Abstract
Selection is expected to be more efficient in species that are more diverse because both the efficiency of natural selection and DNA sequence diversity are expected to depend upon the effective population size. We explore this relationship across a data set of 751 mammal species for which we have mitochondrial polymorphism data. We introduce a method by which we can examine the relationship between our measure of the efficiency of natural selection, the nonsynonymous relative to the synonymous nucleotide site diversity (π(N)/π(S)), and synonymous nucleotide diversity (π(S)), avoiding the statistical non-independence between the two quantities. We show that these two variables are strongly negatively and linearly correlated on a log scale. The slope is such that as π(S) doubles, π(N)/π(S) is reduced by 34%. We show that the slope of this relationship differs between the two phylogenetic groups for which we have the most data, rodents and bats, and that it also differs between species with high and low body mass, and between those with high and low mass-specific metabolic rate.