Figure 1

Amino-acid mutation frequencies in human genes. (a) The expected mutation frequencies based on the neighbor-dependent nucleotide mutation rates. The expected mutation matrix represents the frequencies of amino-acid transitions in the absence of selection. (b) The nonsynonymous benign SNP frequencies using the SNP dataset of Stephens et al. [9]. (c) The genetic disease mutation frequencies based on the Mendelian Inheritance in Man (MIM) database [7]. (d) The interspecies mutation frequencies based on the PAM1 matrix [10]. In the matrices, each square represents a particular amino-acid to amino-acid mutation (for example, Val → Ala). The gray level of the matrix squares is proportional to the number of observed mutations. The matrices were normalized so that the sum over all mutation frequencies for each matrix is equal to 100. The y-axes of the matrices represent the original (wild-type) amino acids; the x-axes represent the mutant amino acids (created as a result of a single-nucleotide mutation). The amino acids (given in single-letter amino-acid code) were ordered along the axes according to the side-chain chemistry [42]: (C) sulfhydryl; (STPAG) small hydrophilic; (NDEQ) acid, acid amide and hydrophilic; (HRK) basic; (MILV) hydrophobic; (FYW) large hydrophobic/aromatic. As a result of the ordering, the mutations close to the matrix diagonal tend to be more conserved.