Fig. 4

Characteristics of epitype-specific methylation patterns. Using the discovery cohort, 39 CpG sites were selected as methylated in the basal-like epitype ET7, and 90 CpG sites were selected as methylated in the luminal epitype ET5 but not in ET7. For a given set of CpG sites (All/Methylated in ET7/Methylated in ET5), a bar indicates the fraction of CpG sites assigned to the respective chromatin state in (a) human embryonic stem cells (H1hESC) and (b) human mammary epithelial cells (HMEC). c Pearson’s correlations between gene expression and methylation levels in the validation cohort for the CpG sites selected as methylated in ET7 and ET5 (left). The numbers on top indicate the number of CpG sites matched to a gene with gene expression data. The average gene expression levels across 661 breast tumors in the validation cohort stratified by epitype and 106 normal breast tissue samples from The Cancer Genome Atlas for the CpG sites methylated in ET7 (center) and methylated in ET5 (right) and matched to a gene with gene expression data. GEX gene expression, CNV copy number variation