運(yùn)用微晶片篩選癌癥基因和探討其上游之調(diào)控

Insilicostudyofcancer-relatedgenesandmicroRNAs

運(yùn)用微晶片篩選癌癥基因及探討其上游之調(diào)控microRNAsKa-LokNg(吳家樂)DepartmentofBiomedicalInformatics(生物與醫(yī)學(xué)資訊學(xué)系)AsiaUniversity

第1頁ContentsMotivationPredictcancergenesbasedonmicroarraymRNAexpressionlevelsmicroRNA(miRNA)canactasanoncogene(OCG)ortumorsuppressorgene(TSG)Identifycancer-relatedmiRNAs,theirtargetgenes,downstreamprotein-proteininteractions(predictionnovelcancerousproteins)(1) Introduction–microarray,cancer,microRNA(2) Methods–inputdata(3) Results (a)cancergenesprediction(Bioconductor),state/breastcancer (b)correlationstudyofmiRNAsandmRNAexpressionlevels (c)ncRNAppi–AplatformforstudyingmicroRNAandtheirtargetgenes’protein-proteininteractions(4) Summary第2頁CentraldogmaofmolecularbiologyPost-transcriptionregulation–microRNAtargetsmRNAtranscriptome第3頁TypesofRNAsRNAmRNAncRNANon-codingRNA.TranscribedRNAwithastructural,functionalorcatalyticrolerRNARibosomalRNAParticipateinproteinsynthesistRNATransferRNAInterfacebetweenmRNA&aminoacidssnRNASmallnuclearRNAIncl.RNAthatformpartofthespliceosomesnoRNASmallnucleolarRNAFoundinnucleolus,involvedinmodificationofrRNAmiRNAMicroRNASmallRNAinvolvedregulationofexpressionOtherIncludinglargeRNAwithrolesinchromotinstructureandimprintingsiRNASmallinterferingRNAActivemoleculesinRNAinterferencestRNASmalltemporalRNA.RNAwitharoleindevelopmentaltimingIntroduction第4頁癌癥旳形成及

97年臺灣前十大重要癌癥死亡因素摘要順位死亡因素CauseofDeath死亡數(shù)百分率癌癥類型CancerType38,913100%1肺癌LungCancer7,77720.0%2肝癌HepatocellularCarcinoma7,65119.7%3結(jié)腸直腸癌ColorectalCancer4,26611.0%4女性乳癌FemaleBreastCancer1,5414.0%5胃癌GastricCancer2,2925.9%6口腔癌OralCavityCancer2,2185.7%7前列腺(攝護(hù)腺)癌ProstateCancer8922.3%8子宮頸癌CervicalCancer7101.8%9食道癌EsophagealCancer1,4333.7%10胰臟癌PancreaticCancer1,3643.5%第5頁ByHanneJarmer,BioCentrum-DTU,TechnicalUniversityofDenmarkcDNAlabeledbyCy3(Green)cDNAlabeledbyCy5(Red)ProbegenesTargetMicroarray–overview第6頁Microarraysareusedtomeasuregeneexpressionlevelsintwodifferentconditions.Green

labelforthecontrolsampleandaredonefortheexperimentalsample.

DNA-cDNAorDNA-mRNAhybridization.

Thehybridisedmicroarrayisexcitedbyalaserandscannedattheappropriatewavelenghtsfortheredandgreendyes

Amountoffluorescenceemitted(intensity)uponlaserexcitation~amountofmRNAboundtoeachspot

Ifthesampleincontrol/experimentalconditionisinabundancegreen/red,whichindicatestherelativeamountoftranscriptforthemRNA(EST)inthesamples.

Ifbothareequalyellow

IfneitherarepresentblackcDNAmicroarrays第7頁Microarraydatageneration,processingandanalysisInformationprocessingImagequantitation–locatingthespotsandmeasuringtheirfluorescenceintensitiesDatanormalizationandintegration–constructionofthegeneexpressionmatrixfromsetsofspotGeneexpressiondataanalysisandmining–findingdifferentiallyexpressedgenes(DEGs)orclustersofsimilarlyexpressedgenesGenerationfromtheseanalysesofnewhypothesesabouttheunderlyingbiologicalprocesses

stimulatesnewhypothesesthatinturnshouldbetestedinfollow-upexperiments/company/pressroom/image_library/biotech.htmlImageanalysisDataanalysisclustering第8頁miRNAgenepri-miRNA(stem-loopstructure)processedbyDroshapre-miRNA(65~90bp)carriedbyExportin5tocytoplasmmaturemiRNA(20~25bp)isgeneratedbytheRNaseIIItypeenzymeDicerdirectedbyRISCtothemiRNAtargetmRNAcleavageorimpedeitstranslationintoproteinIntroduction–biogenesisofmicroRNA第9頁WhenmiRNAplaysanoncogenicrole,ittargetsTSG,controlcelldifferentiationorapoptosisgenes,andleadstotumorformation.ifmiRNAplaysthetumorsuppressorrole,ittargetsOCG,controlcelldifferentiationorapoptosisgenes,soitcansuppresstumorformation.ExpectnegativecorrelationofmiRNAandmRNAexpressionprofilesintegratethehumanmiRNA-targeted(orsiRNA-targetd)mRNAdata,protein-proteininteractions(PPI)records,tissues,pathways,anddiseaseinformationtoestablishadisease-relatedmiRNA(orsiRNA)pathwaydatabaseIntroduction-miRNAscanplaytheroleofanOCGandTSG第10頁Introduction–cancer-relatedmiRNAsCancer-relatedmiRNACancertypeReferencesmiR-17-92cluster,let-7LungcancerMartinetal.,2023,Yanaiharaeta.2023,Takamizawaetal.,2023miR-10b,miR-21,miR-125b,miR-145,miR-155BreastcancerIorioetal.,2023,Sietal.,2023miR-18,miR-122a,miR-224,miR-199a,miR-199a*LivercancerMurakamietal.,2023,Mengetal.,2023,Gramantierietal.,2023miR-195,miR-125a,miR-200a,miR15,miR-16B-CLLCalinetal.,2023Calinetal.2023第11頁AplatformforstudyingmiRNAsandcanceroustargetgenesmiRNAmRNAmiRNA-mRNAanti-correlationpairsAnnotation:TAGknownOCG,TSGorCRGOMIMdiseasegenesKEGGcancerpathwaysAnnotation:miR2Disease–diseaserelatedmiRNAChromosomalfragilesitesmiRNAclustersinfo.CpGislandproximalmiRNATarBASEdataExperimentallyverifiedmiRNA-mRNApairsNCI-60cancerdata:ExpressionprofileofmiRNAandmRNA

BreastCNSColonLungLeukemiaMelanomaOvarianProstateRenalNo.ofCellLines5679610728NumberofcelllinesfortheninecancertypesintheNCI-60datasets第12頁miRNA,targetgene,protein-proteininteraction(PPI)TissuespecificmiRNAorsiRNAtarget,anditsPPIpartnersuptothesecondlevelIftheupstreammiRNA(orsiRNA)isdefective,itseffectcouldbeamplifieddownstream.Asanillustration,giventhatamiRNA(orsiRNA)targetsgeneTG,whichhastwosuccessivePPIpartners,teinsL1

andL2;andsupposethatgenesTGandL2areinvolvedwiththesamedisease,thenitishighlyprobablythatgeneL1isalsorelatedtothesamediseasequantifybyenrichmentanalysismiRNAorsiRNAprotein(mRNAissuppressed)proteinprotein(TF)protein

TGL1L2BP/MFxyzOverlapBP/MFn1n2第13頁InputdataandMethodsDatabases:ArrayExpress64prostatecancertissueand18normalprostatetissuesamples’rawdatafileswithU95Av2TAG(TumorAssociatedGene)NCI-60–miRNAandmRNAgeneexpressionprofilesfor9cancertypesTarBase–miRNAtargets(experimentalverified)miR2DiseaseacomprehensiveresourceofmiRNAderegulationinvarioushumandiseasesOMIM–humandiseaseinformationKEGG–cancerpathwaysinformationncRNAppi

ausefultoolforidentifyingncRNAtargetpathwaysPPIdata(BioIR)–Sevendatabasesareintegrated:HPRD,DIP,BIND,IntAct,MIPS,MINTandBioGRIDGeneOntology(GO)–BiologicalFunction,MolecularProcessannotationsTool:Bioconductor第14頁Research

Protocol第15頁TermEntercommandinRenvironment1library("affy")2library("limma")

3eset<-justRMA()4design<-cbind(normal=c(rep(1,18),rep(0,64)),DM=c(rep(0,18),rep(1,64)))5fit<-lmFit(eset,design)6cont.matrix<-makeContrasts(DMvsNo=DM-normal,levels=design)7fit2<-contrasts.fit(fit,cont.matrix)8fit2<-eBayes(fit2)9topTable(fit2,number=100,adjust="BH")10genenames<-as.character(topTable(fit2,number=100,adjust="BH")$ID)11adj.P_Val<-signif(topTable(fit2,number=100,adjust="BH")$adj.P.Val,digits=3)12logFC<-signif(topTable(fit2,number=100,adjust="BH")$logFC,digits=3)13library("XML")14annotation(eset)15library("annotate")16library("hgu95av2.db")17absts<-pm.getabst(genenames,"hgu95av2.db")18library("annaffy")19atab<-aafTableAnn(genenames,"hgu95av2.db",aaf.handler())20stattable<-aafTable("logFC"=logFC,"adj_P.Val"=adj.P_Val)21table<-merge(atab,stattable)22saveHTML(table,file="report.html",title="Significantgenelistanditsannotationinformation")PredictDEGsusingRandBioconductorcommands第16頁Results–DEGspredictedbyBioconductorTheresultofthetop100DEGs(eitherupordown)Eliminateduplicatedgenes,thepredictedtotalnumberofDEGsis85,andtheadjustedp-valueofallDEGsarelessthan1.9*10-5.TAG∩DEGs14knowncancergenesamongthe85predictedDEGs(16.5%)第17頁Results–miRNAs,DEGsandcancertypesOtherDEGs第18頁Results-TherelationshipamongmiR-20a,TGFBR2andhumanprostatecancer16461460.tw/R_cancer/第19頁AplatformforstudyingmiRNAsandcanceroustargetgenes第20頁AplatformforstudyingmiRNAsandcanceroustargetgenesmiRNAmRNAmiRNA-mRNAanti-correlationpairsAnnotation:TAGknownOCG,TSGorCRGOMIMdiseasegenesKEGGcancerpathwaysAnnotation:miR2Disease–diseaserelatedmiRNAChromosomalfragilesitesmiRNAclustersinfo.CpGislandproximalmiRNATarBASEdataExperimentallyverifiedmiRNA-mRNApairsNCI-60cancerdata:ExpressionprofileofmiRNAandmRNA

BreastCNSColonLungLeukemiaMelanomaOvarianProstateRenalNo.ofCellLines5679610728NumberofcelllinesfortheninecancertypesintheNCI-60datasets第21頁AplatformforstudyingmiRNAsandcanceroustargetgenesForagivencancertissuetype,wecalculatedboththePCCandSRC,r,betweentheisgivenby,wherexiandyidenotetheexpressionintensityofmiRNAandthemiRNA'stargetgenerespectively.OneofthetroubleswithquantifyingthestrengthofcorrelationbyPCCisthatitissusceptibletobeskewedbyoutliers.Outliersthatareasingledatapointcanresultintwogenesappearingtobecorrelated,evenwhenalltheotherdatapointsnot.SRCisanon-parametricstatisticalmethodthatisrobusttooutliers.ThePCCandSRCarecalculatedfor: ThreeAffymetrixchips:U95(A-E),U133A,U133B Normalizationmethods:GCRMA,MAS5,RMA第22頁TestofhypothesisofPCCandSRCThePearsonproduct-momenttabletotestthesignificanceofaPCCresult.Thehypothesisbeingtestedisaone-tailedtest.AdifferenttestisappliedfortheSRCresults.Criticalvaluesforone-tailedtestusingPearsonandSpearmancorrelationatasignificantlevelofaequalto0.05and0.10.第23頁Results–hsa-miR-1:AXL,PCCandSRCcalculationsCaseswherebothPCCandSRCarelessthanorequalto-0.5.第24頁Results–hsa-miR-10b:HOXD10miR2Disease-hsa-mir-10binitiateddiseases,i.e.leukemia,breast,colon,ovarian,prostatecancers.Anotherexample:hsa-miR-21:PTEN(TSG)hsa-miR-15b:BCL2(TSG)hsa-miR-16:BCL2(TSG)第25頁Extension-worksinprogressValidatehowgoodiscorrelationpredictionAddingfurtherinformation–CpGisland,miRNAslocatedaroundCpGislands(i.e.,miR-34b,miR-137,miR-193a,andmiR-203)aresilencedbyDNAhypermethylationinoralcancermiRNAclusters,fragilesitesPositivecorrelatedmiRNA:mRNApairsmayinvolvingTFs第26頁ncRNAppi–miRNA,targetgenes,PPI,and

theprotocolofenrichmentanalysis

Thereisatendencyfortwodirectlyinteractingproteinsparticipateinthesamebiologicalprocessorsharethesamemolecularfunction.LetamiRNAtargetingpathwaydenotedbymiRNA–TG–L1–L2.Weproposetorankthepathwayresultaccordingtothenumberofoverlappingofthebiologicalprocesses(ormolecularfunctions)betweenTGandL1,andbetweenL1andL2.TheJaccardcoefficient(JC)isusedtorankthesignificanceofapathway.

JCofsetAandBisdefinedbywhereanddenotethecardinalityofandrespectively.miRNAorsiRNAprotein(mRNAissuppressed)proteinprotein(TF)proteinJC(TG,L1)JC(L1,L2)第27頁ncRNAppi–TheprotocolofenrichmentanalysisThebiologicalprocess(BP)andmolecularfunction(MF)annotationsarecarriedfromGeneOntology,whichisusedtocharacterizethepathTG–L1–L2,andtheJCforthepathwayisgivenby,whereanddenotetheJCscoreofthebiologicalprocessforsegmentTG–L1,andtheTG–L1–L2pathwayrespectively.第28頁ncRNAppi–Theprotocolofenrichmentanalysis,p-valueWeassignedap-valuetoeveryJCcalculation,thisprovidesameasureofthestatisticalsignificance.Hereishowweestimatethep-value.LetNbetotalnumberofBPfoundinGO.AssumethatTG,

L1andL2havex,yandzBPannotationsrespectively.Also,letn1andn2bethenumberofidenticalBPforTG–L1andL1–L2respectively.Letp1andp2betheprobabilitiesthatTG–L1andL1–L2haven1andn2commonBP(orMF)termsrespectively,whicharedefinedas;andTGL1x-n1n1y-n1N第29頁ncRNAppi–ExtensionofTarBasetargetsLimitationsofmiRNAtargetpredictiontoolsTherearemanytoolsavailableformiRNAtargetgenesprediction,suchasmiRanda,TargetScan,andRNAhybridetc.AmajorproblemofmiRNAtargetgenespredictionisthatthepredictionaccuracyremainsuncertain,therewasreportindicatedthatthefalsepositiveratecouldbeashighas24-39%formiRanda,and22-31%forTargetScan.IfthemiRNA:mRNAtargetingpartisuncertain,thenthe‘Level1’and‘Level2’protein-proteininteractionpathwaysderivedfromPPIdatabasearedoubtful.第30頁ncRNAppi–ExtensionofTarBasetargetsmiRNAtargetpredictiontool–miRandaMaturehumanmiRNAFASTAsequencesisdownloadedfrommiRBase (thelatestversionis13).Then,wepredictthepossibilitiesofmiRNAbindingwithOCG,andTSG.Targetpredictiontool,miRanda,allowsforfiningtuningofcertainparameters,i.e.MFEthreshold,score,shufflestatistics,gapopenandgapextensionscores.WesetMFEthresholdandtheshufflestatisticsto-25kcal/molandONrespectively.Therestoftheparametersaresettotheirdefaultvalues.OncethebindinglistsofOCGandTSGobtained,thentheirPPIpathwayscanberetrievedfromtheBioIRdatabase.第31頁

ncRNAppiprovidesweb-baseddataaccessandallowsdiseaseassignmentforaspecificnodealongmiRNA(siRNA)targetingpathways.ForexampleSelectmiRNAID–hsa-let-7Checksthe‘OMIMDiseasetypeforindividualnode’boxlabeledwith‘Target’and‘Level-2’Choosetheitem‘lungtumor’underthe‘TUMORTYPE’pull-downmenu(OMIM)Select‘Yes’undert