Correlated selection on amino acid deletion and replacement in mammalian protein sequences

A low ratio of nonsynonymous and synonymous substitution rates (dN/dS) at a codon is a sign of functional constraint caused by purifying selection. Intuitively, the functional constraint would also be expected to prevent such a codon from being deleted. Oddly, to the best of our knowledge, the correlation between the rates of deletion and substitution has never actually been estimated. Here, we use 8,595 protein coding-region sequences from 9 mammalian species to examine the relationship between deletion rate and dN/dS. We found significant positive correlations at both the level of sites and genes. We compared our data against controls consisting of simulated coding sequences evolving along identical phylogenetic trees, where the correlation is not included in the model a priori. A much weaker correlation was found in the corresponding simulated sequences, which is probably caused by alignment errors. In the real data, the correlations cannot be explained by alignment errors. Separate investigations on nonsynonymous (dN) and synonymous (dS) substitution rates indicate that the correlation is most likely due to a similarity in patterns of selection rather than mutation rates.