Fig. 2

Training (BCCC) and validation (TCGA) datasets stratified based on DNA methylation data for disease predictor genes. a Sample level data for DNA methylation (GoldenGate) in the BCCC dataset. Results of SAM-supervised classification of ER/PR status from β values are shown. β values for disease predictors (one probe per gene listed in Table 3) are presented as a heat map. Data are shown grouped for hormone receptor-negative samples (N = 22) and hormone receptor-positive samples (N = 53). P values from the t test for the difference between ER/PR-negative and ER/PR-positive disease (Additional file 1: Table S1) are presented on the right. b Sample level data for DNA methylation in the TCGA. Mean β values (when multiple probes were present) are reported for Table 3 genes, stratifying samples according to ER/PR status. P values from the t test are shown on the right