Fig. 8

Analyses of PMDs and hypoDMRs in pan-cancer datasets. A–C TCGA tumormap showing cancer type clusters (A), DNA methylation levels in ESCC- vs. EAC-specific PMDs (B), or in ESCC- vs. EAC-specific hypoDMRs (C). DNA methylation data were obtained from the TCGA project. The TCGA-based clustering scheme denotes pan-gastrointestinal cancers (COAD, READ, STAD, and EAC) and pan-squamous cancers (ESCC, HNSC, LUSC and a subset of CESC and BLCA) are shown A. The number of samples is 8915. The detailed study name of TCGA study abbreviations are listed in https://gdc.cancer.gov/resources-tcga-users/tcga-code-tables/tcga-study-abbreviations. D, E Dot plot quantification of the methylation differences in B and C, respectively. F t-SNE plots showing cancer type clusters, G ATAC-seq levels in ESCC- vs. EAC-specific PMDs or in H ESCC- vs. EAC-specific hypoDMRs across tumor samples. ATAC-seq data were downloaded from the TCGA project. The number of samples is 362. I, J Dot plots quantification of the ATAC-seq values in G and H, respectively. K–M Precision-recall curves for three-class classification characterized by the average methylation of subtype-specific PMDs (K), DMRs (L), or both (M) respectively. The training datasets were from panels B and C