Title

Untangling the Effects of Codon Mutation and Amino Acid Exchangeability

Document Type

Conference Proceeding

Publication Date

12-1-2005

Description

Determining the relative contributions of mutation and selection to evolutionary change is a matter of great practical and theoretical significance. In this paper, we examine relative contributions of codon mutation rates and amino acid exchangeability on the frequencies of each type of amino acid difference in alignments of distantly related proteins, alignments of closely related proteins, and among human SNPs, using a model that incorporates prior estimates of mutation and exchangeability parameters. For the operational exchangeability of amino acids in proteins, we use EX, a measure of protein- level effects from a recent statistical meta-analysis of nearly 10,000 experimental amino acid exchanges. EX is both free of mutational effects and more powerful than commonly used "biochemical distance" measures (7). For distaflt protein relationships, mutational effects (genetic code, transition/transversion bias) and operational exchangeability (EX) account for roughly equal portions of variance in off-diagonal values, the complete model accounting for R2 - 0.35 of the variance. For human/chimpanzee alignments representing closely related proteins relationships, mutational effects (including CpG bias) account for 0.52 of the variance; adding EX to the model increases this to 0.67. For natural variation in human proteins, the variance explained by mutational effects alone, and by mutational effects and operational exchangeability are, respectively, 0.66 and 0.70 for SNPs in HGVBase, and 0.56 and 0.60 for disease-causing missense variants in HGMD. Thus, exchangeability has a stronger relative effect for distant protein evolution than for the cases of closely related proteins or of population variation. A more detailed model for the hominid data suggests that 1) there is a threshold in EX below which substitutions are highly unlikely to be accepted, corresponding to roughly 30 % relative protein activity; 2) selection against missense mutants is a slightly convex function of protein activity, not changing much as long as protein activity is low; and 3) the probability of disease-causing effects decreases nearly linearly with EX.

COinS