Assessing Functional Impacts of Human Coding Variants

Advances in sequencing technology have made it routine to determine all coding variation in an individual human genome. A pressing challenge in the post-genomic era is to functionally characterize these variants, particularly within the disease-associated genes. Within the realm of cancer genome research, a critical problem that remains is how to separate the ‘driver’ from ‘passenger’ mutations and to further understand the functional mechanisms and consequences of driver mutations. I analyzed the missense somatic tumor mutations from 71 whole-genome or whole-exome sequencing studies across 21 cancers. I identified cancer-type-specific mutated domains and mutational hotspots. In some cases, I identified shifts in mutation and domain position between cancer types (but within a given gene product). I also provided clues to mutations’ functional effects. In addition to this, I identified different domain-centric mutational distribution patterns between oncoproteins and tumor suppressor proteins. The systematic correlation of mutations and cancer types at the domain level has the potential to guide more precise cancer treatments. Predictive models were also developed to quantify the impact of antigenicity on the spectrum of tumor missense somatic mutations. I found that somatic mutations are significantly depleted in peptides that are predicted to be displayed by MHC class I proteins, and characterized the dependence of this depletion on expression level. My results indicate that HLA class I alleles are, in general, incompletely dominant. I developed a model that produces an ‘antigenicity score’ for any input somatic coding mutation. These antigenicity scores could guide immunotherapy or aid in developing personalized cancer vaccines. In another collaborative effort to characterize human variation, I developed yeast-based functional assays to assess the functionality of the disease-associated coding variants. I evaluated the ability of wild-type human disease-associated genes to rescue homologous yeast mutants. Complementation between homologous human and yeast genes could often be found in the absence of annotated orthology, and these complementation relationships were of similar value as orthologous relationships for detecting human disease-causing variation. Finally, I found that the ability to detect pathogenic variation from complementation assays was not limited to variants which occur within the aligned region of human and yeast homology.Ph.D.2019-03-01 00:00:0