Correlations between amino-acid preferences and frequencies in natural HIV sequences.

<p><b>(A)</b> Phylogenetic tree of the HIV-1 group-M sequences in the alignment. <b>(B)</b> Correlation between alignment frequencies and preferences. The preferences are the replicate averages re-scaled by the stringency parameter in <a href="http://www.plospathogens.org/article/info:doi/10.1371/journal.ppat.1006114#ppat.1006114.t002" target="_blank">Table 2</a>. <b>(C)</b> The correlation if evolution is simulated along the phylogenetic tree assuming that the preferences correctly describe the actual selection. <b>(D)</b> There is no correlation in a control simulation in which preferences are randomized among sites. <b>(E), (F)</b> Correlation between preferences and alignment frequencies as a function of relative solvent accessibility (RSA). Red lines show the actual correlation. Dark and light gray show the range of correlations in the middle 80% and 100% of 100 simulations. For both plots, data are shown until the subset of sites that meets the RSA cutoff becomes less than 10% of all sites in Env; this is why neither x-axis extends all the way from 0 to 1. Correlation coefficients are Pearson’s <i>R</i>.