Prioritizing Functional Variants in Genetic Association Studies. S. Sengupta, X. Wen, G. Abecasis Biostatistics, University of Michigan, Ann Arbor, MI.

   Genome-wide association studies, which examine millions of genetic variants in thousands of individuals, have identified many complex trait associated loci. Most of these loci include many strongly associated variants and, often, variants which are not tested for association but are known to be in strong linkage disequilibrium with the variants exhibiting strongest evidence of association. The large number of variants actually or potentially showing evidence for association in each locus makes it challenging to prioritize likely functional variants at each locus. We reasoned that causal variants for each trait might share certain genomic features. For example, causal variants for lipid traits might preferentially overlap transcription factor binding sites active in liver, where important steps in lipid metabolism take place. More generally, causal variants for many traits might be non-synonymous variants that alter protein coding sequences. We develop a hierarchical model that identifies genomic features enriched among many associated loci and uses this information to prioritize likely functional variants in each locus. Our models can be fitted to summary statistics from individual studies, making it convenient to incorporate in ongoing genome-wide association study meta-analysis that can include 100,000s of individuals distributed across dozens of studies. We evaluate our method using simulations and application to genome-wide association study data for a variety of metabolic traits.

You may contact the first author (during and after the meeting) at