Fig. 3: Definition of the BCC HER2-low subtype: distinctions from other subtypes, main features, and drivers.

a Barplot showing proportions (in %) of HER2-positive and negative samples as assessed by IHC across the BCC subtypes in TCGA, SCAN-B, and METABRIC cohorts (total n = 5602). p-values for the Pearson’s chi-squared test are shown. b Barplot showing proportions (in %) of HER2-positive and negative samples as assessed by ISH across BCC subtypes in TCGA (n = 394). p-values for the Pearson’s chi-squared test are shown. c Barplot showing proportions (in % of copy number alterations, CNAs) of ERBB2-amplified samples across BCC subtypes in TCGA. p-values for the Pearson’s chi-squared test are shown. d Violin plot showing expression levels of ERBB2, AR, and EGFR (in scaled units) across BCC subtypes (METABRIC, TCGA, and SCAN-B datasets, total n = 6223). ***: p-value < 0.001, **: p-value < 0.01, *: p-value < 0.05, -: p-value < 0.1. Whiskers indicate 25th percentile (bottom) and 75th percentile (top) +/− 1.5 IQR. e Sankey plot depicting the overlap between subtypes of TNBC samples (METABRIC cohort, n = 256) classified by Burstein²² and the BCC. f ERBB2 expression levels across BCC subtypes in the TCGA dataset (n = 970). The color key indicates types of CNA change. ***: p-value < 0.001, **: p-value < 0.01, *: p-value < 0.05, -: p-value < 0.1. g Proportion of alterations in biomarker genes among BCC subtypes (TCGA and METABRIC datasets, n = 2827). h Barplot depicting various types of alterations in biomarker genes across BCC subtypes. The purple scale shows -log10(adjusted p-values) for the right-tailed Fisher’s exact test. i Panorama of events in driver genes across BCC subtypes (TCGA and METABRIC datasets, n = 2906).