Fig. 2

Identification of a 20-gene Notch transcriptomic signature. A Heatmap to represent the number of differentially expressed genes (DEGs) for all possible comparisons of the experimental conditions tested: NOTCH activation status: “Notch on”, “ground state” and “Notch off”; experimental timepoint: 8 h or 72 h (8 h or 72 h), or all samples analysed for differential gene expression analysis together, irrespectively of the experimental timepoint (8 + 72 h). Upregulated genes in brown; downregulated genes in blue. To the right: DEGs for various combinations of datasets as indicated. Gene intersections indicate the number of DEGs; a cross indicates comparisons with not-left skewed P value distributions or less than 100 DEGs. up = upregulation and down = downregulation. B The list of 14 candidate experimental conditions after left skewed P value analysis of the initial 74 experimental conditions. C Plots of coherence scores for three selected candidate signatures (signatures #5, #9 and #11) for different number of genes in each signature. The data show signature scores > 0.2 for the three signatures after training for optimal coherence scores for signature sizes between 10 and 30 genes on the TCGA-BRCA patient dataset. D Schematic depiction of generation of bulk transcriptomic datasets from 19 breast cancer cell lines (10 basal-like; 6 HER2-posivite and 3 Luminal A (LumA)). Transcriptomes were obtained after “Notch on”, “Notch off” or “ground state” conditions, as for the six cell lines in Fig. 1. E Principal component analysis of the transcriptomic data from the 19 cell lines under “Notch on”, “Notch off” or “ground state” conditions at 8 or 72 h, as indicated. A total of 12,038 genes were used for the PCA. F GSEA normalized enrichment score representations of the three candidate signatures in the transcriptomic dataset from the 19 cell lines in E tested against transcriptomes from 8, 72 or 8 + 72 h. Note that signature #11 outperforms the other signatures across the different datasets. For a corresponding analysis by ROC curves of the candidate signatures to classify Notch activation status in the 19-cell line dataset, see Additional file 1: Fig. S3E