Fig. 3: Footprint-C efficiently identifies chromatin structural features.

a Frequency of read mate orientation of different Hi-C datasets when distance <1 kb. b A screenshot showing Footprint-C contact heatmap with tracks from H3K27ac, ZNF143, KLF16, CTCF, RAD21 ChIP-seq, total RNA-seq, DNase-seq and Footprint-C 1D in K562. Gray shades mark the positions of non-CTCF loop anchors. c Number of loops (>20 kb) or stripes detected in Footprint-C, Micro-C, or in situ Hi-C datasets at different sampling size or by different methods. d Overlap of Mustache loops between Footprint-C & Micro-C datasets in K562. e Aggregate plots for common or Footprint-C specific loops. f Fraction of different loop types from common or Footprint-C specific loops. Enh, enhancer. Pro, promoter. g Size of common or Footprint-C specific structural A-A or B-B loops. Min: lower end of violin. Q1: lower bound of box. Q2: line in box. Q3: higher bound of box. Max: higher end of violin. P was calculated by two-sided Wilcoxon–Mann–Whitney test. Effect sizes are shown. h Footprint-C (right up) and Micro-C (left down) contact heatmaps at the same sampling size of 1800 M contacts showing one specific A-A (left) or B-B (right) structural loop, with tracks from genome annotation, PC1 value, CTCF, RAD21 or H3K9me3 ChIP-seq. Black squares mark the two loops. i, Aggregate plots of loops in DMSO or dTAG-7 treated HEK293T. The loops were from Footprint-C in WT HEK293T, and were aggregated at the center of a ±200 kb window at 5-kb resolution. The average enrichment score for loops was shown in top left corner. Source data are provided as a Source Data file.