Fig. 1

Classification of normal breast transcriptomes into Active and Inactive phenotypes and increased adipocyte association with Active samples. a Unsupervised clustering heat map (red = increased gene expression, blue = decreased gene expression) showing batch-specific assignment of Active (orange covariate bar) and Inactive (blue co-variate bar) phenotype samples. Horizontal dendrogram = samples, vertical dendrogram = 1318 classifying genes (gold bars = genes upregulated for Active, gray bars = genes upregulated for Inactive phenotype). b Box plots of normal breast cell composition (% adipocyte, stromal, epithelial nuclei) across all 151 samples relative to their transcriptome phenotype assignments (expression cluster Active or Inactive). c Left panel: two-dimensional TumorMap layouts [19] of all KTB normal breast full transcriptomes after batch-integration and projected relative to 1096 TCGA breast cancer full transcriptomes. Batch-integrated normal breast transcriptome samples colored according to batch (blue = F, yellow = P), and TCGA breast cancer transcriptomes colored according to their PAM50 subtype [20]. Right panel: normal breast full transcriptome TumorMaps presented in isolation before and after batch-correction, showing good sample set integration of both F and P batch samples and preserved spatial separation of Active and Inactive phenotypes, and overlay of the integrated set with color scores (red = high) of their individual cell compositions (% nuclei)