Mutation Patterns in Human Cancer and Coexpressed Genes in the Drosophila Genome

Abstract

For my dissertation I focused on two distinct projects. In the first project, I analyzed early cancer exome datasets to identify patterns of mutation that offer insight into somatic mutational processes in humans. This project began as a critique of the statistical methodology used to identify cancer-causing ("driver") genes in these datasets in one high-profile study. We then analyzed the available data more broadly, and showed that most mutations are not under selection and are therefore a product of the underlying mutational processes in the cancer tissue. Examining these processes in detail, we showed that increased CpG mutation frequency is not a byproduct of aberrant CpG island methylation and that an A->G/T->C associated with gene expression in the germline may be associated with transcription in some cancers. In the second project, I describe a novel approach for identifying clusters of coexpressed genes in eukaryotic genomes and our results applying this model to data from Drosophila melanogaster. We found that two-thirds of genes in the Drosophila genome are coexpressed with a neighbor. The boundaries of these coexpression clusters are enriched for insulator binding sites, and are correlated with physical interaction domains, suggesting that nuclear structure may play a role in coexpression. We hypothesize that coexpression segments may represent a type of substructure within these interaction domains.