Figure 9.

Automated identification of HGT genes. We examined the highest Basic Local Alignment Search Tool (BLAST) scores of homologs within groups of genomes at increasing distances from Escherichia coli. If the BLAST score was substantially lower (by a factor of 1.3) in two consecutive groups relative to its best score in more distant genomes, then the gene was considered a candidate for horizontal gene transfer (HGT). Given such candidates, we then used a quartet test to determine whether the best hit from the more distant genome was actually more closely related to the E. coli gene than were the best hits from intermediate genomes. The quartet test confirmed HGT in 92% of these cases, and for 71% of the genes whose quartet topology indicated HGT the topology was strongly supported (P < 0.05, by Shimodaira-Hasegawa test in tree puzzle [54]). 'HPVS' refers to Haemophilus, Pasteurella, Vibrio, Shewanella, and related species.