Rational characterization of cancer variants in the eukaryotic protein kinase domain
McSkimming, Daniel Ian
MetadataShow full item record
Cancer is a family of diseases which are characterized by genomic instability and deregulation of cellular processes, like proliferation, cell cycle, and apoptosis. Resequencing studies have identified a vast number of somatic mutations, only a small fraction of which are believed to promote oncogenesis, tumor proliferation, and metastasis. The eukaryotic protein kinase domain functions as regulators of nearly every cellular process and is the most common domain found in proteins with cancer variants. Several methods have been developed which use sequence, structure and functional data to classify and prioritize cancer variants for experimental validation, on both the full proteome and the kinase domain, specifically. These methods provide a binary classification (e.g., disease associated vs not disease associated, deleterious vs tolerated), but provide little to no mechanistic insight in how the variants alter protein function. Here, I use and extend the Protein Kinase Ontology (ProKinO) to place cancer variants in the necessary context to understand their mechanistic effects. I use ProKinO, an integrative ontology, to predict the consequences of variants in functional regions of the kinase domain, discovering variants that both activate (EGFR R776C/H) and inactive (PKCβ D523N) the kinase domains of an oncogene and tumor suppressor, respectively. I develop a web-based interactive visualization, the Kinome Viewer (KinView), to enable integrative analysis of cancer variants in the kinase domain. Finally, I use machine learning methods to identify the conformational changes which best separate active from inactive kinase crystal structures. The developed resources and methods provide new insight into the mechanisms with which cancer variants alter the catalysis of phosphorylation, allowing for the rational characterization and prioritization of variants in the kinase domain.