(median of discovery tissues, with each discovery tissue represented as the median of tests across all replication tissues; 90% for iPSC eQTLs). A high-LD proxy for the lead was tested 3.6% of the time (7.3% in iPSC), while the best available cis variant was tested only 1.1% of the time (2.7% in iPSC). Of the rare cases when the best available variant was tested, the selected variant was independent of the original eQTL effects (r2 < 0.1) 0.2% of the time (1.2% in iPSC), indicating that in the vast majority of cases, the same eQTL effect was tested for replication and the choice of variant is unlikely to have a marked effect on the results. Statistics of the replication of the iPSC eQTLs, including numbers of selected proxy variants, are provided in Supplementary Table 5b. If a gene was not tested for eQTLs in the query tissue (for e.g. due to gene not expressed), replication information was defined as missing. A replication profile was derived for each eGene in the discovery tissue, indicating whether the lead eQTL effect replicated (yes | no) or could not be tested (‘NA’). We then extracted eGenes, for which the lead eQTL effect did not
