsites(TFBSs)aregenerallydepletedofmCG(18,23).enhancerprediction|DNAmethylation|bioinformatics|WhethermCGaffectsbindingaffinityisunclearforthemajorityofgeneregulation|icsSignificancenmammals,,whenandwhereageneistranscribedarepri-geneexpressionisprimarilydrivenbytheactivityofdistalregulatorymarilyregulatedbytheactivityofregulatoryDNAelements,orsequences,(1–6).Moreover,-However,ourknowledgeremainslimitedaboutthelocationnome-wideassociationstudies(GWASs)lieinnoncodingregions,,putationalapproach,throughdisruptingenhanceractivity(7,8).Toidentifycausalnon-regulatoryelementpredictionbasedontissue-specificlocalcodingvariantsandunderstandtheirfunctionalconsequences,icmarks(REPTILE),-wideDNAmethylationandhistonemodificationEnhancersareboundbytranscriptionfactors(TFs),,wefoundthatenhancerpredictionsfrom(9).essiblechro-REPTILEaremorelikelytobeactiveinvivoandthepredictedmatinandmarkedbyenrichmentofhistoneH3lysine4mono-(H3K4me1)andH3lysine27acetylation(H3K27ac)–Authorcontributions:;.,.,.,.,.,.,(1012)..,.,.,;.,.,,icmarks;.,.,massivelyparallelsequencing(ChIP-seq)..,;.,.,;-genomebisulfitesequencingdataforbinationsofthesegenome-;;.,.,.,;andprofiles[seereview(13)foralistofrepresentativemethods].-learningalgorithmstolearnthehis-Reviewers:.,VanAndelInstitute;.,.Althoughtheyhaveproventobeuseful,,:******@:/-/DCSupplemental./doi/|1of8TFs,althoughrecentstudiessuggestthattherecanbesignificantMyt1lAqueryregionBMyt1lGenealterationofbindingaffinity(24–26).TheanticorrelationofmCGMyt1landTFbindingispredictiveininferringTFBS(27)anden-mESCs1DMRDMRmESCs2hancers(23,28).-resolution(～)-resolutionhistonemodificationdataderivedfromChIP-()(29).Ourresultsindicatethat,,--------devREPTILEenhancerobservations:(i)activeenhancers,whichareboundbyTFsinPEDLAenhancerEnhancerRFECSenhancerpredictioncertaincellsandtissues,showcell/tissue-specifichypomethy-DELTAenhancerlated,andsuchanticorrelationisaninformativefeatureinpre-CSIANNenhancerDNase---[alsoREPTILEenhancermodelknownasdifferentiallymethylatedregions(DMRs)]stronglyCqueryqueryregionsregionsbed-classifierforDMR-kknownnownenhancersenhancers-classifierforqueryregionoverlapwithenhancers(19,20,30).(ii)Withbase-resolutionmethylomes-kknownnownnegativesnegatives(targetsamplemCGdata,u-labelslabelsofquerqueryyreregionsgionstextandreferences)REPTILEratelydefined,whichmaybeinformativeinidentifyingthepre-()TrainingDMRsbedciselocationofenhancers.(iii)Theknownenhancers(31,32)icmarmarksksbigWig(～2kb)aregenerallymuchlargerthanTFBSs(～10–20bp)andDMRcalling((DNAm,DNAm,H3K4me1H3K4me1,,algorithmH3K4me2,etcetc))“queryregion”todescribesuchlargeregions-sslidinglidingwwindowsindows-()-.(A)Differentiallymethylatedregions(DMRs),queryregionmayhavelittlecontributiontoitsenhanceractivity,typicallysmallerthanqueryregions,serveashigh-resolutionenhancertheepigenomicsignatureofthewholeactivequeryregionmaycandidatesinoverlappedqueryregions.(B)Exampleofaregion(chr12:notbeanidealapproximationtotheepigenomicstateofthebona29,660,800–29,668,600)whereREPTILEusesbase-resolutionDNAmethyl-,(～500bp)topinpointthepossibleregulatorysubregionsgenemodel(GENCODEM2)inthisregionisshownatthetop(“Gene”).“”withinthequeryregionsandtocaptureinformativelocalepi--tissues,,dictiongenerationprocesses().,theREPTILEalgorithminvolvesfourmajorstepsDMRsacrossthesesamples.“Histonemodification”showsthelogtwofold().First,paringthemCGchangeofhistonemodificationChIP-(inwhichenhancerswillbepre-putationalmethodsarevisualizedin“Enhancerprediction.”dicted)andseveraldifferentcell/tissuetypes(whichserveasPredictionsfromREPTILEbestrecapitulatetheopenchromatindatashownreference)(Methods).Next,REPTILEintegratesepigenomicin“DNase-seq.”LightredrectanglesmarktheREPTILEputativeenhancers,dataandrepresentseachDMRorqueryregionasafeaturewhereasthegenomiclocationsofthemidpoints(.,centers)arehigh-vector,whereeachelementisthevalueofeithertheintensityorlightedinred.(C)WorkflowofREPTILE,includingfourmajorsteps.(1)DMRsicmark().Thein-paringtheCGmethylationprofilesoftargetsampleandthereferencesamples.(2)REPTILEintegratesdataininputfilesandtensitydeviationfeaturecapturestheepigenomicvariationbetweenrepresentsqueryregionsandDMRsasfeaturevectors(D).Yellowtextonthecell/tissuetypesandisauniqueaspectofREPTILE,whereastoprightcornershowstheformatforeachinputdatatype.(3)REPTILEexistingmethodsrelyondataofasinglecell/tissuetype().Inthethirdstep,REPTILElearnsamodelofen-enhancersandnegativesequencesaswellastheDMRswithinthem(redhancerepigenomicsignaturesfromthefeaturevaluesof(putative)arrows).(4)Predictionsaregeneratedbasedontheenhancermodel,DMR,knownenhancersandnegativeregionsaswellastheDMRswithinqueryregions,andepigenomicdata(bluearrows).(D)(33)classifiers,“-dev”featuresinthevectoraretheontheirownepigenomicsignature(Methods).Inthelaststep,-hancerconfidencescoresforDMRsandqueryregions,basedonwhichthefinalpredictionsaregenerated(Methods).onhiston