eQTL p-values, effect sizes, and standard errors were obtained by fitting a linear trend test regression between the expression of each gene and all variants 200 kilobases upstream and downstream from each probe. After filtering out the variants with MAF <0.001, monomorphic SNPs, multi-allelic SNPs (as reported in 1000 Genomes or in the Ensembl database) and variants not sufficiently well imputed (Rsq <0.3, as defined by minimac http://genome.sph.umich.edu/wiki/minimac) between both datasets, we applied our colocalisation procedure. We conducted conditional analysis on SNPs with p-values for the expression associations, and repeated the colocalisation test using expression data conditioned on the most significant SNP. The aim of this analysis is to explore whether additional signals for expression other than the main one are shared with the biomarker signal.
