A gene-based test was performed using VEGAS (Liu et al. 2010). Such tests can be more powerful than individual level SNP association because weaker signals from multiple causal variants in a gene will contribute evidence to gene significance whereas in GWAS these would likely be inseparable from random noise. Meta-analysis SNP P-values were used as the input, with the program assigning them to genes, and assessing the combined effect of all SNPs within a gene while accounting for SNP linkage disequilibrium. Almost 18,000 genes were tested, annotated to positions on the UCSC Genome Browser (hg18 assembly) which include regulatory regions located ±50 kb of 5´ and 3´ untranslated regions. Bonferroni significance was set at P<2.8×10−6 based on a correction for the number of genes tested.
