The goal of the proposed research training program is to provide me (Dr. Collin Melton) with additional training in areas that will accelerate my career development as I transition from a post-doctoral fellow in Dr. Michael Snyder's lab to an independent tenure track professor. The key elements of this plan are: Candidate: I have extensive training in experimental and computational approaches to studying biomedicine. Areas of additional focus for career development during the K99 mentored post-doctoral research phase include the acquisition of additional experimental skills and supplemental training in cancer biology, human genetics, human genomics, applied statistics, and parallel computing. Additionally, I will receive training in laboratory management, mentorship, and responsible conduct of research. This well-rounded plan will provide me with a skill set that will enable a facile transition from postdoctoral fellow to tenure track faculty. Environment: I have a valuable advisory committee with experts in the areas of genomics, genetics, and cancer biology to ensure my success in this training program and to guide me through the successful acquisition of a faculty job. These include my mentor Dr. Michael Snyder, my co-mentor Dr. James Ford and two advisors, Dr. Hanlee Ji and Dr. Anshul Kundaje. The environment at Stanford University in the Snyder lab and department of Genetics fosters productivity and collaboration with word class facilities, resources, and researchers. Research: My proposed research plan in cancer genomics is timely, relevant, and innovative. The majority of current research in cancer genomics has made groundbreaking progress in understanding the relevant DNA variation that occurs in coding regions of the genome; however, 97-98% of the human genome does not code for protein. This proposal focuses specifically on studying the regulatory regions of the human genome to identify, characterize, and interpret the impact of point mutations in these regulatory regions. The central hypothesis of this proposal is that point mutations in regulatory regions of the human genome drive cancer formation and the functional consequences of these mutations can be predicted using machine learning algorithms. Aim 1 proposes the statistical identification of regulatory regions which are mutated across cancer samples, Aim 2 proposes functional characterization of the prevalent mutations identified in Aim 1, and Aim 3 extends the analysis of characterizing the effects of mutations genome-wide through use of genomics approaches and proposes the use of machine learning to classify novel mutations as either disrupting, activating, or having no effect on regulatory element activity. Through its use of experimental datasets combined with predictive models for functional consequences of individual cancer variation, this research will further the goal of personalized genome interpretation for cancer therapy.