Fig. 5

Subtype-specific DMRs in esophageal cancer. A Cancer hypoDMRs were identified from the comparison between EAC and ESCC tumors. Regions with FDR < 0.05 and absolute delta methylation levels > 0.2 were identified as DMRs. B Density plots showing the size distribution of hypoDMRs; stacked bar plots displaying fractions of hypoDMRs that overlap with different genomic features. C, D Aggregation plots of ATAC-seq signals from esophageal cancer samples within EAC (C) or ESCC (D) hypoDMRs or random genomic regions (background), which contained 10 times randomly selected regions with the same CpG density. ATAC-seq signals were obtained from the TCGA and normalized with the CPM method. E, F Cistrome-GO enrichment analyses using EAC (E) or ESCC (F) hypoDMRs and upregulated genes in the corresponding subtype. Top 15 most significant pathways are shown. The number of genes enriched in each pathway is shown on the right. Expression datasets are from the TCGA ESCA project. G, H Transcription-factor-binding motif sequences were identified by the ELMER [47] method using EAC (G) or ESCC (H) hypoDMRs as the foreground and random regions as the background. The annotation of the TF family is from the TFClass database [48]. I, J The most strongly enriched TFs in EAC (GATA4) (I) and ESCC (TP63) (J) were chosen for the experimental validation, using TF ChIP-seq, H3K27ac ChIP-seq, and WGBS in matched cell lines. Peaks overlapping with subtype hypoDMRs are shown on the left; the percentages of overlapped peaks are expressed in the column plots. The pie charts at the upper left corner denote the proportion of peaks with TF binding motifs over all peaks overlapping with subtype hypoDMRs. Cell line ChIP-seq and WGBS datasets are listed in the “ Methods” section