Figure S7A shows that the number of significantly detected cis-eQTL probes increases two-fold when 50 PCs were removed from the expression data. There is a long plateau visible (around PC50), where the number of detected cis-eQTLs probes remains approximately constant, irrespective of removing for instance 10 fewer or 10 extra PCs (reported numbers in this figure also include false-positive eQTLs due to potential primer polymorphisms, as we here wanted to solely compare the performance of removing different numbers of PCs). Figure S7B shows that of the initial 5,950 significantly detected cis-eQTL probes (no PCs removed), 4,965 (83.5%) were still detected with 50 PCs subtracted. The 985 initially detected cis-eQTLs probes, yet no longer detected when 50 PCs had been removed from the expression data, all had a low significance (Figure S8). As we controlled the FDR at 0.05 in all analyses it is therefore likely that a considerable amount of these reflect false-positives. Figure S8C shows that for all the overlapping 4,965 detected cis-eQTLs probes between the different analyses, the allelic direction was identical, and effect size on expression correlate well (Pearson r = 0.95) although these were nearly always stronger after having subtracted 50 PCs.
