My goal is to become an independent researcher in methodological development across the disciplines of statistical genetics, epidemiology, and nutritional epidemiology. Specifically, I want to focus on using Bayesian hierarchical models to model cancer risk while accounting for the uncertainty surrounding genetic, epidemiological and nutritional data and their interactions. This goal builds upon my previous statistical training, bayesian modeling and statistical genetics, but requires training in nutrition and nutritional epidemiology, cancer biology, and general epidemiology. At the end of this training period, I will be an established researcher in the nutrition and genetics of cancer prevention. To attain these goals, I have developed a comprehensive educational plan, including expert mentors consisting of a statistical geneticist, an epidemiologist specializing in lung cancer, a nutritional epidemiologist and a Bayesian statistician. The research I have proposed both reinforces my training and provides original, cutting edge methodologies to the field of cancer prevention. The research plan focuses on the following specific aims: 1) To perform a literature review of the current information about dietary patterns, environmental and genetic risk factors that influence the folate metabolism pathways as they relate to lung cancer risk. This review will further my knowledge in the field, and provide the field with a concentrated review of risk factors for lung cancer that relate to diet, folate intake and folate metabolism. 2) To develop Bayesian Hierarchical models to simultaneously identify diet and physical activity patterns and dietary components that associate with lung cancer risk. Using sophisticated models to identify dietary patterns consider nutrients as they naturally occur and are ingested as food. Additionally, we will build models that jointly consider genes, nutritional factors and environmental factors. 3) To develop Bayesian hierarchical models to identify genes and gene by diet interactions involved with folate metabolism that affect lung cancer risk. Bayesian hierarchical models provide sophisticated machinery to investigate interactions that may occur with or without related main effects. The Bayesian methodologies proposed consider sources of variation that frequentist models do not, and can provide more comprehensive risk models. The models and techniques developed can be applied to multiple cancers.
