Detecting selection in the evolution of cancer genomes
Pethick, Joanna Margaret
MetadataShow full item record
Cancer is a disease of the genome, requiring mutation or epimutation of specific genes to develop. The subsequent progression of cancer and response to therapies is also dictated to some degree by new mutation and clonal selection on that novel variation. However, it is thought that the majority of somatic mutations that occur in cancer are inconsequential passengers, and only a subset of functionally important driver mutations are of importance for cancer biology. This project set out to adapt and apply well-established methods from the field of molecular evolution to measure the selective forces driving the development of cancers. The ultimate objective being an improved understanding of which mutations help or hinder the progression of a cancer. Somatic cancer mutations were identified through the analysis of paired tumour and non-tumour whole-exome sequence data from the same individual. Primary data from 1005 patients was processed and complemented with additional publicly available pre-processed somatic variant calls from 4728 patients. Tumours were classified by tissue of origin and also their spectrum of substitution mutations. An advanced evolutionary analysis framework was established, allowing somatic single nucleotide variant data to be analysed as traditional organismal DNA sequence. Estimates of amino acid changing (non-synonymous) and synonymous mutation rates were derived and maximum likelihood tests of selection applied to identify genes and regions of genes subject to selective pressure during oncogenesis. While the meta-analysis of all patients provided unprecedented power for such a study, more refined analyses based on the stratification of patients gave insights into the pathways of importance for specific tissues of origin. Additionally, stratification of patients by the relative frequencies of different mutation types in a tumour also provided insights into how mutation profile influences the sites, genes and pathways that are perturbed in the development of cancer. Of particular interest here, was to test the hypothesis that both (1.) mutation spectrum and (2.) tissue of origin, set the evolutionary trajectory of a cancer. Building on this I sought to estimate their relative contributions. During this work an unexpected, localised mutation pattern was discovered and subsequent analysis demonstrated some loci to be highly susceptible to small segmental deletions in a subset of cancers. In the absence of a justifiable model of neutral segmental deletion it was not possible to infer whether these major mutations could be considered passengers or drivers of cancer progression. In contrast, an advantage of the evolutionary approach applied to nucleotide substitutions in protein coding sequences is that there is a justified model of neutral evolution (synonymous changes). Using this approach, I have not only been able to detect genes harbouring putative cancer driver mutations, but have also found evidence for genes subject to purifying selection in cancers where potentially disruptive mutations appear to be deleterious to cancer progression. Such genes, if they are non-essential in the adult organism, could provide a novel type of target for anti-cancer therapeutics.