The Genom of Homo sapiens.pdf

Evolution of Eukaryotic Gene Repertoire and Gene Structure:Discovering the Unexpected Dynamics of Genome EvolutionI.B. ROGOZIN,* V.N. BABENKO,* N.D. FEDOROVA,* J. D. JACKSON,* A.R. JACOBS,*D.M. KRYLOV,* K.S. MAKAROVA,* R. MAZUMDER,* † S.L. MEKHEDOV,* B.G. MIRKIN, †A.N. NIKOLSKAYA,* B.S. RAO,* S. SMIRNOV,* A.V. SOROKIN,* A.V. SVERDLOV,*S. VASUDEVAN,* Y.I. WOLF,* J.J. YIN,* D.A. NATALE,* AND E.V. KOONIN**National Center for Biotechnology Information, National Library of Medicine, National Institutes of Health,Bethesda, Maryland and † School of Information Systems and Computer Science, Birkbeck College,University of London, London, WC1E 7HX, United KingdomCOMPARATIVE GENOMICS, EVOLUTIONARYCLASSIFICATION OF GENES, ANDPHYLETIC PATTERNSComparative genomics has already changed our understandingof genome evolution. In what might amount to aparadigm shift in evolutionary biology, genome comparisonshave shown that lineage-specific gene loss and horizontalgene transfer (HGT) are not freak incidents ofevolution but extremely common phenomena that, to alarge degree, have shaped the extant genomes, at leastthose of prokaryotes (Doolittle 1999; Gogarten et al.2002; Snel et al. 2002; Mirkin et al. 2003). The extent ofgene loss occurring in certain lineages of prokaryotes,particularly parasites, is astonishing: In some cases,>80% genes in the genome have been lost over ~200 millionyears of evolution (Moran 2002). Horizontal genetransfer is harder to document, but a strong case has beenmade for its extensive contribution to the evolution ofprokaryotes (Ochman et al. 2000; Koonin et al. 2001;Mirkin et al. 2003).Gene exchange between phylogenetically distant eukaryotesdoes not appear to be an important evolutionaryphenomenon. In contrast, the contribution of gene loss tothe evolution of eukaryotic genomes was probably substantial,although the level of genome fluidity observed inprokaryotes is unlikely to have been attained in eukaryoticevolution. A comparison of the genomes of twoyeasts, Saccharomyces cerevisiae and Schizosaccharomycespombe, showed that, in the S. cerevisiae lineage,up to 10% of genes have been lost since the divergence ofthe two species (Aravind et al. 2000). It appears likelythat, in eukaryotic parasites with small genomes, e.g., themicrosporidia, much more massive gene elimination hasoccurred (Katinka et al. 2001). In contrast, the extent ofgene loss in complex, multicellular eukaryotes remainsunclear, although the small number of unique genes in thehuman genome when compared to the mouse genome(and vice versa) suggests considerable stability of thegene repertoire (Waterston et al. 2002). Present address: Protein Identification Resource, Georgetown UniversityMedical Center, 3900 Reservoir Road, NW, Washington, DC20007.We are interested in quantitative analysis of the dynamicsof genome evolution. A prerequisite for suchstudies is a classification of the genes from the sequencedgenomes based on homologous relationships. The twoprincipal categories of homologs are orthologs and paralogs(Fitch 1970; Sonnhammer and Koonin 2002). Orthologsare homologous genes that evolved via verticaldescent from a single ancestral gene in the last commonancestor of the compared species. Paralogs are homologousgenes, which, at some stage of evolution, haveevolved by duplication of an ancestral gene. Orthologyand paralogy are two sides of the same coin because,when a duplication (or a series of duplications) occurs afterthe speciation event that separated the comparedspecies, orthology becomes a relationship between sets ofparalogs, rather than individual genes (genes that belongto orthologous sets are sometimes called co-orthologs)(Sonnhammer and Koonin 2002).Robust identification of orthologs and paralogs is criticalfor the construction of evolutionary scenarios, whichinclude, along with vertical inheritance, lineage-specificgene loss and, possibly, HGT (Snel et al. 2002; Mirkin etal. 2003). The algorithms for the construction of thesescenarios involve, in one form or another, tracing thefates of individual genes, which is feasible only when orthologs(including co-orthologs) are known. In principle,orthologs, including co-orthologs, should be identified byphylogenetic analysis of entire families of homologousproteins, which is expected to define orthologous proteinsets as clades (e.g., Sicheritz-Ponten and Andersson2001). However, for genome-wide protein sets, suchanalysis remains labor-intensive and error-prone. Thus,procedures have been developed for identification of setsof probable orthologs without explicit use of phylogeneticmethods. Generally, these approaches are based onthe notion of a genome-specific best hit (BeT), i.e., theprotein from a target genome, which has the greatest sequencesimilarity to a given protein from the querygenome (Tatusov et al. 1997; Huynen and Bork 1998).The central assumption here is that orthologs have agreater similarity to each other than to any other proteinfrom the respective genomes due to the conservation offunctional constraints. When multiple genomes are analyzed,pairs of probable orthologs detected on the basis ofCold Spring Harbor Symposia on Quantitative Biology, Volume LXVIII. © 2003 Cold Spring Harbor Laboratory Press 0-87969-709-1/04. 293