Bayesian Adaptive Alignment and Inference

Sequencealignmentwithoutthe specification of gap penaltiesor a scoringmatrixis attainedbyusing Bayesianinferenceanda recursivealgorithm.Thisprocedure’srecursivealgorithmsumsoverall possible alignmentsonthe forwardstep to obtainnormalizing constantsessential to Bayesianinferences,andsamples fromthe exactposteriordistribution onthe backward step. Sincebothterminalandinterveningunrelated subsequenceswill often be excludedfroman alignment, the resulting alignmentsmaybe seenas extensionsof local alignments.Analignment’ssignificanceis assessedusingthe Bayesianevidence.Ashuffling simulation showsthat Bayesianevidenceagainst the null hypothesistendsto be a conservativemeasureof significancecomparedto classical p-values.Anapplicationto proteins fromthe GTPasesuperfamilyshowsthat the posteriordistributionof the numberof gapsis oftenfiat andthat the posteriordislributionof the evolutionary distanceis oftenflat andsometimesbimodal.Analignmentof 1GIAwith 1ETUshowsgoodcorrespondence witha struc...

• Type: Conference
• Year: 1997
• Where: ISMB
• Authors: Jun Zhu, Jun S. Liu, Charles E. Lawrence
• Keywords: Computational Biology, ISMB 1997, Possible Alignmentsonthe Forwardstep, Scoringmatrixis Attainedbyusing Bayesianinferenceanda, Witha Structural Alignment