Figure 4

Average difference between two Q _i obtained from samples of genes within a genome ('within'), between sister genomes ('sister'), or between more distantly related genomes ('between'). Panel (a) shows results for simulated sequence evolution. Panels (b)-(h) show results from fully sequenced bacterial genomes; see Methods for details and a list of species included, (a) Distances for pairs randomly-generated matrices, using either the unrestricted model ('unrestricted') or by perturbing a starting matrix by a specified amount (100%, 50%, or 25% per element). (b) Results from one representative genome, PFO (see Methods). Series represent number of genes n used to estimate the rate matrix. By the time n increases to 10 genes (yellow series), the within-genome distance is a factor of 2–4 smaller than any between-genome distance, (c-h) Results for comparisons averaged over all genomes. As the sample size n increases from 1 to 50, the separation between the within-species and between-species distribution increases. This separation is highly significant within each species by n = 10 (for example, for PFO, the difference is about a factor of 2; P < 10^-300).