Supplementary Figure 6: Characterizing and validating interactions identified by Trac-looping.

(a) Ratio of informative PETs associated with promoters of active genes to all informative PETs for data generated by the original Trac-looping (black bars), the PETs generated by the improved Trac-looping (red bars), and PETs from in-house Hi-C data and publicly available Hi-C data (blue bars). (b) Venn-diagram shows the comparison of called interactions from the number of PETs by Trac-looping and H3K4me2 ChIA-PET. (c) Cumulative distribution of number of predicted interaction by Trac-looping as a function of interaction distance. (d) Percentage of Trac-looping predicted interactions confirmed by alternative methods as a function of distance in resting CD4⁺ T cells. Alternative methods include PCHi-C (Javierre et al., 2016), H3K27ac Hi-ChIP (Mumbach et al., 2017), H3K4me2 CHIA-PET (Chepelev et al., 2012) and in situ Hi-C (Javierre et al., 2016). Predictions for Hi-ChIP and H3K4me2 ChIA-PET were made with Mango (bin size=2K, FDR<=0.05), the same method used for Trac-looping. Predictions for the Hi-C data were made with Homer (bin size=2K, FDR<=0.05). Predictions for PCHi-C are from published data. (e) Distribution of interactions predicted from 216M informative PETs, sorted by significance (FDR). (f) Saturation analysis shows interactions with high significance (FDR<1e-5) reached saturation by sequencing 60-80% of the total PETS, while interactions with low significance do not reach saturation at the current sequencing depth. (g) Percentage of functional interactions (enhancer-enhancer, enhancer-promoter and promoter-promoter) increases as the significance of the interactions increases.