To assess what genomic factors are correlated with rates of protein evolution, we estimated partial correlation coefficients in the programming language r (v.2.13.0. [63], package ppcor.r ). We assessed the degree of partial correlation between protein divergence (log(dN/dS)), genetic divergence (mean FSTper gene), genetic diversity (pi, calculated in sites , expression specificity, log(expression level), and the absolute value of log(expression species 1/expression species 2). For genetic diversity (pi), we also considered using the Watterson estimator theta given that this parameter may have a different bias than pi . Both estimators (pi and Watterson) were strongly correlated (r2 = 0.96) and partial correlation coefficient results were largely the same for both. As such, we only report values of pi. We calculated FSTvalues for each SNP (according to using the r package hierfstat and then averaged it per gene.