UC San Diego

Understanding the genetic basis of disease is important, not only, for understanding the molecular mechanisms driving a particular disease phenotype, but also for providing informative prognostic, and diagnostic markers, as well as allowing for the design of personalized therapeutic intervention. Identifying these causative genetic variants is a complex problem because of the relatively small level of risk some variants may contribute, the interplay of variants which may be neutral in isolation, population stratification in purely statistical identification of risk variants, and the overwhelming number of neutral variants present in any individuals genome or tumor genome. A number of computational methods for prioritization of risk factors have been developed, each with a large weakness due to efforts to form generalized predictions. In this dissertation, I describe a specialized prediction method, tailored towards identification of causative polymorphisms in the protein kinase gene family, and demonstrate its applicability to identification of polymorphisms involved in inherited disease as well as cancer. Chapter 1 describes the method itself, Chapter 2 describes its applicability to cancer, and Chapters 3 and 4 delve into further details of the contributions of some of the predictive attributes