To analyze genetic association data at the level of genes and gene sets, we first needed to compute a gene score based on local SNPs. We assigned to each gene g in the genome a set of SNPs that lie within 110 kilobase (kb) upstream and 40 kb downstream of the gene's most extreme transcript boundaries, in attempt to capture signals from potential causal variants affecting regulatory elements, in addition to coding sequence (Figure 1A; see Materials and Methods for boundary choice). Each gene g is then assigned a score , defined in this instantiation as the most significant p-value among the association p-values of all individual SNPs i within the extended gene boundaries (Figure 1B). We used the best SNP rather than an average value, as we expect only one or a few associated variants per gene.
