Our rationale was simple: if the ranked lists resulting from the two studies do not reflect true biologically meaningful patterns, the intersection of these studies would be equivalent to picking loci at random. In other words, the overlap of genes with marginal statistical support for an expression level difference between individuals with low or high lymphocyte counts on the one hand, and loci with marginal support for a genetic association on the other hand, would be no better than that expected by chance alone. Our findings, however, supported the notion that the overlap is higher than expected by chance. Indeed, we identified a set of 117 genes that are likely to play a role in determining variation in lymphocyte counts through changes in gene regulation. Because our test relied on choosing exactly one eQTL for each gene, our approach resulted in the inclusion of weakly associated SNPs with both gene expression and lymphocyte count. On the one hand, this approach was more resistant to possible sampling biases (e.g. due to spurious signals from gene- or SNP-dense regions of the genome).
