樊龙江 附录：生物信息学主要英文术语及释义 AbstractSyntaxNotation(ASN.l)（NCBI发展的许多程序，如显示蛋白质三维结构的Cn3D等所使用的内部格式）Alanguagethatisusedtodescribestructureddatatypesformally,Withinbioinformatits,ithasbeenusedbytheNationalCenterforBiotechnologyInformationtoencodesequences,maps,taxonomicinformation,molecularstructures,andbiographicalinformationinsuchawaythatitcanbeeasilyessedandexchangedputersoftware.essionnumber（记录号）AuniqueidentifierthatisassignedtoasingledatabaseentryforaDNAorproteinsequence.Affinegappenalty（一种设置空位罚分策略）Agappenaltyscorethatisalinearfunctionofgaplength,consistingofagapopeningpenaltyandagapextensionpenaltymultipliedbythelengthofthegap.Usingthispenaltyschemegreatlyenhancestheperformanceofdynamicprogrammingmethodsforsequencealignment.SeealsoGappenalty.Algorithm（算法）Asystematicprocedureforsolvingaprobleminafinitenumberofsteps,typicallyinvolvingarepetitionofoperations.Oncespecified,analgorithmcanbewritteninputerlanguageandrunasaprogram.Alignment（联配/比对/联配）Referstotheprocedureparingtwoormoresequencesbylookingforaseriesofindividualcharactersorcharacterpatternsthatareinthesameorderinthesequences.Ofthetwotypesofalignment,localandglobal,alocalalignmentisgenerallythemostuseful.SeealsoLocalandGlobalalignments.Alignmentscore（联配/比对/联配值）Anputedscorebasedonthenumberofmatches,substitutions,insertions,anddeletions(gaps)withinanalignment.ScoresformatchesandsubstitutionsArederivedfromascoringmatrixsuchastheBLOSUMandPAMmatricesforproteins,andaftinegappenaltiessuitableforthematrixarechosen.Alignmentscoresareinlogoddsunits,oftenbitunits(logtothebase2).Higherscoresdenotebetteralignments.SeealsoSimilarityscore,Distanceinsequenceanalysis.Alphabet（字母表）Thetotalnumberofsymbolsinasequence-4forDNAsequencesand20forproteinsequences.Annotation（注释）Thepredictionofgenesinagenome,includingthelocationofprotein-encodinggenes,thesequenceoftheencodedproteins,anysignificant 125 《生物信息学札记》樊龙江 matchestootherProteinsofknownfunction,andthelocationofRNA-encodinggenes.Predictionsarebasedongenemodels;e.g.,hiddenMarkovmodelsofintronsandexonsinproteinsencodinggenes,andmodelsofsecondarystructureinRNA.AnonymousFTP（匿名FTP）WhenaFTPserviceallowsanyonetologin,itissaidtoprovideanonymousFTPser-vice.AusercanlogintoananonymousFTPserverbytypinganonymousastheusernameandhisE-mailaddressasapassword.MostWebbrowsersnownegotiateanonymousFTPlogonwithoutaskingtheuserforausernameandpassword.SeealsoFTP.ASCIITheAmericanStandardCodeforInformationInterchange(ASCII)encodesentedlettersa-z,A-Z,thenumbersO-9,mostpunctuationmarks,space,andasetofcontrolcharacterssuchascarriagereturnandtab.ASCIIspecifies128charactersthataremappedtothevaluesO-127.ASCIItilesmonlycalledplaintext,meaningthattheyonlyencodetextwithoutextramarkup.BACclone（细菌人工染色体克隆）BacterialartificialchromosomevectorcarryingagenomicDNAinsert,typically100–200kb.Mostofthelarge-insertclonessequencedintheprojectwereBACclones.Back-propagation（反向传输）Whentrainingfeed-forwardworks,aback-propagationalgorithmcanbeusedtomodifyworkweights.Aftereachtraininginputpatternisfedthroughwork,work’soutputparedwiththedesiredoutputandtheamountoferroriscalculated.Thiserrorisback-propagatedthroughworkbyusinganerrorfunctiontocorrectworkweights.SeealsoFeed-forwardwork.Baum-Welchalgorithm（Baum-Welch算法）AnexpectationmaximizationalgorithmthatisusedtotrainhiddenMarkovmodels.Baye’srule（贝叶斯法则）Formsthebasisofconditionalprobabilitybycalculatingthelikelihoodofaneventurringbasedonthehistoryoftheeventandrelevantbackgroundinformation.IntermsoftwoparametersAandB,thetheoremisstatedinanequation:Thecondition-alprobabilityofA,givenB,P(AIB),isequaltotheprobabilityofA,P(A),timestheconditionalprobabilityofB,givenA,P(BIA),dividedbytheprobabilityofB,P(B).P(A)isthehistoricalorpriordistributionvalueofA,P(BIA)isanewpredictionforBforaparticularvalueofA,andP(B)isthesumofthenewlypredictedvaluesforB.P(AIB)isaposteriorprobability,representinganewpredictionforAgiventhepriorknowledgeofAandthenewlydiscoveredrelationshipsbetweenAandB.Bayesiananalysis（贝叶斯分析）Astatisticalprocedureusedtoestimateparametersofanunderlying 126 《生物信息学札记》樊龙江 distributionbasedonanobserveddistribution.SeealsoBaye’srule.Biochips（生物芯片）Miniaturizedarraysoflargenumbersofmolecularsubstrates,oftenoligonucleotides,inadefinedpattern.TheyarealsocalledDNAmicroarraysandmicrochips.Bioinformatics（生物信息学）Themergerofbiotechnologyandinformationtechnologywiththegoalofrevealingnewinsightsandprinciplesinbiology./Thedisciplineofobtaininginformationaboutgenomicorproteinsequencedata.Thismayinvolvesimilaritysearchesofdatabases,paringyourunidentifiedsequencetothesequencesinadatabase,ormakingpredictionsaboutthesequencebasedoncurrentknowledgeofsimilarsequences.Databasesarefrequentlymadepublicallyavailablethroughthe,orlocallyatyourinstitution.Bitscore（二进制值/Bit值）ThevalueS'isderivedfromtherawalignmentscoreSinwhichthestatisticalpropertiesofthescoringsystemusedhavebeentakenintoount.Becausebitscoreshavebeennormalizedwithrespecttothescoringsystem,theycanbeusedparealignmentscoresfromdifferentsearches.BitunitsFrominformationtheory,abitdenotestheamountofinformationrequiredtodistinguishbetweentwoequallylikelypossibilities.Thenumberofbitsofinformation,AJ,requiredtoconveyamessagethathasA4possibilitiesislog2M=Nbits.BLAST（基本局部联配搜索工具，一种主要数据库搜索程序）BasicLocalAlignmentSearchTool.Asetofprograms,usedtoperformfastsimilaritysearches.NucleotidesequencescanparedwithnucleotidesequencesinadatabaseusingBLASTN,forexample.Complexstatisticsareappliedtojudgethesignificanceofeachmatch.Reportedsequencesmaybehomologousto,orrelatedtothequerysequence.TheBLASTPprogramisusedtosearchaproteindatabaseforamatchagainstaqueryproteinsequence.ThereareseveralotherflavoursofBLAST.BLAST2isanewerreleaseofBLAST.Allowsforinsertionsordeletionsinthesequencesbeingaligned.Gappedalignmentsmaybemorebiologicallysignificant.Block（蛋白质家族中保守区域的组块）Conservedungappedpatternsapproximately3-60aminoacidsinlengthinasetofrelatedproteins.BLOSUMmatrices（模块替换矩阵，一种主要替换矩阵）AnalternativetoPAMtables,BLOSUMtableswerederivedusinglocalmultiplealignmentsofmoredistantlyrelatedsequencesthanwereusedforthePAMmatrix.Theseareusedtoassessthesimilarityofsequenceswhenperformingalignments.Boltzmanndistribution（Boltzmann分布）Describesthenumberofmoleculesthathaveenergiesaboveacertainlevel,basedontheBoltzmanngasconstantandtheabsolutetemperature. 127 《生物信息学札记》樊龙江 Boltzmannprobabilityfunction(Boltzmann概率函数)SeeBoltzmanndistribution.BootstrapanalysisAmethodfortestinghowwellaparticulardatasetfitsamodel.Forexample,thevalidityofthebrancharrangementinapredictedictreecanbetestedbyresamplingcolumnsinamultiplesequencealignmenttocreatemanynewalignments.Theappearanceofaparticularbranchintreesgeneratedfromtheseresampledsequencescanthenbemeasured.Alternatively,asequencemaybeleftoutofananalysistodeter-minehowmuchthesequenceinfluencestheresultsofananalysis.Branchlength（分支长度）Insequenceanalysis,thenumberofsequencechangesalongaparticularbranchofaictree.CDSorcds（编码序列）Codingsequence.Chebyshe,dinequalityTheprobabilitythatarandomvariableexceedsitsmeanislessthanorequaltothesquareof1overthenumberofstandarddeviationsfromthemean.Clone（克隆）Populationofidenticalcellsormolecules(e.g.DNA),derivedfromasingleancestor.CloningVector（克隆载体）Amoleculethatcarriesaforeigngeneintoahost,andallows/facilitatesthemultiplicationofthatgeneinahost.Whensequencingagenethathasbeenclonedusingacloningvector(ratherthanbyPCR),careshouldbetakennottoincludethecloningvectorsequencewhenperformingsimilaritysearches.Plasmids,cosmids,phagemids,YACsandPACsareexampletypesofcloningvectors.Clusteranalysis（聚类分析）Amethodforgroupingtogetherasetofobjectsthataremostsimilarfromalargergroupofrelatedobjects.Therelationshipsarebasedonsomecriterionofsimilarityordifference.Forsequences,asimilarityordistancescoreorastatisticalevaluationofthosescoresisused.CobblerAsinglesequencethatrepresentsthemostconservedregionsinamultiplesequencealignment.TheBLOCKSserverusesthecobblersequencetoperformadatabasesimilaritysearchasawaytoreachsequencesthataremoredivergentthanwouldbefoundusingthesinglesequencesinthealignmentforsearches.Codingsystem(works)Regardingworks,acodingsystemneedstobedesignedforrepresentinginputandoutput.Thelevelofessfoundwhentrainingthemodelwillbepartiallydependentonthequalityofthecodingsystemchosen.Codonusage 128 《生物信息学札记》樊龙江 Analysisofthecodonsusedinaparticulargeneanism.COG（直系同源簇）Clustersoforthologousgroupsinasetofgroupsofrelatedsequencesinanismandyeast(S.cerevisiae).Thesegroupsarefoundbywholeparisonsandincludeorthologsandparalogs.SeealsoOrthologsandParalogs.Comparativegenomics（比较基因组学）parisonofgenenumbers,genelocations,andbiologicalfunctionsofgenesinthegenomesofanisms,oneobjectivebeingtoidentifygroupsofgenesthatplayauniquebiologicalroleinaanism.Complexity(ofanalgorithm)（算法的复杂性）Describesthenumberofstepsrequiredbythealgorithmtosolveaproblemasafunctionoftheamountofdata;forexample,thelengthofsequencestobealigned.Conditionalprobability（条件概率）Theprobabilityofaparticularresult(orofaparticularvalueofavariable)givenoneormoreeventsorconditions(orvaluesofothervariables).Conservation（保守）Changesataspecificpositionofanaminoacidor(monly,DNA)sequencethatpreservethephysico-chemicalpropertiesoftheoriginalresidue.Consensus（一致序列）Asinglesequencethatrepresents,ateachsubsequentposition,thevariationfoundwithincorrespondingcolumnsofamultiplesequencealignment.Context-freegrammarsArecursivesetofproductionrulesforgeneratingpatternsofstrings.Theseconsistofasetofterminalcharactersthatareusedtocreatestrings,asetofnonterminalsymbolsthatcorrespondtorulesandactasplaceholdersforpatternsthatcanbegeneratedusingterminalcharacters,asetofrulesforreplacingnonterminalsymbolswithterminalcharacters,andastartsymbol.Contig（序列重叠群/拼接序列）Asetofclonesthatcanbeassembledintoalinearorder.ADNAsequencethatoverlapswithanothercontig.Thefullsetofoverlappingsequences(contigs)canbeputtogethertoobtainthesequenceforalongregionofDNAthatcannotbesequencedinoneruninasequencingassay.Importantinicmappingatthemolecularlevel.CORBA（国际对象管理协作组制定的使OOP对象与网络接口统一起来的一套跨计算机、操作系统、程序语言和网络的共同标准） TheCommonObjectRequestBrokerArchitecture(CORBA)isanopenindustrystandardforworkingwithdistributedobjects,developedbytheObjectManagementGroup.CORBAallowstheinterconnectionofobjectsandapplicationsregardlessputerlanguage,machinearchitecture,orgeographiclocationofputers.Correlationcoefficient（相关系数） 129 《生物信息学札记》樊龙江 Anumericalmeasure,fallingbetween-1and1,ofthedegreeofthelinearrelationshipbetweentwovariables.Apositivevalueindicatesadirectrelationship,anegativevalueindicatesaninverserelationship,andthedistanceofthevalueawayfromzeroindicatesthestrengthoftherelationship.Avaluenearzeroindicatesnorelationshipbetweenthevariables.Covariation(insequences)（共变）CoincidentchangeattwoormoresequencepositionsinrelatedsequencesthatmayinfluencethesecondarystructuresofRNAorproteinmolecules.Coverage(ordepth)（覆盖率/厚度）Theaveragenumberoftimesanucleotideisrepresentedbyahigh-qualitybaseinacollectionofrandomrawsequence.Operationally,a'high-qualitybase'isdefinedasonewithanuracyofatleast99%(correspondingtoaPHREDscoreofatleast20).Database（数据库）puterizedstorehouseofdatathatprovidesastandardizedwayforlocating,adding,removing,andchangingdata.SeealsoObject-orienteddatabase,Relationaldatabase.DendogramAformofatreethatlistsparedobjects(e.g.,sequencesorgenesinamicroarrayanalysis)inaverticalorderandjoinsrelatedonesbylevelsofbranchesextendingtoonesideofthelist.Depth（厚度）SeecoverageDirichletmixturesDefinedastheconjugationalpriorofamultinomialdistribution.Oneuseisforpredictingtheexpectedpatternofaminoacidvariationfoundinthematchstateofahid-denMarkovmodel(representingonecolumnofamultiplesequencealignmentofproteins),basedonpriordistributionsfoundinconservedproteindomains(blocks).Distanceinsequenceanalysis（序列距离）Thenumberofobservedchangesinanoptimalalignmentoftwosequences,usuallynotcountinggaps.DNASequencing（DNA测序）TheexperimentalprocessofdeterminingthenucleotidesequenceofaregionofDNA.Thisisdonebylabellingeachnucleotide(
A,C,GorT)witheitheraradioactiveorfluorescentmarkerwhichidentifiesit.Thereareseveralmethodsofapplyingthistechnology,eachwiththeiradvantagesanddisadvantages.Formoreinformation,refertoacurrenttextbook.Highthroughputlaboratoriesfrequentlyuseautomatedsequencers,whicharecapableofrapidlyreadinglargenumbersoftemplates.Sometimes,thesequencesmaybegeneratedmorequicklythantheycanbecharacterised.Domain（功能域）Adiscreteportionofaproteinassumedtofoldindependentlyoftherestoftheproteinandpossessingitsownfunction. 130 《生物信息学札记》樊龙江 Dotmatrix（点标矩阵图）Dotmatrixdiagramsprovideagraphicalmethodparingtwosequences.Onesequenceiswrittenhorizontallyacrossofthegraphandtheotheralongtheleft-handside.Dotsareplacedwithinthegraphattheintersectionofthesameletterappearinginbothsequences.Aseriesofdiagonallinesinthegraphindicateregionsofalignment.Thematrixmaybefilteredtorevealthemost-alikeregionsbyscoringaminimalthresholdnumberofmatcheswithinasequencewindow.Draftgenomesequence（基因组序列草图）Thesequenceproducedbiningtheinformationfromtheindividualsequencedclones(bycreatingmergedsequencecontigsandthenemployinglinkinginformationtocreatescaffolds)andpositioningthesequencealongthephysicalmapofthechromosomes.DUST（一种低复杂性区段过滤程序）Aprogramforfilteringplexityregionsfromnucleicacidsequences.Dynamicprogramming（动态规划法）Adynamicprogrammingalgorithmsolvesaproblembiningsolutionstosub-problemsthatputedonceandsavedinatableormatrix.Dynamicprogrammingistypicallyusedwhenaproblemhasmanypossiblesolutionsandanoptimaloneneedstobefound.Thisalgorithmisusedforproducingsequencealignments,givenascoringsystemforparisons.EMBL（欧洲分子生物学实验室，EMBL数据库是主要公共核酸序列数据库之一） EuropeanMolecularBiologyLaboratories.MaintaintheEMBLdatabase,oneofthemajorpublicsequencedatabases.（欧洲分子生物学网络）EuropeanMolecularBiologyNetwork:/wasestablishedin1988,andprovidesservicesincludinglocalmoleculardatabasesandsoftwareformolecularbiologistsinEurope.Thereareseverallargeoutpostsof,includingEXPASY.Entropy（熵）Frominformationtheory,ameasureoftheunpredictablenatureofasetofpossibleelements.Thehigherthelevelofvariationwithintheset,thehighertheentropy.ErdosandRenyilawInatossofa“fair”coin,thenumberofheadsinarowthatcanbeexpectedisthelogarithmofthenumberoftossestothebase2.Thelawmaybegeneralizedformorethantwopossibleesbychangingthebaseofthelogarithmtothenumberofes.Thislawwasusedtoanalyzethenumberofmatchesandmismatchesthatcanbeexpectedbetweenrandomsequencesasabasisforscoringthestatisticalsignificanceofasequencealignment.EST（表达序列标签的缩写） 131 《生物信息学札记》樊龙江 SeeExpressedSequenceTagExpectvalue(E)（E值）Evalue.ThenumberofdifferentalignentswithscoresequivalenttoorbetterthanSthatareexpectedtourinadatabasesearchbychance.ThelowertheEvalue,themoresignificantthescore.Inadatabasesimilaritysearch,theprobabilitythatanalignmentscoreasgoodastheonefoundbetweenaquerysequenceandadatabasesequencewouldbefoundinasparisonsbetweenrandomsequencesaswasdonetofindthematchingsequence.Inothertypesofsequenceanalysis,Ehasasimilarmeaning.Expectationmaximization(sequenceanalysis)Analgorithmforlocatingsimilarsequencepatternsinasetofsequences.Aguessedalignmentofthesequencesisfirstusedtogenerateanexpectedscoringmatrixrepresentingthedistributionofsequencecharactersineachcolumnofthealignment,thispatternismatchedtoeachsequence,andthescoringmatrixvaluesarethenupdatedtomaximizethealignmentofthematrixtothesequences.Theprocedureisrepeateduntilthereisnofurtherimprovement.Exon（外显子）CodingregionofDNA.SeeCDS.ExpressedSequenceTag(EST)（表达序列标签）Randomlyselected,partialcDNAsequence;representsit'scorrespondingmRNA.dbESTisalargedatabaseofESTsatGenBank,NCBI.FASTA（一种主要数据库搜索程序）Thefirstwidelyusedalgorithmfordatabasesimilaritysearching.Theprogramlooksforoptimallocalalignmentsbyscanningthesequenceforsmallmatchescalled"words".Initially,thescoresofsegmentsinwhichtherearemultiplewordhitsarecalculated("init1").Laterthescoresofseveralsegmentsmaybesummedtogeneratean"initn"score.Anoptimizedalignmentthatincludesgapsisshownintheoutputas"opt".Thesensitivityandspeedofthesearchareinverselyrelatedandcontrolledbythe"k-tup"variablewhichspecifiesthesizeofa"word".(PearsonandLipman)Extremevaluedistribution（极值分布）Somemeasurementsarefoundtofollowadistributionthathasalongtailwhichdecaysathighvaluesmuchmoreslowlythanthatfoundinanormaldistribution.Thisslow-fallingtypeiscalledtheextremevaluedistribution.Thealignmentscoresbetweenunrelatedorrandomsequencesareanexample.Thesescorescanreachveryhighvalues,particularlywhenalargenumberparisonsaremade,asinadatabasesimilaritysearch.Theprobabilityofaparticularscoremaybeuratelypredictedbytheextremevaluedistribution,whichfollowsadoublenegativeexponentialfunctionafterGumbel.Falsenegative（假阴性）Anegativedatapointcollectedinadatasetthatwasincorrectlyreportedduetoafailureofthetestinavoidingnegativeresults. 132 《生物信息学札记》樊龙江 Falsepositive（假阳性）Apositivedatapointcollectedinadatasetthatwasincorrectlyreportedduetoafailureofthetest.Ifthetesthadcorrectlymeasuredthedatapoint,thedatawouldhavebeenrecordedasnegative.Feed-forwardwork（反向传输神经网络）Organizesnodesintosequencelayersinwhichthenodesineachlayerarefullyconnectedwiththenodesinthenextlayer,exceptforthefinaloutputlayer.Inputisfedfromtheinputlayerthroughthelayersinsequenceina“feed-forward”direction,resultinginoutputatthefinallayer.Seealsowork.Filtering(windowsize)Duringpair-wisesequencealignmentusingthedotmatrixmethod,randommatchescanbefilteredoutbyusingaslidingwindowparethetwosequences.Ratherparingasinglesequencepositionatatime,awindowofadjacentpositionsinthetwosequencesparedandadot,indicatingamatch,isgeneratedonlyifacertainminimalnumberofmatchesur.Filtering（过滤）AlsoknownasMasking.Theprocessofhidingregionsof(nucleicacidoraminoacid)sequencehavingcharacteristicsthatfrequentlyleadtospurioushighscores.SeeSEGandDUST.Finishedsequence（完成序列）Completesequenceofacloneorgenome,withanuracyofatleast99.99%andnogaps.FourieranalysisStudiestheapproximationsandpositionoffunctionsusingtrigonometricpolynomials.Format(file)（格式）Differentprogramsrequirethatinformationbespecifiedtotheminaformalmanner,usingparticularkeywordsandordering.Thisspecificationisafileformat.Forward-backwardalgorithmUsedtotrainahiddenMarkovmodelbyaligningthemodelwithtrainingsequences.Thealgorithmthenrefinesthemodeltoreducetheerrorwhenfittedtothegivendatausingagradientdescentapproach.FTP(FileTransferProtocol)（文件传输协议）AllowsapersontotransferfilesfromputertoanotheracrossworkusinganFTP-capableclientprogram.TheFTPclientprogramcanmunicatewithmachinesthatrunanFTPserver.Theserver,inturn,willmakeaspecificportionofitstilesystemavailableforFTPess,providingthattheclientisabletosupplyarecognizedusernameandpasswordtotheserver.Fullshotgunclone（鸟枪法克隆）Alarge-insertcloneforwhichfullshotgunsequencehasbeenproduced. 133 《生物信息学札记》樊龙江 Functionalgenomics（功能基因组学）Assessmentofthefunctionofgenesidentifiedbyparisons.Thefunctionofanewlyidentifiedgeneistestedbyintroducingmutationsintothegeneandthenexaminingtheresultantanismforanalteredphenotype.gap（空位/间隙/缺口）Aspaceintroducedintoanalignmentpensateforinsertionsanddeletionsinonesequencerelativetoanother.Topreventtheumulationoftoomanygapsinanalignment,introductionofagapcausesthedeductionofafixedamount(thegapscore)fromthealignmentscore.Extensionofthegaptopassadditionalnucleotidesoraminoacidisalsopenalizedinthescoringofanalignment.Gappenalty（空位罚分）Anumericscoreusedinsequencealignmentprogramstopenalizethepresenceofgapswithinanalignment.Thevalueofagappenaltyaffectshowoftengapsappearinalignmentsproducedbythealgorithm.Mostalignmentprogramssuggestgappenaltiesthatareappropriateforparticularscoringmatrices.icalgorithm（遗传算法）Akindofsearchalgorithmthatwasinspiredbytheprinciplesofevolution.Apopulationofinitialsolutionsisencodedandthealgorithmsearchesthroughthesebyapplyingapre-definedfitnessmeasurementtoeachsolution,selectingthosewiththehighestfitnessforreproduction.Newsolutionscanbegeneratedduringthisphasebycrossoverandmutationoperations,definedintheencodedsolutions.icmap（遗传图谱）Agenomemapinwhichpolymorphiclociarepositionedrelativetooneanotheronthebasisofthefrequencywithwhichtheybineduringmeiosis.Theunitofdistanceisans(cM),denotinga1%chanceofbination.Genome（基因组）Theicmaterialofanism,containedinonehaploidsetofchromosomes.GibbssamplingmethodAnalgorithmforfindingconservedpatternswithinasetofrelatedsequences.Aguessedalignmentofallbutonesequenceismadeandusedtogenerateascoringmatrixthatrepresentsthealignment.Thematrixisthenmatchedtotheleft-outsequence,andaprobablelocationofthecorrespondingpatternisfound.Thispredictionistheninputintoanewalignmentandanotherscoringmatrixisproducedandtestedonanewleft-outsequence.Theprocessisrepeateduntilthereisnofurtherimprovementinthematrix.Globalalignment（整体联配）Attemptstomatchasmanycharactersaspossible,fromendtoend,inasetoftwoor 134 《生物信息学札记》樊龙江 moresequences.Gopher(一个文档发布系统，允许检索和显示文本文件)Graphtheory（图论）Abranchofmathematicswhichdealswithproblemsthatinvolveagraphworkstructure.Agraphisdefinedbyasetofnodes(orpoints)andasetofarcs(linesoredges)joiningthenodes.Insequenceandgenomeanalysis,graphtheoryisusedforsequencealignmentsandclusteringalikegenes.GSS（基因综述序列）Genomesurveysequence.GUI（图形用户界面）Graphicaluserinterface.H（相对熵值）Histherelativeentropyofthetargetandbackgroundresiduefrequencies.(KarlinandAltschul,1990).Hcanbethoughtofasameasureoftheaverageinformation(inbits)availableperpositionthatdistinguishesanalignmentfromchance.AthighvaluesofH,shortalignmentscanbedistinguishedbychance,whereasatlowerHvalues,alongeralignmentmaybenecessary.(Altschul,1991)Half-bitsSomescoringmatricesareinhalf-bitunits.Theseunitsarelogarithmstothebase2ofoddsscorestimes2.Heuristic（启发式方法）Aprocedurethatprogressesalongempiricallinesbyusingrulesofthumbtoreachasolution.Thesolutionisnotguaranteedtobeoptimal.Hexadecimalsystem（16制系统）Thebase16countingsystemthatusesthedigitsO-9followedbythelettersA-
F.HGMP（人类基因组图谱计划）HumanGenomeMappingProject.HiddenMarkovModel(HMM)（隐马尔可夫模型）Insequenceanalysis,aHMMisusuallyaprobabilisticmodelofamultiplesequencealignment,butcanalsobeamodelofperiodicpatternsinasinglesequence,representing,forexample,patternsfoundintheexonsofagene.Inamodelofmultiplesequencealignments,eachcolumnofsymbolsinthealignmentisrepresentedbyafrequencydistributionofthesymbolscalledastate,andinsertionsanddeletionsbyotherstates.Onethenmovesthroughthemodelalongaparticularpathfromstatetostatetryingtomatchagivensequence.Thenextmatchingsymbolischosenfromeachstate,recordingitsprobability(frequency)andalsotheprobabilityofgoingtothatparticularstatefromapreviousone(thetransitionprobability).Stateandtransitionprobabilitiesarethenmultipliedtoobtainaprobabilityofthegivensequence.Generallyspeaking,aHMMisastatisticalmodelforanorderedsequenceofsymbols,actingasastochasticstatemachinethatgeneratesasymboleachtimeatransitionismadefromonestatetothenext.Transitionsbetween 135 《生物信息学札记》樊龙江 statesarespecifiedbytransitionprobabilities.Hiddenlayer（隐藏层）Aninnerlayerwithinaworkthatreceivesitsinputandsendsitsoutputtootherlayerswithinwork.Onefunctionofthehiddenlayeristodetectcovariationwithintheinputdata,suchaspatternsofaminoacidcovariationthatareassociatedwithaparticulartypeofsecondarystructureinproteins.Hierarchicalclustering（分级聚类）Theclusteringorgroupingofobjectsbasedonsomesinglecriterionofsimilarityordifference.Anexampleistheclusteringofgenesinamicroarrayexperimentbasedonthecorrelationbetweentheirexpressionpatterns.Thedistancemethodusedinicanalysisisanotherexample.HillclimbingAnonoptimalsearchalgorithmthatselectsthesingularbestpossiblesolutionatagivenstateorstep.Thesolutionmayresultinalocallybestsolutionthatisnotagloballybestsolution.Homology（同源性）Aponentinanisms(e.g.,geneswithstronglysimilarsequences)thatcanbeattributedtomonancestoroftheanismsduringevolution.Horizontaltransfer（水平转移）Thetransferoficmaterialbetweentwodistinctspeciesthatdonotordinarilyexchangeicmaterial.ThetransferredDNAesestablishedintherecipientgenomeandcanbedetectedbyanovelichistoryandcodon-paredtotherestofthegenome.HSP（高比值片段对）High-scoringsegmentpair.Localalignmentswithnogapsthatachieveoneofalignmentscoresinagivensearch.HTGS/HGT（高通量基因组序列）High-throughoutgenomesequencesHTML（超文本标识语言）TheHyper-TextMarkupLanguage(HTML)providesastructuraldescriptionofadocumentusingaspecifiedtagset.HTMLcurrentlyservesasthelinguafrancafordescribinghypertextWebpagedocuments.HyperplaneAgeneralizationofthetwo-dimensionalplanetoNdimensions.HypercubeAgeneralizationofthethree-dimensionalcubetoNdimensions.Identity（相同性/相同率）Theextenttowhichtwo(nucleotideoraminoacid)sequencesareinvariant.Indel（插入或删除的缩略语）Aninsertionordeletioninasequencealignment.Informationcontent(ofascoringmatrix)Arepresentationofthedegreeofsequenceconservationinacolumnofa 136 《生物信息学札记》樊龙江 scoringmatrixrepresentinganalignmentofrelatedsequences.Itisalsothenumberofquestionsthatmustbeaskedtomatchthecolumntoapositioninatestsequence.Forbases,themax-imumpossiblenumberis2,andforproteins,4.32(logarithmtothebase2ofthenumberofpossiblesequencecharacters).Informationtheory（信息理论）Abranchofmathematicsthatmeasuresinformationintermsofbits,theminimalamountofplexityneededtoencodeagivenpieceofinformation.Inputlayer（输入层）Theinitiallayerinafeed-forward.Thislayerencodesinputinformationthatwillbefedthroughworkmodel.InterfacedefinitionlanguageUsedtodefineaninterfacetoanobjectmodelinaprogramminglanguageneutralform,whereaninterfaceisanabstractionofaservicedefinedonlybytheoperationsthatcanbeperformedonit.（因特网）workinfrastructure,consistingofcablesinterconnectedbyrouters,thatpro-videsglobalconnectivityforputersandworksputers.Asecondsenseofthewordistheputerresourcesavailableoverthiswork.InterpolatedMarkovmodelAtypeofMarkovmodelofsequencesthatexaminessequencesforpatternsofvariablelengthinordertodiscriminatebestbetweengenesandnon-genesequences.（内部网）Intron（内含子）Non-codingregionofDNA.Iterative（反复的/迭代的）Asequenceofoperationsinaprocedurethatisperformedrepeatedly.Java（一种由SUNMicrosystem开发的编程语言）K（BLAST程序的一个统计参数）AstatisticalparameterusedincalculatingBLASTscoresthatcanbethoughtofasanaturalscaleforsearchspacesize.ThevalueKisusedinconvertingarawscore(S)toabitscore(S').K-tuple（字/字长）Identicalshortstretchesofsequences,alsocalledwords.lambda（λ，BLAST程序的一个统计参数）AstatisticalparameterusedincalculatingBLASTscoresthatcanbethoughtofasanaturalscaleforscoringsystem.Thevaluelambdaisusedinconvertingarawscore(S)toabitscore(S').LAN（局域网）Localwork.Likelihood（似然性） 137 《生物信息学札记》樊龙江 Thehypotheticalprobabilitythataneventwhichhasalreadyurredwouldyieldaspecifice.Unlikeprobability,whichreferstofutureevents,likelihoodreferstopastevents.LineardiscriminantanalysisAnanalysisinwhichastraightlineislocatedonagraphbetweentwosetsofdatapointsinalocationthatbestseparatesthedatapointsintotwogroups.Localalignment（局部联配）Attemptstoalignregionsofsequenceswiththehighestdensityofmatches.Indoingso,oneormoreislandsofsubalignmentsarecreatedinthealignedsequences.Logoddsscore（概率对数值）Thelogarithmofanoddsscore.SeealsoOddsscore.LowComplexityRegion(LCR)（低复杂性区段）Regionsofpositionincludinghomopolymericruns,short-periodrepeats,andmoresubtleoverrepresentationofoneorafewresidues.TheSEGprogramisusedtomaskorfilterLCRsinaminoacidqueries.TheDUSTprogramisusedtomaskorfilterLCRsinnucleicacidqueries.Machinelearning（机器学习）Thetrainingofputationalmodelofaprocessorclassificationschemetodistinguishbetweenalternativepossibilities.Markovchain（马尔可夫链）Describesaprocessthatcanbeinoneofanumberofstatesatanygiventime.TheMarkovchainisdefinedbyprobabilitiesforeachtransitionurring;thatis,probabilitiesoftheurrenceofstatesjgiventhatthecurrentstateisspSubstitutionsinnucleicacidandproteinsequencesaregenerallyassumedtofollowaMarkovchaininthateachsitechangesindependentlyoftheprevioushistoryofthesite.Withthismodel,thenumberandtypesofsubstitutionsobservedoverarelativelyshortperiodofevolutionarytimecanbeextrapolatedtolongerperiodsoftime.Inperformingsequencealignmentsandcalculatingthestatisticalsignificanceofalignmentscores,sequencesareassumedtobeMarkovchainsinwhichthechoiceofonesequencepositionisnotinfluencedbyanother.Masking（过滤）AlsoknownasFiltering.Theremovalofrepeatedorplexityregionsfromasequenceinordertoimprovethesensitivityofsequencesimilaritysearchesperformedwiththatsequence.Maximumlikelihood(phylogeny,alignment)（最大似然法）Themostlikelye(treeoralignment),givenaprobabilisticmodelofevolutionarychangeinDNAsequences.Maximumparsimony（最大简约法）Theminimumnumberofevolutionarystepsrequiredtogeneratetheobservedvariationinasetofsequences,asfoundparisonofthenumberofstepsinallpossibleictrees.Methodofmoments 138 《生物信息学札记》樊龙江 Themeanorexpectedvalueofavariableisthefirstmomentofthevaluesofthevariablearoundthemean,definedasthatnumberfromwhichthesumofdeviationstoallvaluesiszero.Thestandarddeviationisthesecondmomentofthevaluesaboutthemean,andsoon.MinimumspanningtreeGivenasetofrelatedobjectsclassifiedbysomesimilarityordifferencescore,themini-mumspanningtreejoinsthemost-alikeobjectsonadjacentouterbranchesofatreeandthensequentiallyjoinsless-alikeobjectsbymoreinwardbranches.Thetreebranchlengthsarecalculatedbythesameneighbor-joiningalgorithmthatisusedtobuildictreesofsequencesfromadistancematrix.Thesumoftheresultingbranchlengthsbetweeneachpairofobjectswillbeapproximatelythatfoundbytheclassificationscheme.MMDB（分子建模数据库）MolecularModellingDatabase.AtaxonomyassigneddatabaseofPDB(seePDB)files,andrelatedinformation.Molecularclockhypothesis（分子钟假设）Thehypothesisthatsequenceschangeatthesamerateinthebranchesofanevolutionarytree.MonteCarlo（蒙特卡罗法）Amethodthatsamplespossiblesolutionstoplexproblemasawaytoestimateamoregeneralsolution.Motif（模序）Ashortconservedregioninaproteinsequence.Motifsarefrequentlyhighlyconservedpartsofdomains.MultipleSequenceAlignment（多序列联配）Analignmentofthreeormoresequenceswithgapsinsertedinthesequencessuchthatresiduesmonstructuralpositionsand/orancestralresiduesarealignedinthesamecolumn.ClustalWisoneofthemostwidelyusedmultiplesequencealignmentprogramsMutationdatamatrix（突变数据矩阵，即PAM矩阵）Ascoringpiledfromtheobservationofpointmutationsbetweenalignedsequences.AlsoreferstoaDayhoffPAMmatrixinwhichthescoresaregivenaslogoddsscores.N50length（N50长度，即覆盖50%所有核苷酸的最大序列重叠群长度）Ameasureofthecontiglength(orscaffoldlength)containinga'typical'nucleotide.Specifically,itisthemaximumlengthLsuchthat50%ofallnucleotideslieincontigs(orscaffolds)ofsizeatleastL.Nats(naturallogarithm)Anumberexpressedinunitsofthenaturallogarithm.NCBI（美国国家生物技术信息中心）NationalCenterforBiotechnologyInformation(USA).CreatedbytheUnitedStatesCongressin1988,todevelopinformationsystemstosupportthe 139 《生物信息学札记》樊龙江 biologicalmunity.Needleman-Wunschalgorithm（Needleman-Wunsch算法）Usesdynamicprogrammingtofindglobalalignmentsbetweensequences.Neighbor-joiningmethod（邻接法）Clusterstogetheralikepairswithinagroupofrelatedobjects(e.g.,geneswithsimilarsequences)tocreateatreewhosebranchesreflectthedegreesofdifferenceamongtheobjects.work（神经网络）Fromartificialintelligencealgorithms,techniquesthatinvolveasetofmanysimpleunitsthatholdsymbolicdata,whichareinterconnectedbyworkoflinksassociatedwithnumericweights.Unitsoperateonlyontheirsymbolicdataandontheinputsthattheyreceivethroughtheirconnections.Mostworksuseatrainingalgorithm(seeBack-propagation)toadjustconnectionweights,allowingworktolearnassociationsbetweenvariousinputandoutputpatterns.SeealsoFeed-forwardwork.NIH（美国国家卫生研究院）NationalInstitutesofHealth(USA).Noise（噪音）Insequenceanalysis,asmallamountofrandomlygeneratedvariationinsequencesthatisaddedtoamodelofthesequences;e.g.,ahiddenMarkovmodelorscoringmatrix,inordertoavoidthemodeloverfittingthesequences.SeealsoOverfitting.Normaldistribution（正态分布）Thedistributionfoundformanytypesofdatasuchasbodyweight,size,andexamscores.Thedistributionisabell-shapedcurvethatisdescribedbyameanandstandarddeviationofthemean.Localsequencealignmentscoresbetweenunrelatedorrandomsequencesdonotfollowthisdistributionbutinsteadtheextremevaluedistributionwhichhasamuchextendedtailforhigherscores.SeealsoExtremevaluedistribution.ObjectManagementGroup(OMG)（国际对象管理协作组）Anot-for-profitcorporationthatwasformedtoponent-basedsoftwarebyintroducingstandardizedobjectsoftware.TheOMGestablishesindustryguidelinesanddetailedobjectmanagementspecificationsinordertoprovidemonframeworkforapplicationdevelopment.WithinOMGisaLifeSciencesResearchgroup,aconsortiumrepresentingpanies,academicinstitutions,softwarevendors,andhardwarevendorswhoareworkingtogethertomunicationandinter-operabilityputationalresourcesinlifesciencesresearch.SeeCORBA.Object-orienteddatabase（面向对象数据库）Unlikerelationaldatabases(seeentry),whichuseatabularstructure,object-orienteddatabasesattempttomodelthestructureofagivendatasetascloselyaspossible.Indoingso,object-orienteddatabasestendtoreducetheappearanceofduplicateddataandplexityofquerystructureoftenfoundinrelationaldatabases. 140 《生物信息学札记》樊龙江 Oddsscore（概率/几率值）Theratioofthelikelihoodsoftwoeventsores.Insequencealignmentsandscoringmatrices,theoddsscoreformatchingtwosequencecharactersistheratioofthefrequencywithwhichthecharactersarealignedinrelatedsequencesdividedbythefrequencywithwhichthosesametwocharactersalignbychancealone,giventhefrequencyofurrenceofeachinthesequences.Oddsscoresforasetofindividuallyalignedpositionsareobtainedbymultiplyingtheoddsscoresforeachposition.Oddsscoresareoftenconvertedtologarithmstocreatelogoddsscoresthatcanbeaddedtoobtainthelogoddsscoreofasequencealignment.OMIM（一种人类遗传疾病数据库）OnlineMendelianInheritanceinMan.Databaseoficdiseaseswithreferencestomolecularmedicine,cellbiology,biochemistryandclinicaldetailsofthediseases.Optimalalignment（最佳联配）Thehighest-scoringalignmentfoundbyanalgorithmcapableofproducingmultiplesolutions.Thisisthebestpossiblealignmentthatcanbefound,givenanyparameterssuppliedbytheusertothesequencealignmentprogram.ORF（开放阅读框）OpenReadingFrame.Aseriesofcodons(basetriplets)whichcanbetranslatedintoaprotein.Therearesixpotentialreadingframesofanunidentifedsequence;TBLASTN(seeBLAST)transalatesanucleotidesequenceinallsixreadingframes,intoaprotein,thenattemptstoaligntheresultstosequenecesinaproteindatabase,returningtheresultsasanucleotidesequence.Themostlikelyreadingframecanbeidentifiedusingon-linesoftware(e.g.ORFFinder).Orthologous（直系同源）Homologoussequencesindifferentspeciesthatarosefrommonancestralgeneduringspeciation;mayormaynotberesponsibleforasimilarfunction.Apairofgenesfoundintwospeciesareorthologouswhentheencodedproteinsare60-80%identicalinanalignment.Theproteinsalmostcertainlyhavethesamethree-dimensionalstructure,domainstructure,andbiologicalfunction,andtheencodinggeneshaveoriginatedfrommonancestorgeneatanearlierevolutionarytime.Twoorthologs1andIIingenomesAandB,respectively,maybeidentifiedwhenpletegenomesoftwospeciesareavailable:(1)inadatabasesimilaritysearchofalloftheproteomeofBusingIasaquery,IIisthebesthitfound,and
(2)Iisthebesthitwhen11isusedasaqueryoftheproteomeofB.Thebesthitisthedatabasesequencewiththehighestexpectvalue(E).Orthologyisalsopredictedbyaverycloseicrelationshipbetweensequencesorbyaclusteranalysis.ComparetoParalogs.SeealsoClusteranalysis.Outputlayer（输出层）Thefinallayerofaworkinwhichsignalsfromlowerlevelsinworkareinputintooutputstateswheretheyareweightedandsummedto 141 《生物信息学札记》樊龙江 giveanoutputsignal.Forexample,theoutputsignalmightbethepredictionofonetypeofproteinsecondarystructureforthecentralaminoacidinasequencewindow.OverfittingCanurwhenusingalearningalgorithmtotrainamodelsuchasaorhid-denMarkovmodel.Overfittingreferstothemodelingtoohighlyrepresentativeofthetrainingdataandthusnolongerrepresentativeoftheoverallrangeofdatathatissupposedtobemodeled.Pvalue（P值/概率值）Theprobabilityofanalignmenturringwiththescoreinquestionorbetter.Thepvalueiscalculatedbyrelatingtheobservedalignmentscore,S,totheexpecteddistributionofHSPscoresparisonsofrandomsequencesofthesamelengthpositionasthequerytothedatabase.ThemosthighlysignificantPvalueswillbethosecloseto0.PvaluesandEvaluesaredifferentwaysofrepresentingthesignificanceofthealignment.Pair-wisesequencealignment（双序列联配）Analignmentperformedbetweentwosequences.PAM（可接受突变百分率/可以观察到的突变百分率，它可作为一种进化时间单位） PercenteptedMutation.AunitintroducedbyDayhoffetal.toquantifytheamountofevolutionarychangeinaproteinsequence.1.0PAMunit,istheamountofevolutionwhichwillchange,onaverage,1%ofaminoacidsinaproteinsequence.APAM(x)substitutionmatrixisalook-uptableinwhichscoresforeachaminoacidsubstitutionhavebeencalculatedbasedonthefrequencyofthatsubstitutionincloselyrelatedproteinsthathaveexperiencedacertainamount(x)ofevolutionarydivergence.Paralogous（旁系同源）Homologoussequenceswithinasinglespeciesthatarosebygeneduplication.Genesthatarerelatedthroughgeneduplicationevents.Theseeventsmayleadtotheproductionofafamilyofrelatedproteinswithsimilarbiologicalfunctionswithinaspecies.Paralogousgenefamilieswithinaspeciesareidentifiedbyusinganindividualproteinasaqueryinadatabasesimilaritysearchoftheentireproteomeofanism.Theprocessisrepeatedfortheentireproteomeandtheresultingsetsofrelatedproteinsarethensearchedforclustersthataremostlikelytohaveaconserveddomainstructureandshouldrepresentaparalogousgenefamily.ParametricsequencealignmentAnalgorithmthatfindsarangeofpossiblealignmentsbasedonvaryingtheparametersofthescoringsystemformatches,mismatches,andgappenalties.AnexampleistheBayesblockaligner.PDB（主要蛋白质结构数据库之一）BrookhavenProteinDataBank.Adatabaseandformatoffileswhichdescribethe3Dstructureofaproteinornucleicacid,asdeterminedbyX-raycrystallographyornuclearicresonance(NMR)imaging.The 142 《生物信息学札记》樊龙江 moleculesdescribedbythefilesareusuallyviewedlocallybydedicatedsoftware,butcansometimesbevisualisedontheworldwideweb.Pearsoncorrelationcoefficent（Pearson相关系数）Ameasureofthecorrelationbetweentwovariablesthatreflectsthedegreetowhichthetwovariablesarerelated.Forexample,thecoefficientisusedasameasureofsimilarityofgeneexpressioninamicroarrayexperiment.SeealsoCorrelationcoefficient.PercentidentityThepercentageofthecolumnsinanalignmentoftwosequencesthatincludesidenticalaminoacids.Columnsinthealignmentthatincludegapsarenotscoredinthecalculation.Percentsimilarity（相似百分率）Thepercentageofthecolumnsinanalignmentoftwosequencesthatincludeseitheridenticalaminoacidsoraminoacidsthatarefrequentlyfoundsubstitutedforeachotherinsequencesofrelatedproteins(conservativesubstitutions).ThesesubstitutionsmaybefoundinanaminoacidsubstitutionmatrixsuchastheDayhoffPAMandHenikoffBLOSUMmatrices.Columnsinthealignmentthatincludegapsarenotscoredinthecalculation.Perceptron（感知器，模拟人类视神经控制系统的图形识别机)Aworkinwhichinputandoutputstatesaredirectlyconnectedwithoutinterveninghiddenlayers.PHRED（一种广泛应用的原始序列分析程序，可以对序列的各个碱基进行识别和质量评价） Awidelyputerprogramthatanalysesrawsequencetoproducea'basecall'withanassociated'qualityscore'foreachpositioninthesequence.APHREDqualityscoreofXcorrespondstoanerrorprobabilityofapproximately10-X/10.Thus,aPHREDqualityscoreof30correspondsto99.9%uracyforthebasecallintherawread.PHRAP（一种广泛应用的原始序列组装程序）Awidelyputerprogramthatassemblesrawsequenceintosequencecontigsandassignstoeachpositioninthesequenceanassociated'qualityscore',onthebasisofthePHREDscoresoftherawsequencereads.APHRAPqualityscoreofXcorrespondstoanerrorprobabilityofapproximately10-X/10.Thus,aPHRAPqualityscoreof30correspondsto99.9%uracyforabaseintheassembledsequence.icstudies（系统发育研究）PIR（主要蛋白质序列数据库之
一，翻译自GenBank）AdatabaseoftranslatedGenBanknucleotidesequences.PIRisaredundant(seeRedundancy)proteinsequencedatabase.Thedatabaseisdividedintofourcategories:PIR1-Classifiedandannotated.PIR2-Annotated.PIR3-Unverified.PIR4-Unencodedoruntranslated.Poissondistribution（帕松分布）Usedtopredicttheurrenceofinfrequenteventsoveralongperiodoftime 143 《生物信息学札记》樊龙江 orwhentherearealargenumberoftrials.Insequenceanalysis,itisusedtocalculatethechancethatonepairofalargenumberofpairsofunrelatedsequencesmaygiveahighlocalalignmentscore.Position-specificscoringmatrix(PSSM)（特定位点记分矩阵，PSI-BLAST等搜索程序使用） ThePSSMgivesthelog-oddsscoreforfindingaparticularmatchingaminoacidinatargetsequence.Representsthevariationfoundinthecolumnsofanalignmentofasetofrelatedsequences.Eachsubsequentmatrixcolumncorrespondstothenextcolumninthealignmentandeachrowcorrespondstoaparticularsequencecharacter(oneoffourbasesinDNAsequencesor20aminoacidsinproteinsequences).Matrixvaluesarelogoddsscoresobtainedbydividingthecountsoftheresidueinthealignment,dividingbytheexpectednumberofcountsbasedonposition,andconvertingtheratiotoalogscore.Thematrixismovedalongsequencestofindsimilarregionsbyaddingthematchinglogoddsscoresandlookingforhighvalues.Thereisnoallowanceforgaps.Alsocalledaweightmatrixorscoringmatrix.Posterior(Bayesiananalysis)AconditionalprobabilitybasedonpriorknowledgeandnewlyevaluatedrelationshipsamongvariablesusingBayesrule.SeealsoBayesrule.Prior(Bayesiananalysis)Theexpecteddistributionofavariablebasedonpreviousdata.Profile（分布型）Amatrixrepresentationofaconservedregioninamultiplesequencealignmentthatallowsforgapsinthealignment.Therowsincludescoresformatchingsequentialcolumnsofthealignmenttoatestsequence.Thecolumnsinclude