Extended Data Fig. 7. Identification of putative CRE-gene pairs.

a, Bar charts showing the fraction of predicted H3K27ac- and H3K27me3- associated cCRE-gene pairs supported by the CEMBA datasets. P-value, two-tailed Fisher’s exact test. b-e, Bar charts showing the numbers of cCREs per targeted genes for (b) H3K27ac- and (c) H3K27me3- associated cCRE-gene pairs, and the numbers of predicted targeted genes per cCRE for (d) H3K27ac- and (e) H3K27me3- associated cCRE-gene pairs. f, Representative genome browser view of Gad2 gene locus, both H3K27ac- and H3K27me3-associated cCREs were shown. TSS-proximal region is indicated with green box and cCREs are marked with blue (H3K27ac-specific), brown (H3K27me3-specific) or purple (shared) boxes. g, Top enriched de novo TF motifs and GO terms of cCREs in H3K27ac-specific, shared and H3K27me3-specific pairs. h, Stacked bar plots showing the fraction of genomic regions for cCREs with potential active and repressive functions. P-value, two-tailed Fisher’s exact test. i,j, Bar charts showing the distribution of the distance between cCRE and TSS of predicted target genes from (i) H3K27ac- and (j) H3K27me3- associated cCRE-gene pairs. k, Spearman’s correlations coefficients between reads densities of cCREs and promoters of putative target genes across cell types for H3K4me1 and H3K9me3. P-value, two-tailed Wilcoxon signed-rank test. The boxes were drawn from lower quartile (Q1) to upper quartile (Q3) with the middle line denote the median, whiskers with maximum 1.5 IQR. n = 22 cell types. l, Heatmap showing the histone modification levels at cCREs with potential repressive roles in expression of putative target gene. cCRE were grouped using K-means clustering based on histone modification levels. m, Heatmap showing the expression levels of corresponding putative target genes of cCREs in (l). n, Top enriched Gene Ontology terms for genes in (m) and the top enriched de novo motifs for cCREs from each group in (l).