The Exchangeability of Amino Acids in Proteins

Abstract

The comparative analysis of protein sequences depends crucially on measures of amino acid similarity or distance. Many such measures exist, yet it is not known how well these measures reflect the operational exchangeability of amino acids in proteins, since most are derived by methods that confound a variety of effects, including effects of mutation. In pursuit of a pure measure of exchangeability, we present (1) a compilation of data on the effects of 9671 amino acid exchanges engineered and assayed in a set of 12 proteins; (2) a statistical procedure to combine results from diverse assays of exchange effects; (3) a matrix of "experimental exchangeability" values EX ij derived from applying this procedure to the compiled data; and (4) a set of three tests designed to evaluate the power of an exchangeability measure to (i) predict the effects of amino acid exchanges in the laboratory, (ii) account for the disease-causing potential of missense mutations in the human population, and (iii) model the probability of fixation of missense mutations in evolution. EX not only captures useful information on exchangeability while remaining free of other effects, but also outperforms all measures tested except for the best-performing alignment scoring matrix, which is comparable in performance. M EASURES of the pairwise distance (or similarity) of rate at which mutation introduces new alleles (Kimura amino acids provide the basis for scoring schemes 1983). in the alignment of sequences (Henikoff and Henikoff Although mutational effects are rarely treated as im-1993) and in other types of comparative analysis (Wen portant phenomena in their own right, they appear to et Yang et al. 1998; Atchley et al. 2001; Alexbe extremely important. For instance, each nonidentiandre and Zhulin 2003). A great many such matrices cal amino acid pair can be assigned a "genetic code disexist: an incomplete listing available from the AAIndex tance" G ij ʦ {1, 2, 3} equal to the minimum number of database (Kawashima and Kanehisa 2000) includes 83 nucleotides that must be changed to switch from amino matrices of pairwise amino acid similarity or distance acid i to amino acid j; the different categories someand 494 indices of amino acid properties. Formally retimes are referred to as "singlet," "doublet," and "triplet" lated to these are various schemes to distinguish "conexchanges. The practical importance of genetic code disservative" from "radical" amino acid changes (Hughes tance is amply demonstrated by the effectiveness of Fitch's et al. 1990; Hughes 1992;Rand et al. 2000; Zhang 2000).matrix of "mutational" distance (Fitch 1966) as a source A tacit assumption has been that the ultimate yardof match scores for protein sequence alignment (Feng stick for measuring amino acid similarity is the propenet al. 1985). Furthermore, pairs of amino acids with the sity for evolutionary change from one amino acid to ansame genetic code distance may differ in the density other. However, evolutionary transition probabilities, of minimum-length muta...

Consensus Genome-Wide Expression Quantitative Trait Loci and Their Relationship with Human Complex Trait Disease

“…Expression effects for the eQTL data analyzed here are usually relatively small, with a median value of 2.2-fold change in the level of expression. In contrast, high-impact missense variants typically change in vivo activity of a protein by 5-to 10-fold, sometimes more (Yampolsky and Stoltzfus, 2005). Where both mechanisms are present in a locus, a high-impact mechanism does not necessarily dominate.…”

Correlating disease-related mutations to their effect on protein stability: A large-scale analysis of the human proteome

“…Pp correlation values increase when asymmetric substitution matrices are considered [Overington et al, 1992;Koshi and Golstein 1995;Yampolsky and Stoltzfus, 2005] and the best correlation is obtained with the environment-specific substitution matrix for accessible residues [Overington et al, 1992]. Indeed some 32 SAP types (23% of the total) corresponding to 16 mutation pairs have a Pp difference (|P(X → Y) -P(Y → X)|) >0.3, sufficient to include one of the two mutation types below or above the 0.5 perturbing threshold value (Fig.…”

Proteomic characterization of a temperature-sensitive conditional lethal in Drosophila melanogaster

“…However, fine-tuned processes are also especially vulnerable to expression of deleterious alleles. The most common cause for temperature sensitivity is suggested to be amino acid changes affecting activity and stability of proteins, which will severely affect efficiency of energy metabolism (Hochachka and Somero, 1968;Yampolsky and Stoltzfus, 2005). The two differentially regulated proteins, Adh-related and porin, are coded by genes located in the identified QTL region, known to segregate variation affecting expression of the lethal effect.…”