Expression level, cellular compartment and metabolic network position all influence the average selective constraint on mammalian enzymes

We measure the association between selective constraint, estimated by the ratio of nonsynonymous (Ka) to synonymous (Ks) substitutions, and several, primarily metabolic, measures of gene function. We find significant differences between the selective constraints acting on enzyme-coding genes from different cellular compartments, with the nucleus showing higher constraint than genes from either the cytoplasm or the mitochondria. Among metabolic genes, the centrality of an enzyme in the metabolic network is significantly correlated with Ka/Ks. In contrast to yeasts, gene expression magnitude does not appear to be the primary predictor of selective constraint in these organisms.Our results imply that the relationship between selective constraint and enzyme centrality is complex: the strength of selective constraint acting on mammalian genes is quite variable and does not appear to exclusively follow patterns seen in other organisms.The rate and manner of evolutionary change has long been a matter of keen interest to biologists [1]. Kimura provided theoretical underpinnings to molecular evolution by relating rates of sequence substitution, population parameters and mutation rates [2,3]. Thus, Kimura's neutral theory [4] predicts that mutations having no fitness effect will become fixed in a population at a rate equal to the mutation rate. Such neutral mutations therefore provide a standard for measuring the action of natural selection: regions changing more slowly than neutral ones are inferred to be experiencing purifying selection (e.g., selective constraint), those changing more rapidly, adaptive evolution. While the relative contributions of genetic drift, adaptive evolution and purifying selection to population differentiation are still debated, [5], there is general agreement that the patterns of selection vary both across species as well as among genes in the same species [6].Regarding interspecific variation, Lynch and Conery [7] argue that much of the variation