Fig. 2: HiC-DC+ accurately identifies enhancer–promoter interactions from H3K27ac HiChIP.

a Evaluation of method performance by per gene auPR values. For each target gene, we ranked the candidate regulatory elements tested by CRISPRi-FlowFISH in K562 cells¹¹ based on the significance (P value) of corresponding 5 kb HiChIP interactions (merged replicates of GSM2705043, GSM2705044, and GSM2705045) with the target promoter as estimated by each method (n = 22 genes; Methods). Centers of the boxes indicate median values, the lower and upper hinges correspond to the first and third quartiles and the upper (lower) whiskers extend from the hinge to the largest (smallest) value no further than 1.5 times the distance between the first and third quartiles. b Comparison of performance for identifying promoter–enhancer loops with large effect sizes on target gene expression as assessed by CRISPRi-FlowFISH. Each dot in the scatterplots represents one tested promoter–enhancer pair: true positives resulting in decreased target gene expression upon enhancer inactivation are shown in red, true positives resulting in increased target gene expression shown in blue, false positives in gray. c Promoter–enhancer loops identified by HiC-DC+ and FitHiChIP (FDR <0.01) are shown as arcs for PLP2 gene, along with all candidate enhancers tested by CRISPRi-FlowFISH¹¹ and signal tracks for K562 DNase-seq (GSM816655) and H3K27ac ChIP-seq (GSM733656). d Promoter–enhancer loops identified by HiC-DC+ and FitHiChIP (FDR < 10⁻⁷) are shown as arcs for FTL gene, along with all candidate enhancers tested by CRISPRi-FlowFISH and K562 DNase-seq and H3K27ac ChIP-seq signal tracks.