We used the R package coloc (Giambartolomei et al., 2014) to assess statistical evidence of colocalization between each of four GLGC traits (LDL-C, HDL-C, TG, and TC) and each of eight top GLGC HLC eGenes (ANGPTL3, CPNE1, FRK, FUT2, IGF2R, SYPL2, UBE2L3, and VKORC1). We used P-values obtained from single-tissue eQTL analyses to calculate approximate Bayes factors as implemented in coloc. We estimated credible sets from these associations at 95% confidence as suggested previously (Wellcome Trust Case Control Consortium, 2012). We used minor allele frequencies estimated by the 1000 Genomes Project (Phase 3). All SNPs were compared and matched for genomic positions in GRCh37/hg19 coordinates and for reference and alternate alleles between the GWAS and eQTL summary statistics. Any mismatching SNPs were: (1) flipped to match strands; (2) allele-switched in cases where the effect allele of the GWAS SNP was not the effect allele of the corresponding eQTL SNP, with the corresponding beta estimate multiplied by −1; or (3) excluded if we could not resolve the discrepancies.
