Associated Variants

Disease-Associated Variants from the NHGRI GWAS Catalog
We obtained trait-associated SNPs from the GWAS Catalog25 on November 5, 2013. We included only high-frequency (MAF > 5%) bi-allelic autosomal SNPs with a genome-wide significant (p < 5 × 10−8) association with any trait. We included only studies where Europeans contributed to the majority of the final samples to simplify LD calculations. We conducted LD calculations across the selected SNPs by using 379 EUR samples from the 1000 Genomes Project21 and only bi-allelic SNPs with at least five copies of the minor allele. We ensured that these SNPs were independent by randomly excluding one SNP for each pair of SNPs if r2 > 0.1 or if the distance between the SNPs was <100 kb. Finally, we excluded phenotypes with fewer than ten independent SNP associations after our filtering criteria. This resulted in 1,416 SNPs in our test set.

Variants Associated with Height, Rheumatoid Arthritis, and Breast Cancer
We used 689 SNPs associated with height,26 89 SNPs associated with rheumatoid arthritis (RA [MIM: 180300]) in Europeans alone or shared between Europeans and Japanese,27 and 69 SNPs associated with breast cancer (MIM: 114480).28

eQTLs
We assembled a set of 923,022 cis-eQTL SNPs associated with whole-blood gene expression at a FDR of 0.5.29 For each reported eQTL gene, we selected the single SNP most significantly associated with its expression and then performed LD pruning30 (by using a window of 1,000 SNPs, sliding by one SNP at a time, and excluding one SNP in a pair if r2 > 0.1) to ensure independence of the SNPs in our final SNP set. This resulted in a final dataset of 6,381 eQTL SNPs.