Phylogenetic and Functional Assessment of Orthologs Inference Projects and Methods

Citation: Altenhoff AM, Dessimoz C ( Phylogenetic and Functional Assessment of Orthologs Inference Projects and Methods Adrian M. Altenhoff 0 Christophe Dessimoz 0 Jonathan A. Eisen, University of California Davis, United States of America 0 Institute of Computational Science, ETH Zurich, and Swiss Institute of Bioinformatics , Zu rich , Switzerland Accurate genome-wide identification of orthologs is a central problem in comparative genomics, a fact reflected by the numerous orthology identification projects developed in recent years. However, only a few reports have compared their accuracy, and indeed, several recent efforts have not yet been systematically evaluated. Furthermore, orthology is typically only assessed in terms of function conservation, despite the phylogeny-based original definition of Fitch. We collected and mapped the results of nine leading orthology projects and methods (COG, KOG, Inparanoid, OrthoMCL, Ensembl Compara, Homologene, RoundUp, EggNOG, and OMA) and two standard methods (bidirectional best-hit and reciprocal smallest distance). We systematically compared their predictions with respect to both phylogeny and function, using six different tests. This required the mapping of millions of sequences, the handling of hundreds of millions of predicted pairs of orthologs, and the computation of tens of thousands of trees. In phylogenetic analysis or in functional analysis where high specificity is required, we find that OMA and Homologene perform best. At lower functional specificity but higher coverage level, OrthoMCL outperforms Ensembl Compara, and to a lesser extent Inparanoid. Lastly, the large coverage of the recent EggNOG can be of interest to build broad functional grouping, but the method is not specific enough for phylogenetic or detailed function analyses. In terms of general methodology, we observe that the more sophisticated tree reconstruction/ reconciliation approach of Ensembl Compara was at times outperformed by pairwise comparison approaches, even in phylogenetic tests. Furthermore, we show that standard bidirectional best-hit often outperforms projects with more complex algorithms. First, the present study provides guidance for the broad community of orthology data users as to which database best suits their needs. Second, it introduces new methodology to verify orthology. And third, it sets performance standards for current and future approaches. - The identification of orthologs is an important problem in the field of comparative genomics. Many studies, such as gene function prediction, phylogenetic analyses, and genomics context analyses, depend on accurate predictions of orthology. A large variety of methods for predicting orthologs and the resulting databases have appeared in recent years [18]. But although the accuracy of the predictions highly impacts any downstream analyses, there are only few comparative studies of the quality of the different prediction algorithms [9,10]. This paucity can be attributed to at least three major challenges. The first challenge resides in the multiple and sometimes intrinsically conflicting definitions of orthology [1113]. The original definition of Fitch [14] is based on the evolutionary history of genes: two genes are orthologs if they diverged through a speciation event. On the other hand, given that orthologs often have similar function, many people uses the term orthologs to refer to genes with conserved function. Yet another definition is used in some studies of genome rearrangement, in which the ortholog refers, in the event of a duplication, to the original sequence [15], which remains in its genomic context. The second challenge resides in the difficulty of validating the predictions. Take the case of phylogenetic orthology. Gene tree inference can be a notoriously difficult task, but it is usually precisely in difficult cases that the performances of methods can be differentiated. Indeed, in simple cases, most methods perform equally well. Validation of the definition based on function is not easier: orthology is in this context arguably impossible to verify because there is no universally applicable, unequivocal definition of conserved function, that is, the required similarity in terms of regulation, chemical activity, interaction partners, etc. for two genes to qualify as orthologs often varies across studies. For instance, in some wet lab experiments [16,17], two genes are only considered orthologs if they have the ability to complement each others function. The third challenge is of practical nature: to compare the different orthology inference projects, their methods must either be replicated on a common set of data, or the results produced by the different databases must be mapped to each other for comparison. Replication is not always possible, because some projects depend on human curation, or are not documented in detail. Mapping data is complicated by the lack of homogeneity in the sources of genomic data used by the different projects. The resulting intersection sets are often relatively small and may not be representative. In the present article, we provide an in-depth comparison of the prediction from 11 major projects, including OMA [4], our own orthology inference effort. We try to address the aforementioned challenges by testing phylogenetic and functional definitions of orthologs, using a variety of tests. We took the approach of The identification of orthologs, pairs of homologous genes in different species that started diverging through speciation events, is a central problem in genomics with applications in many research areas, including comparative genomics, phylogenetics, protein function annotation, and genome rearrangement. An increasing number of projects aim at inferring orthologs from complete genomes, but little is known about their relative accuracy or coverage. Because the exact evolutionary history of entire genomes remains largely unknown, predictions can only be validated indirectly, that is, in the context of the different applications of orthology. The few comparison studies published so far have asssessed orthology exclusively from the expectation that orthologs have conserved protein function. In the present work, we introduce methodology to verify orthology in terms of phylogeny and perform a comprehensive comparison of nine leading ortholog inference projects and two methods using both phylogenetic and functional tests. The results show large variations among the different projects in terms of performances, which indicates that the choice of orthology database can have a strong impact on any downstream analysis. comparing the inferred orthologs available from the different projects, which required mapping the data between projects. The rest of this introduction provides a description of the projects retained here, a review on the representation of orthology in those projects so to provide a common basis for comparison, and fina (...truncated)