Fig. 4.

Variation of REs among clades can explain η variation. (a) Simulations with RE values based on the Grantham matrix (see Materials and Methods). For each of the ten new RE sets at a given level of deviation from or shuffled from the original values, five replicate sequence evolution simulations are conducted and the corresponding η estimates are plotted. Different RE sets at each deviation level and from each independent shuffle are distinguished by different (randomly assigned) colors. (b) The η’s estimated from the 90 clades simulated using the corresponding RE values of the real clades are plotted against the η’s estimated from the real clades. Dots are colored by the corresponding taxonomic group of the clades, as shown in the legend. The dashed red line indicates y = x. The y-axis value of each dot is the mean estimate from ten replicate simulations. (c) The expected η computed from the estimated REs and codon frequencies of each clade is plotted against the η estimated by the likelihood method from the alignment of the clade. The dashed red line indicates y = x. The y-axis value of each dot is the mean estimate from ten replicate simulations.