Figure 5.

Augmented eQTL identification by refined thresholding for P-values in tissue- or cell-type-specific hotspots. The fold of enrichment for three benchmark sets in seven types of genomic regions defined by cCREs (CTCF: CTCF-only, K4m3: DNase-H3K4me3, enhD: distal enhancer-like signature, enhP: proximal enhancer-like signature, prom: promoter-like signature) and eQTL-hotspots (C1 and C2) were compared across 36 tissues or cell types. (A) eQTLs with allelic fold change (aFC), (B) fine-mapping eQTLs and (C) GWAS SNPs. (D) A Q-Q plot illustrates the deviation of the distributions of the original association test P-values for eSNPs in hotspot-C1 and C2 against C0 (expected) in blood. The x-axis represents the −log₁₀P of non-hotspot (C0) as ‘expected’, and the y-axis represents the observed −log₁₀P. The dashed lines correspond to −log₁₀FDR and with a slope equal to 1, the FDR was set to 0.001 (Methods). The y-coordinate of the intersection point of the quantiles of the observed P-values indicates the refined thresholds for significant eSNP associations. (E) The fold of enrichment for fine-mapping eQTL associations in the eSNP–eGene pairs based on four different calling schemes. (F) The fold of enrichment for GWAS SNPs in eSNP sets based on four different calling schemes. Each colored dot represents a tissue or cell type; P-values were calculated using the one-sided Wilcoxon rank-sum test and were adjusted using ‘bonferroni’. ∙, adjusted P < 0.1; ^*, adjusted P < 0.05; ^**, adjusted P < 0.01; ^***, adjusted P < 0.001; ^****, adjusted P < 0.0001; ns, not significant. Hotspot: eSNPs defined within each type of hotspot by refined thresholds; Same ranking, WG: eSNPs defined by the same number of top-ranking SNPs in the whole genome as that defined by hotspots; Empirical threshold, WG: eSNPs defined by the whole genome, empirical thresholds of 5.0 × 10^–8; FDR < 0.001, WG: eSNPs defined by the whole genome thresholds corresponding to the same level of FDR (0.001).