Fig. 1
Genes and signatures distinguishing DCIS and invasive cancer cell segments. a Hierarchical clustering dendrogram illustrating the relationship between cancer cell segments (top) and stromal segments (bottom). Sample ID and tissue type is indicated by color. b Differentially expressed genes between DCISinv and invasive cancer cell segments using mixed effect models. The model’s estimate (effect size) is shown on the x-axis and -log10(FDR) is shown on the y-axis. Genes significantly upregulated in invasive segments (FDR < 0.1) are shown in red and genes significantly upregulated in DCISinv segments are shown in purple. c Hallmark gene sets enriched between invasive cancer cell segments and DCISinv. Normalized Enrichment Scores are shown for the top five signatures higher expressed in invasive segments (NES > 0) and the top five signatures higher expressed in DCISinv (NES < 0). Colored bars indicate significant signatures (FDR < 0.1). d Network of leading-edge genes enriched in DCISinv cancer cell segments. Leading edge genes from glycolysis and oxidative phosphorylation signatures upregulated in DCIS compared to IDC. Node size is correlated to the inverse of the p value from DGE, and gray shading according to the estimate from the mixed effect models. e Significantly differentially expressed genes between DCISinv and invasive cancer cell segments (FDR < 0.1). The heatmap represents gene expression values centered across segments and hierarchically clustered across genes and segments (distance method: euclidean, clustering method: complete). Top annotation indicates sample ID and tissue type