Selective Effects on Highly Expressed Genes

Many different selective effects on DNA and proteins infl uence the frequency of codons and amino acids in coding sequences. Selection is often stronger on highly expressed genes. Hence, by comparing highand low-expression genes it is possible to distinguish the factors that are selected by evolution. It has been proposed that highly expressed genes should (i) preferentially use codons matching abundant tRNAs (translational effi ciency), (ii) preferentially use amino acids with low cost of synthesis, (iii) be under stronger selection to maintain the required amino acid content, and (iv) be selected for translational robustness. These effects act simultaneously and can be contradictory. We develop a model that combines these factors, and use Akaike’s Information Criterion for model selection. We consider pairs of paralogues that arose by wholegenome duplication in Saccharmyces cerevisiae. A codon-based model is used that includes asymmetric effects due to selection on highly expressed genes. The largest effect is translational effi ciency, which is found to strongly infl uence synonymous, but not non-synonymous rates. Minimization of the cost of amino acid synthesis is implicated. However, when a more general measure of selection for amino acid usage is used, the cost minimization effect becomes redundant. Small effects that we attribute to selection for translational robustness can be identifi ed as an improvement in the model fi t on top of the effects of translational effi ciency and amino acid usage.