Figure 7. Non-metric multidimensional scaling analysis of amino acid word usage in microbial genomes divided by lifestyle (free-living vs. pathogenic) and gram-staining (+ vs −).

A linear discriminate analysis of the MDS coordinates was utilized for Monte Carlo, bootstrap iterations (10,000) of the separation among group centroids when observations are randomly distributed among groups. Probability values indicate the likelihood that the observed centroid separation could arise by random chance alone, and the (p<1e-06) values are indicated by the gray circles.