We assessed this for trans-eQTLs as well. An important aspect to consider is that trans-eQTL SNPs might affect multiple genes. If these effects are substantial (either in effect size or the number of affected genes), it is likely that a certain PC will capture this. Removal of such PCs from the expression data will therefore unintentionally result in the inability to detect these trans-eQTLs. In order to avoid such false-negatives we first performed a QTL analysis on the first 50 PCs (that had been removed from the expression data for the cis-eQTL analysis) to assess whether some of these PCs are under genetic control (genome-wide analysis, controlling FDR at 0.05). We did this for the large HT12 and the smaller H8v2 expression data separately, as PCA had been applied independently to these datasets. We observed that out of the first 25 PCs in the HT12 data three PCs and in the H8v2 two PCs were to some extent genetically determined (r2>5%). This was different for PCAs 26–50 in the HT12 data: 11 PCs were under substantial genetic control (Figure S9a).
