We first evaluated the power (sensitivity) of the method to identify enrichment of modest associations. We considered models in which there is low detection power with single SNP analysis. We varied as parameters gene set size, fraction of genes in the set assigned a causal SNP (referred to as causal genes), effect size of the causal SNPs, and total number of causal genes. We performed multiple computer simulations where small effect sizes were randomly assigned to SNPs near different fractions of genes in a given gene set, against a background of randomized DGI for all SNPs i. Power was estimated for a given set of parameters as the fraction of simulation runs in which gene set enrichment was detected (described in Materials and Methods).
