the UCSC RefGene (hg19) coordinates. Variants missing LD information were discarded. For the pathway analysis, we considered 1320 canonical pathways collected from the Molecular Signature Database 22 (manually curated pathways originating from Reactome 23, KEGG 24, and BioCarta 25 databases), and restricted the analysis to pathways comprising 10 to 300 genes. We chose a genome-wide significance level of 0.05, and used the Benjamini and Hochberg (BH) method 26 to correct the gene- and pathway based p-values for multiple testing. The BH method is expected to show better power relative to a Bonferroni correction, which is conservative when the number of tests is large and the tests are correlated 26, 27.
