Nonsynonymous, synonymous and nonsense mutations in human cancer-related genes undergo stronger purifying selections than expectation

Background: Nonsynonymous mutations change the protein sequences and are frequently subjected to natural selection. The same goes for nonsense mutations that introduce pre-mature stop codons into CDSs (coding sequences). Synonymous mutations, however, are intuitively thought to be functionally silent and evolutionarily neutral. Now researchers know that the optimized synonymous codon usage is advantageous in the speedy mRNA translation process. With the advent of NGS technique, the explosion of NGS data generated from the tumor tissues help researchers identify driver mutations in cancer-related genes, but relatively less attention is paid to the SNP data in healthy human populations when studying cancer. Methods: Here, we analyzed the publically available human SNPs. We classified these SNPs according to their functional and evolutionary categories. By simply dividing the human genes into cancer-related genes and other genes, we compared the features of nonsynonymous, synonymous and nonsense mutations in these two gene sets from multiple aspects. Results: We provided lines of evidence that the nonsynonymous, synonymous and nonsense mutations in cancer-related genes undergo stronger purifying selection when compared to the expected pattern in other genes. The lower nonsynonymous to synonymous ratio observed in cancer-related genes suggests the suppression of amino acid substitutions in these genes. The synonymous SNPs, after excluding those in splicing regions, exhibit preferred changes in codon usage and higher codon frequencies in cancer-related genes compared to other genes, indicating the constraint exerted on these mutations. Nonsense mutations are less frequent and located closer to stop codons in cancer-related genes than in other genes, which putatively minimize their deleterious effects. Conclusion: Our study demonstrated the evolutionary constraint on mutations in CDS of cancer-related genes without the requirement of data from cancer tissues or patients. Our work provides novel perspectives on interpreting the constraint on mutations in cancer-related genes. We reveal extra constraint on synonymous mutations in cancer-related genes which is related to codon usage bias and is in addition to the splicing effect.