Fig. 1: Systematic identification of transcriptionally deregulated PRC2⁺-CGI and PRC2⁻-CGI genes across human cancers.

a An integrated pipeline for identification of three different classes of CGI promoters. The cancer type information is listed in Supplementary Table 1. b The criteria specified in Fig. 1a were applied to each cancer type separately, and the numbers of CGI genes in each class are shown for each TCGA cancer type. c–e Expression boxplots and Integrative Genomics Viewer (IGV) plots show representative colon adenocarcinoma (COAD) genes for three classes: c hypermethylated PRC2⁺-CGI, d upregulated PRC2⁺-CGI and e upregulated PRC2⁻-CGI. ChIP-Seq data are from Roadmap and ENCODE projects. Expression and methylation data are from TCGA. For c–e, box plots indicate the median (middle line), 25th, 75th percentile (box) and 5th and 95th percentile (whiskers); n = 41 biologically independent nonmalignant samples and 456 colon tumor samples. f Three classes of genes identified in COAD show different H3K27me3 and H3K27ac patterns in normal colonic mucosa. The heatmap was ordered by H3K27me3 signal within each class. The barplots show the mean H3K27me3 signal, mean H3K27ac signal, mean expression in normal colonic mucosa and promoter fractions overlapping with lamina-associated domains (LADs). Black lines in the barplots show the trend for a moving window of 100 genes. g The fractions of three classes of genes overlapping with LADs across all cancer types.