To quantify the degree to which an xQTL SNP influences more than one molecular phenotype, we first identified the list of xQTL SNPs for a “discovery” phenotype and then estimated the π1 statistics of the SNP-feature associations for a “test” phenotype that share the same xQTL SNPs. Since an xQTL SNP might be tested for association with multiple cis features, e.g. an mQTL SNP was, on average, tested for association with 18 gene expression levels, a decision on which SNP-feature associations to include in the π1 estimation was necessary (see Supplementary Information). In particular, we examined the distance between each pair of “discovery” SNP and “test” feature, and found this distance to be a prime determinant of cross-phenotype sharing. For example, the strongest associated eQTL gene for each mQTL SNP is often the gene closest to the mQTL SNP (Figure 3C, Figure S5). Based on this observation, we estimated π1 to be 0.41–0.63 when we considered only the closest feature to each xQTL SNP (Figure 3D). Also, we examined the effect of window size by restricting the haQTL analyses to
