© 2005 Nature Publishing Group REVIEWSNATURE REVIEWS | GENETICS VOLUME 6 | MAY 2005 | 361Understanding phylogenetic relationships betweenorganisms is a prerequisite of almost any evolutionarystudy, as contemporary species all share a common his-tory through their ancestry.The idea of phylogeny fol-lows directly from the theory of evolution presented byCharles Darwin in The Origin of Species1: the only illus-tration in his famous book is the first representation ofevolutionary relationships among species, in the formof a phylogenetic tree.The subsequent enthusiasm ofbiologists for the phylogenetic concept is illustrated bythe publication of Ernst Haeckel’s famous ‘trees’as earlyas 1866 (REF. 2).Today, phylogenetics — the reconstruction of evolu-tionary history — relies on using mathematical meth-ods to infer the past from features of contemporaryspecies,with only the fossil record to provide a windowto the evolutionary history of life on our planet. Thisreconstruction involves the identification ofHOMOLOGOUSCHARACTERSthat are shared between different organisms,and the inference of phylogenetic trees from the com-parison of these characters using reconstruction meth-ods (BOX 1). The accuracy of the inference is thereforeheavily dependent on the quality of models for theevolution of such characters.Because the underlyingmechanisms are not yet well understood,reconstructingthe evolutionary history of life on Earth solely on thebasis of the information provided by living organismshas turned out to be difficult.Until the 1970s,which brought the dawn of molec-ular techniques for sequencing proteins and DNA,phylogenetic reconstruction was essentially based onthe analysis of morphological or ultrastructural char-acters.The comparative anatomy of fossils and extantspecies has proved powerful in some respects; forexample,the main groups of animals and plants havebeen delineated fairly easily using these methods.How-ever, this approach is hampered by the limited number ofreliable homologous characters;these are almost non-existent in microorganisms3and are rare even in complexorganisms.The introduction of the use of molecular data inphylogenetics4led to a revolution. In the late 1980s,access to DNA sequences increased the number ofhomologous characters that could be compared fromless than 100 to more than 1,000,greatly improving theresolving power of phylogenetic inference.A few genesbecame reference markers.In particular,owing to itsconsiderable degree of conservation across all organ-isms,the gene that encodes small subunit ribosomalRNA (SSU rRNA) was extensively used for the classifi-cation of microorganisms and allowed the recognitionof the Archaea as a third distinct domain of the tree ofPHYLOGENOMICS AND THERECONSTRUCTION OF THE TREE OF LIFEFrédéric Delsuc, Henner Brinkmann and Hervé PhilippeAbstract | As more complete genomes are sequenced, phylogenetic analysis is entering a newera — that of phylogenomics. One branch of this expanding field aims to reconstruct theevolutionary history of organisms on the basis of the analysis of their genomes. Recent studieshave demonstrated the power of this approach, which has the potential to provide answers toseveral fundamental evolutionary questions. However, challenges for the future have also beenrevealed. The very nature of the evolutionary history of organisms and the limitations of currentphylogenetic reconstruction methods mean that part of the tree of life might prove difficult, if notimpossible, to resolve with confidence.HOMOLOGOUS CHARACTERSHomologous characters arethose that are descended from acommon ancestor.Canadian Institute forAdvanced Research,Département de Biochimie,Centre Robert-Cedergren,Université de Montréal,Succursale Centre-Ville,Montréal, Québec H3C3J7,Canada.Correspondence to H.P.e-mail:[email protected]:10.1038/nrg1603© 2005 Nature Publishing Group PRIOR PROBABILITYThe probability of a hypothesis(or parameter value) withoutreference to the available data.This can be derived from firstprinciples, or based on generalknowledge or previousexperiments.NODENodes of phylogenetic treesrepresent taxonomic units.Internal nodes (or branches)refer to hypothetical ancestors,whereas terminal nodes (orleaves) generally correspond toextant species.INCONSISTENCYA phylogenetic reconstructionmethod is statisticallyinconsistent if it convergestowards supporting an incorrectsolution with increasingconfidence as more data isanalysed.HOMOPLASYIdentical character states (forexample,the same nucleotidebase in a DNA sequence) that arenot the result of commonancestry (not homologous), butthat arose independently indifferent ancestors by convergentmutations.CONVERGENCEThe independent evolution ofsimilar character states inevolutionarily distinct lineages.REVERSALThe independent reacquisitionof the ancestral character state ina given evolutionary lineage.362 | MAY 2005 | VOLUME 6 www.nature.com/reviews/geneticsREVIEWSIn this review,we describe current methods for phy-logenomic inference and discuss their merits and pitfallsin light of their recent application to diverse phylogeneticproblems.The recent improvements in the resolution ofthe tree of life owing to large-scale studies in each of thethree domains — Archaea,Bacteria and Eukaryota —are discussed.Despite holding considerable promise,thephylogenomic approach also has potential problems thatstem from the limitations of current phylogenetic recon-struction methods. We present examples of methodINCONSISTENCY that lead to tree-reconstruction artefacts,and tentatively propose solutions to these issues.Finally,we discuss the future potential of phylogenomics andspecifically address the issue of how to corroborateresults from phylogenomic analyses.Current methods for phylogenomics – an overviewThe two crucial steps of classical phylogenetic infer-ence — the identification of homologous charactersand tree reconstruction — are generally preserved inphylogenomics.Therefore,as for phylogenies that arebased on morphological data or single genes,the relia-bility of a phylogenomic tree depends on the qualityof the characters and the accuracy of the reconstruc-tion methods.Theoretically,reliable characters canbe considered as those that have undergone only afew changes over time (ideally, a single change).Multiple changes create HOMOPLASY (noise) in the formofCONVERGENCE and REVERSAL, masking the genuinephylogenetic signal9.The three main types of standard
View Full Document
Unlocking...