Identifying driver mutations in cancers

All cancers depend upon mutations in critical genes, which confer a selective advantage to the tumour cell. The key to understanding the contribution of a disease-associated mutation to the development and progression of cancer comes from an understanding of the consequences of that mutation on the function of the affected protein, and the impact on the pathways in which that protein is involved. Using data from over 30 different cancers from whole-exome sequencing cancer genomic projects, I analysed over one million somatic mutations. I identified mutational hotspots within domain families by mapping small mutations to equivalent positions in multiple sequence alignments of protein domains. I found that gain of function mutations from oncogenes and loss of function mutations from tumour suppressors are normally found in different domain families and when observed in the same domain families, hotspot mutations are located at different positions within the multiple sequence alignment of the domain. Next, I investigated the ability of seven prediction algorithms to discriminate between driver missense mutations in oncogenes and tumour suppressors. Using 19 features to describe these mutations, I then developed a random forest classifier, MOKCaRF, to distinguish between gain of function and loss of function missense mutations in cancer. MOKCaRF performs significantly better than existing algorithms. I then evaluated the ability of six existing prediction tools to distinguish between pathogenic and neutral mutations for both inframe insertion and inframe deletion mutations. I developed my own classifiers using 11 features that perform better than the current algorithms. Finally, using the algorithms that I developed, as well as changes in copy number and expression data for each gene, I analysed samples from 50 lung cancer patients to identify the actionable targets and potential new drug targets for each tumour.