A growing number of molecular evolutionary studies are estimating the proportion of adaptive amino acid substitutions (alpha) from comparisons of ratios of polymorphic and fixed DNA mutations. Here, we examine how violations of two of the model assumptions, neutral evolution of synonymous mutations and stationary base composition, affect alpha estimation. We simulated the evolution of coding sequences assuming weak selection on synonymous codon usage bias and neutral protein evolution, alpha = 0. We show that weak selection on synonymous mutations can give polymorphism/divergence ratios that yield alpha-hat (estimated alpha) considerably larger than its true value. Nonstationary evolution (changes in population size, selection, or mutation) can exacerbate such biases or, in some scenarios, give biases in the opposite direction, alpha-hat alpha. These results demonstrate that two factors that appear to be prevalent among taxa, weak selection on synonymous mutations and non-steady-state nucleotide composition, should be considered when estimating a. Estimates of the proportion of adaptive amino acid fixations from large-scale analyses of Drosophila melanogaster polymorphism and divergence data are positively correlated with codon usage bias. Such patterns are consistent with alpha-hat inflation from weak selection on synonymous mutations and/or mutational changes within the examined gene trees.
展开▼
