分子進(jìn)化樹(shù)構(gòu)建方法課件

第五章系譜分析生物信息學(xué)第五章生物信息學(xué)12.系統(tǒng)發(fā)生分析（Phylogeneticanalysis)分析基因或蛋白質(zhì)的進(jìn)化關(guān)系系統(tǒng)發(fā)生（進(jìn)化）樹(shù)（phylogenetictree）Atreeshowingtheevolutionaryrelationshipsamongvariousbiologicalspeciesorotherentitiesthatarebelievedtohaveacommonancestor.2.系統(tǒng)發(fā)生分析（Phylogeneticanal2經(jīng)典進(jìn)化生物學(xué)：比較：形態(tài)、生理結(jié)構(gòu)、化石分子進(jìn)化生物學(xué)：比較DNA和蛋白質(zhì)序列研究系統(tǒng)發(fā)生的方法經(jīng)典進(jìn)化生物學(xué)：研究系統(tǒng)發(fā)生的方法3Residuesthatarelinedupindifferentsequencesareconsideredtoshareacommonancestry(i.e.,theyarederivedfromacommonancestralresidue).AnAlignmentisanhypothesisofpositionalhomologybetweenbases/AminoAcidsEasyonlywithsubstitutionsDifficultalsowithindelsResiduesthatarelinedupin4=((A,(B,C)),(D,E))Newickformat節(jié)點(diǎn)Node分支BranchABCDE末端節(jié)點(diǎn)

可以是物種，群體，或者蛋白質(zhì)、DNA、RNA分子等OTU祖先節(jié)點(diǎn)/樹(shù)根Root系統(tǒng)發(fā)生樹(shù)術(shù)語(yǔ)內(nèi)部節(jié)點(diǎn)/分歧點(diǎn)該分支可能的祖先HTU=((A,(B,C)),(D,E))Newick5Aclade（進(jìn)化支）isagroupoforganismsthatincludesanancestorandalldescendentsofthatancestor.geneticchangenomeaningPhylogram CladogramtimeTaxonATaxonBTaxonCTaxonD111635TaxonATaxonBTaxonCTaxonDTaxonATaxonBTaxonCTaxonDUltrametrictree

超度量樹(shù)進(jìn)化樹(shù)分支樹(shù)系統(tǒng)發(fā)生樹(shù)術(shù)語(yǔ)Aclade（進(jìn)化支）isagroupoforg6Rootedtreevs.Unrootedtreetwomajorwaystoroottrees:ABCD102352d(A,D)=10+3+5=18Midpoint=18/2=9Bymidpointordistance有根樹(shù)ACBD無(wú)根樹(shù)系統(tǒng)發(fā)生樹(shù)術(shù)語(yǔ)outgroup外群、外圍支Rootedtreevs.Unrootedtreet7plantplantplantfungusanimalanimalanimalUnrootedtreerootRootedtreebacteriumanimalanimalanimalfungusplantplantplantMonophyleticgroupMonophyleticgroupRootedtreevs.Unrootedtreeplantplantplantfungusanimalani8選擇外群

（Outgroup）選擇一個(gè)或多個(gè)已知與分析序列關(guān)系較遠(yuǎn)的序列作為外類群外類群可以輔助定位樹(shù)根外類群序列必須與進(jìn)化樹(shù)上其它序列同源，但外類群序列與這些序列間的差異必須比這些序列之間的差異更顯著。eukaryoteeukaryoteeukaryoteeukaryotearchaeaarchaeaarchaeabacteriaoutgroup外群Howtorootatree?選擇外群

（Outgroup）選擇一個(gè)或多個(gè)已知與分析序列關(guān)9系統(tǒng)發(fā)育樹(shù)構(gòu)建步驟多序列比對(duì)（自動(dòng)比對(duì)、手工校正）選擇建樹(shù)方法（替代模型）建立進(jìn)化樹(shù)進(jìn)化樹(shù)評(píng)估最大簡(jiǎn)約法(maximumparsimony,MP)距離法(distance)最大似然法(maximumlikelihood,ML)貝葉斯法(Bayesianinference)統(tǒng)計(jì)分析BootstrapLikelihoodRatioTest……UPGMA鄰近法(Neighbor-joining,NJ)最小進(jìn)化法(minimumevolution)系統(tǒng)發(fā)育樹(shù)構(gòu)建步驟多序列比對(duì)（自動(dòng)比對(duì)、手工校正）選擇建樹(shù)方10距離法 距離法又稱距離矩陣法，首先通過(guò)各個(gè)序列之間的比較，根據(jù)一定的假設(shè)（進(jìn)化距離模型）推導(dǎo)得出分類群之間的進(jìn)化距離，構(gòu)建一個(gè)進(jìn)化距離矩陣。進(jìn)化樹(shù)的構(gòu)建則是基于這個(gè)矩陣中的進(jìn)化距離關(guān)系。CatDogRatDog3Rat45Cow676CatDogRat11224Cow計(jì)算序列的距離，建立距離矩陣通過(guò)距離矩陣建進(jìn)化樹(shù)距離法 距離法又稱距離矩陣法，首先通過(guò)各個(gè)序列之間的比較，11Step1.計(jì)算序列的距離，建立距離矩陣Uncorrected“p”distance(=observedpercentsequencedifference)Kimura2-parameterdistance(estimateofthetruenumberofsubstitutionsbetweentaxa)對(duì)位排列，去除空格（選擇替代模型）Step1.計(jì)算序列的距離，建立距離矩陣Uncorrect12由進(jìn)化距離構(gòu)建進(jìn)化樹(shù)的方法有很多，常見(jiàn)有：1.UnweightedPairGroupMethodwithArithmeticmean

(UPGMA)

2.Neighbor-JoiningMethod(NJ法/鄰位連接法)

3.MinimumEvolution(MP法/最小進(jìn)化法)Step2.通過(guò)矩陣建樹(shù)由進(jìn)化距離構(gòu)建進(jìn)化樹(shù)的方法有很多，常見(jiàn)有：Step2.通過(guò)13 最大簡(jiǎn)約法（MP）最早源于形態(tài)性狀研究，現(xiàn)在已經(jīng)推廣到分子序列的進(jìn)化分析中。最大簡(jiǎn)約法的理論基礎(chǔ)是奧卡姆（Ockham）哲學(xué)原則，對(duì)所有可能的拓?fù)浣Y(jié)構(gòu)進(jìn)行計(jì)算，找出所需替代數(shù)最小的那個(gè)拓?fù)浣Y(jié)構(gòu)，作為最優(yōu)樹(shù)。

最大簡(jiǎn)約法(MaximumParsimony)Findthetreethatexplainstheobservedsequenceswithaminimalnumberofsubstitutions 最大簡(jiǎn)約法（MP）最早源于形態(tài)性狀研究，現(xiàn)在已經(jīng)推廣到分14Sequence1TGCSequence2TACSequence3AGGSequence4AAG1

2

3PositionMP法建樹(shù)流程If1and2aregroupedatotaloffourchangesareneeded.

If1and3aregroupedatotaloffivechangesareneeded.

If1and4aregroupedatotalofsixchangesareneeded.Position1

(1,2):1change;

(1,3)or(1,4):2changesPosition2

(1,3):1change;

(1,2)or(1,4):2changesPosition3

(1,2):1change;

(1,3)or(1,4):2changesSequence1TGCSequence2TACSequen15456BESTMP法建樹(shù)步驟456BESTMP法建樹(shù)步驟16最大似然法(MaximumLikelihood)最大似然法(ML)最早應(yīng)用于對(duì)基因頻率數(shù)據(jù)的分析上。其原理為選取一個(gè)特定的替代模型來(lái)分析給定的一組序列數(shù)據(jù)，使得獲得的每一個(gè)拓?fù)浣Y(jié)構(gòu)的似然率都為最大值，然后再挑出其中似然率最大的拓?fù)浣Y(jié)構(gòu)作為最優(yōu)樹(shù)。最大似然法(MaximumLikelihood)最大似然17ML法建樹(shù)流程CAGATGCCATGCML法建樹(shù)流程CAGATGCCATGC18PickanEvolutionaryModelForeachposition,GenerateallpossibletreestructuresBasedontheEvolutionaryModel,calculateLikelihoodoftheseTreesandSumthemtogettheColumnLikelihoodforeachOTUcluster.CalculateTreeLikelihoodbymultiplyingthelikelihoodforeachpositionChooseTreewithGreatestLikelihoodInferringthemaximumlikelihoodtreePickanEvolutionaryModelInf19Holder&Lewis(2003)NatureReviewsGenetics4,275-284Bayesianinference:Whatistheprobabilitythatthemodel/theoryiscorrectgiventheobserveddata?Pr(T|D)MaximumLikelihood:

Whatistheprobabilityofseeingtheobserveddata(D)givenamodel/theory(T)?Pr(D|T)SpeedNoneedforbootstrapping構(gòu)建進(jìn)化樹(shù)的新方法——貝葉斯推斷

(Bayesianinference)與ML相比，BI的優(yōu)勢(shì)：Holder&Lewis(2003)NatureRev20ComparisonofMethodsDistanceMaximumparsimonyMaximumlikelihoodUsesonlypairwisedistancesUsesonlysharedderivedcharactersUsesalldataMinimizesdistancebetweennearestneighborsMinimizestotaldistanceMaximizestreelikelihoodgivenspecificparametervaluesVeryfastSlowVeryslowEasilytrappedinlocaloptimaAssumptionsfailwhenevolutionisrapidHighlydependentonassumedevolutionmodelGoodforgeneratingtentativetree,orchoosingamongmultipletreesBestoptionwhentractable(<30taxa,homoplasyrare)GoodforverysmalldatasetsandfortestingtreesbuiltusingothermethodsComparisonofMethodsDistanceM21Bioinformatics:SequenceandGenomeAnalysis,2ndedition,byDavidW.Mount.p254ChoosingaMethodforPhylogeneticPrediction/cgi/content/full/2008/5/pdb.ip49MolecularBiologyandEvolution200522(3):792-802Bioinformatics:SequenceandG22AssessingtreereliabilityPhylogeneticreconstructionisaproblemofstatisticalinference.Onemustassessthereliabilityoftheinferredphylogenyanditscomponentparts.Questions:(1)howreliableisthetree?(2)whichpartsofthetreearereliable?(3)isthistreesignificantlybetterthananotherone?Assessingtreereliability23Astatisticaltechniquethatusesintensiverandomresamplingofdatatoestimateastatisticwhoseunderlyingdistributionisunknown.評(píng)估進(jìn)化樹(shù)的可靠性——自展法（bootstrappingmethod）從排列的多序列中隨機(jī)有放回的抽取某一列，構(gòu)成相同長(zhǎng)度的新的排列序列重復(fù)上面的過(guò)程，得到多組新的序列對(duì)這些新的序列進(jìn)行建樹(shù)，再觀察這些樹(shù)與原始樹(shù)是否有差異，以此評(píng)價(jià)建樹(shù)的可靠性Astatisticaltechniquethatu24TheBootstrapComputationalmethodtoestimatetheconfidencelevelofacertainphylogenetictree.rat GAGGCTTATChuman GTGGCTTATCturtle GTGCCCTATGfruitfly CTCGCCTTTGoak ATCGCTCTTGduckweed ATCCCTCCGG 0123456789SamplerathumanturtlefruitflyoakduckweedInferredtreeMorereplicates(between100-1000)rat GGAAGGGGCThuman GGTTGGGGCTturtle GGTTGGGCCCfruitfly CCTTCCCGCCoak AATTCCCGCTduckweed AATTCCCCCT

0011222345Pseudosample1rat CCTTTTAAAThuman CCTTTTAAATturtle CCCCCTAAAT fruitfly CCCCCTTTTToak CCTTTCTTTTduckweed CCTTTCCCCG

4455567778Pseudosample2TheBootstrapComputationalme25自展法檢驗(yàn)流程Bootstrappingdoesn’treallyassesstheaccuracyofatree,onlyindicatestheconsistencyofthedata對(duì)ML法而言，自展法太耗時(shí)，可用aLRT法檢驗(yàn)進(jìn)化樹(shù)的可靠性Anisimova&Gascuel(2006)Syst.Biol.55(4):539-552自展法檢驗(yàn)流程Bootstrappingdoesn’tr26MSA程序可對(duì)任何序列進(jìn)行比對(duì)，選擇什么樣的序列進(jìn)行比對(duì)非常重要！！用于構(gòu)建進(jìn)化樹(shù)的序列必須是同源序列MSA是構(gòu)建分子進(jìn)化樹(shù)的關(guān)鍵步驟MSA程序可對(duì)任何序列進(jìn)行比對(duì)，選擇什么樣的序列進(jìn)行比對(duì)非常27分子進(jìn)化樹(shù)構(gòu)建（ClustalW）頁(yè)面下方顯示CladogramTree點(diǎn)擊“ShowasPhylogramTree”展示PhylogramTree不推薦：僅提供距離法建樹(shù)，且沒(méi)有進(jìn)行評(píng)估輸入比對(duì)后的序列（或上載Alignments文件）EBI的ClustalW2-phylogeny分析網(wǎng)頁(yè)http://www.ebi.ac.uk/Tools/phylogeny/clustalw2_phylogeny/分子進(jìn)化樹(shù)構(gòu)建（ClustalW）頁(yè)面下方顯示Cladog28看圖工具下載“Phyliptreefile”（ph文件）TreeView進(jìn)化樹(shù)編輯打印軟件（在http://taxonomy.zoology.gla.ac.uk/rod/treeview.html）輸入比對(duì)后的序列（或上載Alignments文件）用TreeView軟件打開(kāi)上述文件可以不同格式展示進(jìn)化樹(shù)（1、2、3）EBI的ClustalW2-phylogeny分析網(wǎng)頁(yè)看圖工具下載“Phyliptreefile”（ph文件）29PHYLIP/phylip.html免費(fèi)的集成進(jìn)化分析工具PAUP/商業(yè)軟件，集成的進(jìn)化分析工具M(jìn)EGA/免費(fèi)的圖形化集成進(jìn)化分析工具PHYMLhttp://atgc.lirmm.fr/phyml/最快的ML建樹(shù)工具PAMLhttp://abacus.gene.ucl.ac.uk/software/paml.htmlML建樹(shù)工具Tree-puzzle

http://www.tree-puzzle.de/較快的ML建樹(shù)工具M(jìn)rBayes/基于貝葉斯方法的建樹(shù)工具分子進(jìn)化分析軟件更多工具/phylip/software.htmlPHYLIPhttp://evolution.g30提供最大簡(jiǎn)約法（MP）、最大似然法（ML）和距離法三種建樹(shù)方法。其中距離法包括鄰接法（NJ）、最小進(jìn)化法（ME）和UPGMA三種算法。分子進(jìn)化樹(shù)構(gòu)建方法優(yōu)點(diǎn)：圖形界面，集序列查詢、比對(duì)、進(jìn)化樹(shù)構(gòu)建為一體，幫助文件詳盡，免費(fèi)/提供最大簡(jiǎn)約法（MP）、最大似然法（ML）和距離法三種建樹(shù)方31分子進(jìn)化樹(shù)構(gòu)建方法課件32Buffon(1707-1788)NaturalHistoryofAnimals

Buffon(1707-1788)33始祖鳥(niǎo)化石復(fù)原圖始祖鳥(niǎo)化石復(fù)原圖342.7%difference2.7%difference35xl,Xenopuslaevis;xt,Xenopustropicalis;gg,Gallusgallus;rn,Rattusnorvegicus;mm,Musmusculus;hs,Homosapiens.BMCEvolutionaryBiology20077:164xl,Xenopuslaevis;BMCEvolut36分子進(jìn)化樹(shù)構(gòu)建方法課件37分子進(jìn)化樹(shù)構(gòu)建方法課件38分子進(jìn)化樹(shù)構(gòu)建方法課件39由于同一位點(diǎn)多重替代（multiplesubstitution）的發(fā)生，觀測(cè)到的差異比實(shí)際替代數(shù)要小原始序列后代序列13mutations

=

3differencesDegreeofdivergenceTotalnumberofsubstitutions由于同一位點(diǎn)多重替代（multiplesubstituti40為了估算出正確的分歧時(shí)間（期望替代數(shù)），必須對(duì)觀測(cè)到的替代數(shù)進(jìn)行校正在進(jìn)化的任意時(shí)間點(diǎn)，任意位點(diǎn)的核苷酸都可能發(fā)生回復(fù)和平行突變。替代模型Substitutionmodel為了估算出正確的分歧時(shí)間（期望替代數(shù)），必須對(duì)觀測(cè)到的替代數(shù)41替代模型替代模型42第五章系譜分析生物信息學(xué)第五章生物信息學(xué)432.系統(tǒng)發(fā)生分析（Phylogeneticanalysis)分析基因或蛋白質(zhì)的進(jìn)化關(guān)系系統(tǒng)發(fā)生（進(jìn)化）樹(shù)（phylogenetictree）Atreeshowingtheevolutionaryrelationshipsamongvariousbiologicalspeciesorotherentitiesthatarebelievedtohaveacommonancestor.2.系統(tǒng)發(fā)生分析（Phylogeneticanal44經(jīng)典進(jìn)化生物學(xué)：比較：形態(tài)、生理結(jié)構(gòu)、化石分子進(jìn)化生物學(xué)：比較DNA和蛋白質(zhì)序列研究系統(tǒng)發(fā)生的方法經(jīng)典進(jìn)化生物學(xué)：研究系統(tǒng)發(fā)生的方法45Residuesthatarelinedupindifferentsequencesareconsideredtoshareacommonancestry(i.e.,theyarederivedfromacommonancestralresidue).AnAlignmentisanhypothesisofpositionalhomologybetweenbases/AminoAcidsEasyonlywithsubstitutionsDifficultalsowithindelsResiduesthatarelinedupin46=((A,(B,C)),(D,E))Newickformat節(jié)點(diǎn)Node分支BranchABCDE末端節(jié)點(diǎn)

可以是物種，群體，或者蛋白質(zhì)、DNA、RNA分子等OTU祖先節(jié)點(diǎn)/樹(shù)根Root系統(tǒng)發(fā)生樹(shù)術(shù)語(yǔ)內(nèi)部節(jié)點(diǎn)/分歧點(diǎn)該分支可能的祖先HTU=((A,(B,C)),(D,E))Newick47Aclade（進(jìn)化支）isagroupoforganismsthatincludesanancestorandalldescendentsofthatancestor.geneticchangenomeaningPhylogram CladogramtimeTaxonATaxonBTaxonCTaxonD111635TaxonATaxonBTaxonCTaxonDTaxonATaxonBTaxonCTaxonDUltrametrictree

超度量樹(shù)進(jìn)化樹(shù)分支樹(shù)系統(tǒng)發(fā)生樹(shù)術(shù)語(yǔ)Aclade（進(jìn)化支）isagroupoforg48Rootedtreevs.Unrootedtreetwomajorwaystoroottrees:ABCD102352d(A,D)=10+3+5=18Midpoint=18/2=9Bymidpointordistance有根樹(shù)ACBD無(wú)根樹(shù)系統(tǒng)發(fā)生樹(shù)術(shù)語(yǔ)outgroup外群、外圍支Rootedtreevs.Unrootedtreet49plantplantplantfungusanimalanimalanimalUnrootedtreerootRootedtreebacteriumanimalanimalanimalfungusplantplantplantMonophyleticgroupMonophyleticgroupRootedtreevs.Unrootedtreeplantplantplantfungusanimalani50選擇外群

（Outgroup）選擇一個(gè)或多個(gè)已知與分析序列關(guān)系較遠(yuǎn)的序列作為外類群外類群可以輔助定位樹(shù)根外類群序列必須與進(jìn)化樹(shù)上其它序列同源，但外類群序列與這些序列間的差異必須比這些序列之間的差異更顯著。eukaryoteeukaryoteeukaryoteeukaryotearchaeaarchaeaarchaeabacteriaoutgroup外群Howtorootatree?選擇外群

（Outgroup）選擇一個(gè)或多個(gè)已知與分析序列關(guān)51系統(tǒng)發(fā)育樹(shù)構(gòu)建步驟多序列比對(duì)（自動(dòng)比對(duì)、手工校正）選擇建樹(shù)方法（替代模型）建立進(jìn)化樹(shù)進(jìn)化樹(shù)評(píng)估最大簡(jiǎn)約法(maximumparsimony,MP)距離法(distance)最大似然法(maximumlikelihood,ML)貝葉斯法(Bayesianinference)統(tǒng)計(jì)分析BootstrapLikelihoodRatioTest……UPGMA鄰近法(Neighbor-joining,NJ)最小進(jìn)化法(minimumevolution)系統(tǒng)發(fā)育樹(shù)構(gòu)建步驟多序列比對(duì)（自動(dòng)比對(duì)、手工校正）選擇建樹(shù)方52距離法 距離法又稱距離矩陣法，首先通過(guò)各個(gè)序列之間的比較，根據(jù)一定的假設(shè)（進(jìn)化距離模型）推導(dǎo)得出分類群之間的進(jìn)化距離，構(gòu)建一個(gè)進(jìn)化距離矩陣。進(jìn)化樹(shù)的構(gòu)建則是基于這個(gè)矩陣中的進(jìn)化距離關(guān)系。CatDogRatDog3Rat45Cow676CatDogRat11224Cow計(jì)算序列的距離，建立距離矩陣通過(guò)距離矩陣建進(jìn)化樹(shù)距離法 距離法又稱距離矩陣法，首先通過(guò)各個(gè)序列之間的比較，53Step1.計(jì)算序列的距離，建立距離矩陣Uncorrected“p”distance(=observedpercentsequencedifference)Kimura2-parameterdistance(estimateofthetruenumberofsubstitutionsbetweentaxa)對(duì)位排列，去除空格（選擇替代模型）Step1.計(jì)算序列的距離，建立距離矩陣Uncorrect54由進(jìn)化距離構(gòu)建進(jìn)化樹(shù)的方法有很多，常見(jiàn)有：1.UnweightedPairGroupMethodwithArithmeticmean

(UPGMA)

2.Neighbor-JoiningMethod(NJ法/鄰位連接法)

3.MinimumEvolution(MP法/最小進(jìn)化法)Step2.通過(guò)矩陣建樹(shù)由進(jìn)化距離構(gòu)建進(jìn)化樹(shù)的方法有很多，常見(jiàn)有：Step2.通過(guò)55 最大簡(jiǎn)約法（MP）最早源于形態(tài)性狀研究，現(xiàn)在已經(jīng)推廣到分子序列的進(jìn)化分析中。最大簡(jiǎn)約法的理論基礎(chǔ)是奧卡姆（Ockham）哲學(xué)原則，對(duì)所有可能的拓?fù)浣Y(jié)構(gòu)進(jìn)行計(jì)算，找出所需替代數(shù)最小的那個(gè)拓?fù)浣Y(jié)構(gòu)，作為最優(yōu)樹(shù)。

最大簡(jiǎn)約法(MaximumParsimony)Findthetreethatexplainstheobservedsequenceswithaminimalnumberofsubstitutions 最大簡(jiǎn)約法（MP）最早源于形態(tài)性狀研究，現(xiàn)在已經(jīng)推廣到分56Sequence1TGCSequence2TACSequence3AGGSequence4AAG1

2

3PositionMP法建樹(shù)流程If1and2aregroupedatotaloffourchangesareneeded.

If1and3aregroupedatotaloffivechangesareneeded.

If1and4aregroupedatotalofsixchangesareneeded.Position1

(1,2):1change;

(1,3)or(1,4):2changesPosition2

(1,3):1change;

(1,2)or(1,4):2changesPosition3

(1,2):1change;

(1,3)or(1,4):2changesSequence1TGCSequence2TACSequen57456BESTMP法建樹(shù)步驟456BESTMP法建樹(shù)步驟58最大似然法(MaximumLikelihood)最大似然法(ML)最早應(yīng)用于對(duì)基因頻率數(shù)據(jù)的分析上。其原理為選取一個(gè)特定的替代模型來(lái)分析給定的一組序列數(shù)據(jù)，使得獲得的每一個(gè)拓?fù)浣Y(jié)構(gòu)的似然率都為最大值，然后再挑出其中似然率最大的拓?fù)浣Y(jié)構(gòu)作為最優(yōu)樹(shù)。最大似然法(MaximumLikelihood)最大似然59ML法建樹(shù)流程CAGATGCCATGCML法建樹(shù)流程CAGATGCCATGC60PickanEvolutionaryModelForeachposition,GenerateallpossibletreestructuresBasedontheEvolutionaryModel,calculateLikelihoodoftheseTreesandSumthemtogettheColumnLikelihoodforeachOTUcluster.CalculateTreeLikelihoodbymultiplyingthelikelihoodforeachpositionChooseTreewithGreatestLikelihoodInferringthemaximumlikelihoodtreePickanEvolutionaryModelInf61Holder&Lewis(2003)NatureReviewsGenetics4,275-284Bayesianinference:Whatistheprobabilitythatthemodel/theoryiscorrectgiventheobserveddata?Pr(T|D)MaximumLikelihood:

Whatistheprobabilityofseeingtheobserveddata(D)givenamodel/theory(T)?Pr(D|T)SpeedNoneedforbootstrapping構(gòu)建進(jìn)化樹(shù)的新方法——貝葉斯推斷

(Bayesianinference)與ML相比，BI的優(yōu)勢(shì)：Holder&Lewis(2003)NatureRev62ComparisonofMethodsDistanceMaximumparsimonyMaximumlikelihoodUsesonlypairwisedistancesUsesonlysharedderivedcharactersUsesalldataMinimizesdistancebetweennearestneighborsMinimizestotaldistanceMaximizestreelikelihoodgivenspecificparametervaluesVeryfastSlowVeryslowEasilytrappedinlocaloptimaAssumptionsfailwhenevolutionisrapidHighlydependentonassumedevolutionmodelGoodforgeneratingtentativetree,orchoosingamongmultipletreesBestoptionwhentractable(<30taxa,homoplasyrare)GoodforverysmalldatasetsandfortestingtreesbuiltusingothermethodsComparisonofMethodsDistanceM63Bioinformatics:SequenceandGenomeAnalysis,2ndedition,byDavidW.Mount.p254ChoosingaMethodforPhylogeneticPrediction/cgi/content/full/2008/5/pdb.ip49MolecularBiologyandEvolution200522(3):792-802Bioinformatics:SequenceandG64AssessingtreereliabilityPhylogeneticreconstructionisaproblemofstatisticalinference.Onemustassessthereliabilityoftheinferredphylogenyanditscomponentparts.Questions:(1)howreliableisthetree?(2)whichpartsofthetreearereliable?(3)isthistreesignificantlybetterthananotherone?Assessingtreereliability65Astatisticaltechniquethatusesintensiverandomresamplingofdatatoestimateastatisticwhoseunderlyingdistributionisunknown.評(píng)估進(jìn)化樹(shù)的可靠性——自展法（bootstrappingmethod）從排列的多序列中隨機(jī)有放回的抽取某一列，構(gòu)成相同長(zhǎng)度的新的排列序列重復(fù)上面的過(guò)程，得到多組新的序列對(duì)這些新的序列進(jìn)行建樹(shù)，再觀察這些樹(shù)與原始樹(shù)是否有差異，以此評(píng)價(jià)建樹(shù)的可靠性Astatisticaltechniquethatu66TheBootstrapComputationalmethodtoestimatetheconfidencelevelofacertainphylogenetictree.rat GAGGCTTATChuman GTGGCTTATCturtle GTGCCCTATGfruitfly CTCGCCTTTGoak ATCGCTCTTGduckweed ATCCCTCCGG 0123456789SamplerathumanturtlefruitflyoakduckweedInferredtreeMorereplicates(between100-1000)rat GGAAGGGGCThuman GGTTGGGGCTturtle GGTTGGGCCCfruitfly CCTTCCCGCCoak AATTCCCGCTduckweed AATTCCCCCT

0011222345Pseudosample1rat CCTTTTAAAThuman CCTTTTAAATturtle CCCCCTAAAT fruitfly CCCCCTTTTToak CCTTTCTTTTduckweed CCTTTCCCCG

4455567778Pseudosample2TheBootstrapComputationalme67自展法檢驗(yàn)流程Bootstrappingdoesn’treallyassesstheaccuracyofatree,onlyindicatestheconsistencyofthedata對(duì)ML法而言，自展法太耗時(shí)，可用aLRT法檢驗(yàn)進(jìn)化樹(shù)的可靠性Anisimova&Gascuel(2006)Syst.Biol.55(4):539-552自展法檢驗(yàn)流程Bootstrappingdoesn’tr68MSA程序可對(duì)任何序列進(jìn)行比對(duì)，選擇什么樣的序列進(jìn)行比對(duì)非常重要?。∮糜跇?gòu)建進(jìn)化樹(shù)的序列必須是同源序列MSA是構(gòu)建分子進(jìn)化樹(shù)的關(guān)鍵步驟MSA程序可對(duì)任何序列進(jìn)行比對(duì)，選擇什么樣的序列進(jìn)行比對(duì)非常69分子進(jìn)化樹(shù)構(gòu)建（ClustalW）頁(yè)面下方顯示CladogramTree點(diǎn)擊“ShowasPhylogramTree”展示PhylogramTree不推薦：僅提供距離法建樹(shù)，且沒(méi)有進(jìn)行評(píng)估輸入比對(duì)后的序列（或上載Alignments文件）EBI的ClustalW2-phylogeny分析網(wǎng)頁(yè)http://www.ebi.ac.uk/Tools/phylogeny/clustalw2_phylogeny/分子進(jìn)化樹(shù)構(gòu)建（ClustalW）