Hierarchical Modeling in Linkage Disequilibrium Mapping. G. Chen, E. Jorgenson, J. Witte. Institute for Human Genetics, UCSF, San Francisco, CA.
   The completion of the HapMap project and advances in high-throughput genotyping methods has made feasible genome-wide association (GWA) studies. Such studies generally evaluate the relationship between hundreds of thousands of single nucleotide polymorphisms (SNPs) and one or more phenotypes. A critical unresolved issue in GWAs is how to analyze the enormous amount of information generated in a manner that is most likely to detect causal variants. The conventional analysis approach entails estimating the association between each SNP (or multiple SNPs) and a phenotype, and then using the corresponding p-values to prioritize the results. This approach, however, ignores existing information about the SNPs, can lead to spurious results as well as suffer from low power. To address these issues, we propose here a hierarchical model: adding a prior model that incorporates known information about the SNPs into a conventional analysis. In particular, we develop a hierarchical model that uses the following information on the SNPs: 1) potential functionality; 2) whether non-synonymous; 3) evolutionary conservation; 4) previously associated or in linkage region; 5) LD with neighboring SNPs. We show empirically how integrating this information in a hierarchical model may improve the ability of GWAs to determine the location of causal variants.