Grant: Operating Grant

In large-scale genetic association studies, thousands of letters of DNA code are measured for hundreds of people with a widespread disease and for hundreds of people without the disease. The goal of this research is to create mathematical methods of accurately determining which DNA letters are associated with the disease in the sense that people with a variation in a certain DNA letter are at higher or lower risk of contracting the disease. We will use the new methods to aid in the discovery of DNA letters associated with three widespread health problems: coronary artery disease, severe obesity, and colorectal cancer. That will be accomplished on the basis of data from several large-scale genetic association studies. We will determine what such discoveries reveal about the causes of the health problems, which can lead to prevention and treatment interventions. In addition, the methods we develop will be provided to other research groups to help them discover the causes of other widespread diseases. The methods currently used to interpret large-scale genetic association data were designed to avoid all false discoveries of association, but only at the expense of missing many detectable associations. Advances in methods of controlling the false discovery rate (FDR) increase the number of detected true associations by allowing false discoveries as long as the overall average proportion of discoveries that are false is below set rate such as 5%. Likewise, the local FDR (LFDR), as a probability that a given DNA letter is not associated with the health problem, would be highly informative for inferring causes of disease. We will develop reliable methods of estimating the LFDR and the strengths of association for new prevention and treatment interventions.