High-throughput sequencing has proven to be an effective tool in the diagnosis of many cancers. Identifying driver mutations from passenger mutations however remains a significant challenge, with current approaches of limited use in distinguishing pathogenic from benign. Missense variants are particularly difficult to predict due to the vast range of effects these can have. Conservation-based approaches to predicting missense variant consequences are widely used, however these rely on the suitability and depth of aligned sequences, and may not identify regions of biological importance specific only to humans.
Using gnomAD[1], the largest database of human standing variation, we have created a sequence-based measure of intolerance to missense variation across over 18,000 unique human genes named the Missense Tolerance Ratio (MTR)[2]. We have shown that patient-ascertained variants preferentially clustered in intolerant, low scoring MTR regions.
To further improve upon the MTR estimates, we are quantifying missense intolerance over protein tertiary structures to create a more sensitive measure, and to identify important novel structural and functional features. We hypothesise that the MTR score can therefore be used in drugability site identification, by identifying features not previously identified as fundamental to the resulting product from a gene.