| Identifying the causes of complex diseases is the impetus of biomedical research. Today, with the advances in computational power and biotechnology, identifying the genetic determinants of these diseases is now a possibility. In this dissertation, we focus on developing computational methods to increase the power of genomic data sets to identify genes associated with complex diseases. First, we describe the basic methods used in statistical genetic research and the impact that microarray technology has had on the field. Then, we provide an introduction to the analysis of genome wide association studies. Following the introduction, we provide a simulation experiment to demonstrate how incorporating gene annotation into the analysis can increase the power to find gene associations with the disease phenotype. These simulations include multiple loci as the causal basis of the association, a scenario that is expected biologically but has been greatly ignored in previous simulation studies. Using these gene level associations, we perform gene set enrichment analyses to identify the biological pathways responsible for the disease, and use this data to identify the influential genes. Our results are compared across three datasets of Crohn's disease patients. Lastly, we switch to a linkage study and describe an algorithm to predict a clinical trait (obesity) given gene expression values. We show that high dimensional regression procedures provide greater power than removing missing data or replacing missing data with the mean of available data. With the completed (i.e., including the imputed values) data, we suggest the application of gene and pathway information to identify causal genes of obesity. |
