Fig. 2

Genomic features of DREs identified across cancers. a Bar chart showing the number of promoter DRE-gene pairs (top) and distal DRE-gene pairs (bottom) identified from the TCGA cancer cohorts. Blue bars indicate the fraction of the DRE-gene pairs shared by more than one cancer type, while red bars indicate the fraction of the cancer-type specific DRE-gene pairs. b The negative and positive correlation between DRE methylation and its target gene expression are shown by red and blue, respectively. Promoter pairs are mainly negatively correlated (88% in total). c Representative results showing the preferred chromatin state of distal DREs in liver (HepG2) and breast (HMEC) cancer cell lines. The distal DREs for each cell line were inferred from those identified from the corresponding TCGA cohort, LIHC and BRCA, respectively, here in this example. The number of distal DREs were counted at each chromatin state, with the heatmap color and number indicating the enrichment p value of distal DREs in each state. Results for other cancer types can be found in Additional file 1: Figure S2. d Representative results in liver cancer cell line HepG2, showing increased chromatin signals for H3K27ac, H3K4me1, p300, and DNase I hypersensitivity at genomic regions surrounding the negatively correlated distal DREs, in contrast to increased chromatin signals for repressive marks H3K9me3 and H3K27me3 surrounding the positively correlated distal DREs. P values were calculated by t-test comparing the signals of negatively correlated distal DREs vs that of an all probes control. The results for other cancer cell lines can be found in Additional file 1: Figure S4. e Average conservation score of distal DRE flanking regions. PhastCons conservation score was in the range of 0–1 (non- to perfectly conserved). P values are calculated between distal DREs and all probes (as control) by t-test. f The precision of DRE-target pairs predicted by MICMIC was determined by calculating the positive predictive value (PPV) in comparison with other chromatin interaction data, including IM-PET, Hi-C, RAD21-cohesin, and ChIA-PET (see “Methods,” only negatively correlated pairs considered here)