Coevolutionary response of the melanocortin system to selection on target genes. The table shows the number of nucleotides of a target gene that belong to a codon under positive selection (in rows) and induced a coevolutionary response in the other nine genes of the melanocortin system (in columns). For example in the Table 1, 10 nucleotides, which are part of a codon under positive selection in MC1R, induced a change in 10 sites in MC5R during vertebrate evolution. For each target gene, we give the number of pairs of sites that coevolved (note that a given site within a codon can be implicated in more than one pair of sites). We also provide the number of sites that belong to a codon under positive selection of a target gene that induced an evolutionary response in at least one of the other genes of the melanocortin system (this number can be smaller than the number of pairs because the same site can be involved in several pairs). Finally, we give the length in nucleotides of each human target gene and the percentage (in relation to this sequence length) of the number of different sites that belong to a codon under positive selection of a target gene and that induced an evolutionary response, which indicates the mean percentage of sites that coevolved with other genes of the melanocortin system. The sequences used in the coevolution and selection analyses where trimmed to remove conserved sites and regions of the alignment containing ambiguities. The length of the nucleotide sequence is therefore shorter than the total length given in Supplementary Materials Table S2. To evaluate the robustness of our method in assessing coevolution, we applied the ΔAIC threshold based on the 0.95 percentiles of the null distribution of ΔAIC obtained by simulations (the percentiles 0.975 and 0.90 are reported in Supplementary Materials Table S3). Whether the frequency of number of pairs or the frequency of number of sites per gene was significantly different than the frequency estimated for the rest of the genes in the melanocortin system was tested using Pearson’s χ2 tests corrected for multiple testing using the Benjamini-Hochberg approach. Genes with frequencies in the number of pairs or the number of sites significantly above the frequency for the rest of the genes are denoted with an ‘a’ superscript and those with frequencies below are denoted with a ‘b’ superscript. No superscript denotes non-significant differences.