Skip to main content
Fig. 7 | Genome Biology

Fig. 7

From: 3DeFDR: statistical methods for identifying cell type-specific looping interactions in 5C and Hi-C data

Fig. 7

Characterization of performance of 3DeFDR-HiC method using simulated Hi-C data. a Heatmaps showing a single example loop in simulations generated using varying effect sizes. The difference between any heatmap and the baseline loop strength shown in the far-left panel becomes more pronounced as effect size increases. b MA plots resulting from analysis of simulations of two artificial conditions (“A” and “B”) generated using varying effect sizes, with red and blue points representing interactions called as differential by our method at a false discovery rate of 1%. No interactions are called differential when no loops are truly differential (effect size + 0%). The number of interactions called as differential increases with increasing effect size, though the true proportion of differential interactions remains fixed at 40% in the simulations shown here. c Receiver operating characteristic (ROC) curves showing performance of our method on simulations generated using varying effect sizes. Like in b, the true proportion of differential interactions remains fixed at 40%. The x-axis shows the false-positive rate (FPR), or one minus the specificity. The y-axis shows the true positive rate (TPR), or sensitivity. The area under the receiver operating characteristic curve (AUROC) for each curve is shown in parentheses in the legend. d False discovery rate (FDR) control curves showing FDR control characteristics of our method on simulations generated using varying effect sizes, colored as in (C). The x-axis shows a range of FDR thresholds, while the y-axis shows the actual FDR we observe in the differential calls made by our method at that FDR threshold. Methods that control FDR should stay below the dashed gray line. All FDR control curves should show an FDR of 60% at an FDR threshold of 100%, since only 40% of loops in each simulation are truly differential. e Power curves showing the proportion of truly differential interactions called differential by our method (y-axis) as a function of the FDR threshold used for thresholding (x-axis) in simulations generated using varying effect sizes, colored as in c

Back to article page