Heatmap and boxplot of RPM signal and ChIP-seq signal at different 4C regions

4C regions are classified as gained, lost, and unchanged according to their presence in experiment versus control condition. For GSK343-treated K562 4C data, the experiment condition is GSK343-treated K562, while the control condition is DMSO-treated K562. Similarly, for CRISPR KO K562 4C data, the experiment condition is either FGF18-MRR1-A1 KO or IGF2-MRR2-A1 KO clones, while the control condition is the EV clones. Gained, 4C interactions only present in the experiment condition but not in the control condition; Lost, 4C interactions only present in the control condition but not in experiment condition; unchanged, 4C interactions present in both control and experiment conditions. In this comparison, only those 4C interactions with p value < 0.05 are considered.

For RPM signal heatmap, different types of 4C regions (gained/lost/unchanged) are first classified according to their H3K27ac/H3K27me3 ChIP-seq signal levels in the control condition. Tertiles are used to classify these 4C regions into high, medium, or low category. The 4C intensity in RPM of each 4C regions are shown in a color-scaled manner.

For the ChIP-seq signal heatmap, ChIP-seq signal different types of 4C regions (gained/lost/unchanged) are calculated as the area of signal at these regions. Deeptools computeMatrix⁸⁹ is used to calculated ChIP-seq signal at these 4C regions, and then the total signal areas are calculated as: Total Signal Area = sum (Sig *BS), where BS is the size of bins used when summarizing RPKM, and Sig is the ChIP-seq signal in RPKM at individual bin. For ChIP-seq signal boxplot, the same 4C region (gained/lost/unchanged) in different conditions are connected using gray line. Wilcoxon paired test are used, and p value are indicated accordingly: ns, p > 0.05, *p < = 0.05, **p < = 0.01, ***p < = 0.001, ****p < = 0.0001.