基于logistic回歸分析的P53下游基因判別分析模型.pdf

1、重慶郵電大學碩士學位論文基于logistic回歸分析的P53下游基因判別分析模型姓名：白明澤申請學位級別：碩士專業(yè)：計算機應用技術指導教師：譚軍20060610重慶郵電大學碩士論文摘要AbstractP53proteinisageneraltranscriptionfactorwhichCallaccomplishmanyphysiologicalfunctionsthroughregulatingdownstreamgenes’exp

2、ressionThosefunctionsmostlyrelatedtonegativeregulatingthecellcycle，replication&repairofDNA，apoptosis，restrainingthegrowthofbloodvesselandresponsetocelltressResearchesonp53indicatethatp53isa11importanttumorsuppressorgene，

3、definingthecompletegeneregulatorynetworkistheultimatechallengetop53biologyAbout100humangeneswhoseexpressionispositivelyregulatedbyp53havebeenidentifiedAlthoughthenumberofidentifiedP53downstreamgeneskeepsgrowing，itisconce

4、ivablethatlargefractionoftheP53downstreamgeneshavenotyetbeenidentifiedDuetobiologyexperimentstoidentifythosegenesistediousandexpensiveAcomputationalmethodwhichsynthesistheknownP53downstreamgenes’featureiseconomicandeffec

5、tiveforidentifyingtheotherswhichhavenotbeenidentified，andwillbringbreakthroughinthisareaInthispapertotal63humanP53’Sbindingsequencesand13clonesequenceswhichcailbindwitllP53identifiedinE1一Deiry’Sexperimenthasbeencollected

6、AdiscriminantanalysismodelforP53downstreamgenesbasedonlogisticregressionanalysisisproposedThecandidatefeaturesofprimarysequencearecalculatedbyselectingpropermodelsincludingPWMmodel，frequencydistributionmodel，consensusseq

7、uencemodelandthelengthofinsertsequenceinthemotifInviewoftheP53bindingsequences’particularitytwoPWMmatricesareadoptedtomodelingthetwodecamersineverymotifandacrossvalidatemethodisusedtoaffirmthemotifineveryknownbindingsequ

8、enceThenthosemotifs’featuresareconsideredastheobjectsofthelogisticregressionanalysisThroughastepwiseselectionprocessofferedbySPSS，theoptimalfeaturesincludingthetwodecamers’PWMscorearedeterminedfromcandidatefeaturesetsThe

9、modelhasbeentrainedandtestedontheselectedpositiveandnegativedatasetsbythejackknifemethod，andtheaveragepredictionaccuracyis9391％TheresultsindicatethatOurmethodisauniversalalgorithmthatoutperformsmostofthesimpleapproachesF