Depletion of somatic mutations in splicing-associated sequences in cancer genomes

One of the main goals of large scale cancer genome projects is to exhaustively identify all cancer driver mutations which can then be target for cancer therapy. Just as the ratio of nonsynonymous to synonymous substitutions (Ka/Ks, dN/dS) has been a cornerstone of comparative genetics for the identification of proteins under positive selection, so too cancer driver mutations are identified by looking for nonsynonymous mutations at high frequency compared to the presumed background mutation rate estimated from synonymous mutations rate. When comparing sequences between species this is now known not to be true: indeed more Ka/Ks > 1 peaks are due to purifying selection on synonymous sites than due to protein level positive selection. Is the same purifying selection happening in cancers? If it is, the long held assumption that the local synonymous rate is the local mutation rate would be wrong. We thus sought to test the hypothesis that selection against splice altering mutations occurs in cancers. After careful analysis of synonymous somatic mutations (SSMs) identified in exomes and whole genome sequencing data from over 4,000 tumours across 15 different cancer types which are publically available from The Cancer Genome Atlas (TCGA) and the International Cancer Genome Concortium (ICGC), we conclude that the local synonymous mutation profile is not an unbiased reflection of the local mutational profile which leads to identification of many false positives.