Gene sequences useful for predicting relatedness of whole genomes in bacteria

Abstract

Thirty-two protein-encoding genes that are distributed widely among bacterial genomes were tested for the potential usefulness of their DNA sequences in assigning bacterial strains to species. From publicly available data, it was possible to make 49 pairwise comparisons of whole bacterial genomes that were related at the genus or subgenus level. DNA sequence identity scores for eight of the genes correlated strongly with overall sequence identity scores for the genome pairs. Even single-gene alignments could predict overall genome relatedness with a high degree of precision and accuracy. Predictions could be refined further by including two or three genes in the analysis. The proposal that sequence analysis of a small set of protein-encoding genes could reliably assign novel strains or isolates to bacterial species is strongly supported.