Many important health problems in modern society such as obesity, diabetes mellitus, heart disease, and polycistic kidney disease are partially influenced by genetic factors but are characterized by complex rather than simple Mendelian inheritance. One hallmark characteristic of such disorders is population heterogeneity in the causative influences on phenotypic variation. We seek to develop statistical methods that will capitalize upon population heterogeneity in the analysis of complex quantitative traits. Because unidentified subpopulations threaten the validity of genetic association studies, the presence of population heterogeneity has been considered a liability. However, conditioning on correctly identified subpopulations will greatly enhance the modeling of genotypic effects. Research into subpopulation identification and modeling will provide techniques that will 1) decrease potential confounding by population stratification, 2) increase statistical power to detect associations between marker loci and phenotypes, and 3) allow for a richer and more precise estimation of genetic effects in heterogeneous populations. To accomplish this goal, the following specific aims are proposed: 1) To examine the utility of probability-based clustering algorithms in identifying subpopulations; 2) To examine the utility of the distance based clustering algorithms in identifying subpopulations; 3) To develop and examine the properties of Weighted Empirical Bayes estimates of the genotypic effect size of candidate genes upon continuous phenotypes for the genetic subpopulations; 4) To compare the statistical properties of traditional Bayes estimators of genotypic effect with those of Weighted Empirical Bayes Estimators; 5) To develop software for subpopulation identification and weighted empirical Bayes estimation of subpopulation effects; and 6) To apply these methods to real datasets in obesity research.
These specific aims are complimentary to the candidate's, Dr. Redden, long term career goals of becoming a leader in the development and application of statistical genetics methods in the estimation and testing of associations between candidate genes and continuous phenotypes. To this end, training activities and a thorough career development plan for Dr. Redden have been proposed. At the end of this five year research initiative, Dr. Redden will be well positioned to conduct cross-disciplinary research in obesity, genetics, and statistics.
