查看文章

Genome sequencing efforts, coupled with technological advances and cost reductions, have led to the discovery of an increasing number of disease-related genetic variants. For the vast majority of these variants there is no known molecular mechanism for how they are related to the disease. This problem is particularly evident for diseases with complex genotype-phenotype relationships, such as cancer. Fortunately, the parallel growth of data on protein families, structures, interactions, modifications, and other aspects of function, in addition to the development of new computational methods provide the means to predict or identify disease variant mechanism. In this thesis, I first present a systematic analysis of a large dataset of pan-cancer missense mutations to investigate whether positive selection of certain types of amino acid substitutions can reveal interaction-disrupting cancer driver mutations. Hundreds of mechanistically interesting variants were identified in both potentially novel cancer-associated proteins and well-established cancer driver genes. I discuss new insights and for some instances, attempt functional interpretations by integrating information on protein structure and interactions that suggest putative novel mechanisms that question the classical oncogene/tumour suppressor paradigm. There is a wealth of publicly available resources that already provide valuable information on all aspects that define gene and protein function. This information has been collected from thousands of experiments or publications and has usually been manually verified or predicted using new approaches. This means that interpreting variants can be …