Identifying Molecular Mechanisms of Disease Action in Association Studies
Genome wide association studies (GWAS) have emerged as a very successful tool for identifying genetic variants that are associated with disease. Many of the identified SNPs, however, have only modest effects on disease risk or heritability. There is an emerging need to move from the "one SNP at a time" paradigm to models that consider the effect of multiple variants across genes and pathways. Performing genome-wide searches of high order SNP combinations requires testing large number of hypotheses with often limited sample sizes, which leads to poor statistical power. Furthermore, the computational search becomes unmanageable for more than a few hundreds of SNPs. To circumvent these problems, we have designed an algorithm that examines the molecular mechanisms of action underlying the association between a genetic variant and a disease, by searching for conditional changes in the transcriptional activity of essential gene regulators. By considering the cumulative associations in a particular pathway, the algorithm can identify specific regulatory programs associated with a disease and provide insights into its molecular mechanisms. We have applied this method to neuroblastoma (NBL) — one of the most common solid tumors in children. We will present preliminary data pointing to intriguing associations with the BARD1 locus that harbors several SNPs associated to high risk NBL.