Hierarchical Clustering Procedure for Grouping Orthologous Domains in Multiple Genomes

Identification of orthologous genes, which are defined as homologous genes derived from speciation in the history of evolution [1], is a crucial step for comparative analysis of multiple genomes. Basically, evidence to identify orthologs includes: i) highest level of similarity, ii) conservation of gene arrangement on the chromosome, iii) consistency between gene phylogeny and species phylogeny. However, because of very large evolutionary distances as well as complex evolutionary events such as genome rearrangements and horizontal gene transfers, ortholog identification among various microbial genomes is not so trivial task. Many of the previous works (e.g. [2, 5]) mainly relied on the best hit relationships among all-against-all protein sequence comparisons between two genomes, but generally they required several additional procedures such as domain splitting, gene order comparison and phylogenetic consideration. On the other hand, numerous clustering procedures have been developed for homology domain identification (e.g. [4]), but generally they are not suitable for ortholog grouping. Here, I propose a hierarchical clustering approach which is flexible enough to incorporate many of the above ortholog criteria. Especially, the procedure is suitable for splitting genes into homologous domains minimally required for ortholog grouping.