@ARTICLE{10.3389/fgene.2017.00074, AUTHOR={Van den Eynden, Jimmy and Larsson, Erik}, TITLE={Mutational Signatures Are Critical for Proper Estimation of Purifying Selection Pressures in Cancer Somatic Mutation Data When Using the dN/dS Metric}, JOURNAL={Frontiers in Genetics}, VOLUME={8}, YEAR={2017}, URL={https://www.frontiersin.org/articles/10.3389/fgene.2017.00074}, DOI={10.3389/fgene.2017.00074}, ISSN={1664-8021}, ABSTRACT={Large cancer genome sequencing initiatives have led to the identification of cancer driver genes based on signals of positive selection in somatic mutation data. Additionally, the identification of purifying (negative) selection has the potential to identify essential genes that may be of therapeutic interest. The most widely used way of quantifying selection pressures in protein-coding genes is the dN/dS metric, which compares non-synonymous to synonymous substitution rates. In this study, we examine whether and how this metric is influenced by the mutational processes that have been active during tumor evolution. We use exome sequencing data from six different cancer types from The Cancer Genome Atlas (TCGA) and demonstrate that dN/dS in its basic form, where uniform base substitution probabilities are assumed, is in fact strongly biased by these mutational processes. This is particularly true in malignant melanoma, where the mutational signature is characterized by a high amount of UV-induced cytosine to thymine mutations at dipyrimidine dinucleotides. This increases the likelihood of random synonymous mutations occurring in hydrophobic amino acid codons, leading to reduced dN/dS ratios in genes encoding membrane proteins and falsely suggesting purifying selection in these genes. When this effect is corrected for by taking mutational signature-derived substitution probabilities into account, purifying selection was found to be limited and similar in all cancer types studied. Our results demonstrate that it is crucial to take mutational signatures into account when applying the dN/dS metric to cancer somatic mutation data.} }