Phylogenomic clustering for selecting non-redundant genomes for comparative genomics

Motivation: Analyses in comparative genomics often require non-redundant genome datasets. Eliminating redundancy is not as simple as keeping one strain for each named species because genomes might be redundant at a higher taxonomic level than that of species for some analyses; some strains with different species names can be as similar as most strains sharing a species name, whereas some strains sharing a species name can be so different that they should be put into different groups; and some genomes lack a species name. Results: We have implemented a method and Web server that clusters a genome dataset into groups of redundant genomes at different thresholds based on a few phylogenomic distance measures.