Neighbor preferences of amino acids and context-dependent effects of amino acid substitutions in human, mouse, and dog

Abstract

Amino acids show apparent propensities toward their neighbors. In addition to preferences of amino acids for their neighborhood context, amino acid substitutions are also considered to be context-dependent. However, context-dependence patterns of amino acid substitutions still remain poorly understood. Using relative entropy, we investigated the neighbor preferences of 20 amino acids and the context-dependent effects of amino acid substitutions with protein sequences in human, mouse, and dog. For 20 amino acids, the highest relative entropy was mostly observed at the nearest adjacent site of either N- or C-terminus except C and G. C showed the highest relative entropy at the third flanking site and periodic pattern was detected at G flanking sites. Furthermore, neighbor preference patterns of amino acids varied greatly in different secondary structures. We then comprehensively investigated the context-dependent effects of amino acid substitutions. Our results showed that nearly half of 380 substitution types were evidently context dependent, and the context-dependent patterns relied on protein secondary structures. Among 20 amino acids, P elicited the greatest effect on amino acid substitutions. The underlying mechanisms of context-dependent effects of amino acid substitutions were possibly mutation bias at a DNA level and natural selection. Our findings may improve secondary structure prediction algorithms and protein design; moreover, this study provided useful information to develop empirical models of protein evolution that consider dependence between residues.

Figure 1

Neighbor preference pattern of the amino acid A. (A) Relative entropies of the neighboring sites of A of N- and C-termini. The vertical coordinate represents the value of relative entropy, and the horizontal coordinate represents the relative distance to the 0 site. The flanking sites of N-terminus are indicated by negative numbers and the flanking sites of C-terminus are indicated by positive numbers. The gray lines are the thresholds representing 0.001 significance level. The standard deviation of the relative entropy in each site is indicated by black error bars; (B) Boxplot of the 20 relative entropies calculated by at each flanking site of the amino acid A. The bold line in each box represents the median of the 20 values. The top and bottom lines of each box indicate the upper and lower quartiles, respectively. The upper and lower whiskers represent the largest data point which was less than the sum of the upper quartile plus 1.5 times the interquartile range (IQR), and the lowest data point which was greater than the lower quartile minus 1.5 IQR, respectively. In order to determine which amino acids were apparently preferred (or not preferred) in the neighboring sites, the outliers were represented by the corresponding one letter codes of amino acids.

Figure 2

Neighbor preference pattern of the amino acid C. (A) Relative entropies of the neighboring sites of C of N- and C-termini; (B) Boxplot of the 20 relative entropies at each flanking site of the amino acid C.

Figure 3

Neighbor preference pattern of the amino acid G. (A) Relative entropies of the neighboring sites of G of N- and C-termini; (B) Boxplot of the 20 relative entropies at each flanking site of the amino acid G.

Figure 4

Neighbor preference patterns of V, T, and E in different secondary structures. (A) Neighbor preference pattern of the amino acid V in α-helix; (B) Neighbor preference pattern of the amino acid T in β-strand; and (C) Neighbor preference pattern of the amino acid E in coil. (Boxplots of these amino acids can be found in Figures S2–S4).

Figure 5

Context-dependence pattern of S→P substitution. (A) Relative entropies of the neighboring sites of the S→P substitution of N- and C-termini; (B) Boxplot of the 20 relative entropies at each flanking site of the S→P substitution.

Figure 6

Context-dependence pattern of G→E substitution in α-helix and coil. (A) Context-dependence pattern of the G→E substitution in α-helix; (B) Context-dependence pattern of the G→E substitution in coil. (Boxplots of this amino acid substitution type in α-helix and coil can be found in Figures S6 and S7).