Fig. 4: Distinctly responsive patient samples are marked by dramatic FOXA1 loss, HER3 and HER2 decreases and increased immune signatures.

a Schematic overview of LCCC1214 (Clinical trials identifier NCT01875666) for the analysis of adaptive responses to HER2 targeted therapy. b Normalized RNAseq data from matched patient samples was used to generate log2 fold changes for expressed genes (post-treatment vs. pre-treatment). The fold change in expression was used for the principal component analysis (PCA). Labels indicate patient number and dot color indicates treatment arm. The blue dashed line indicates distinct matched samples (from patients 116, 119, and 123) determined to be strongly responsive based on their distinct expression changes. c Log₂ fold change for FOXA1 for matched patient samples is plotted. The blue dashed line indicates the strongly responsive sample pairs. d Log₂ fold changes of the top 5000 differentially expressed genes for matched patient samples were used for unsupervised hierarchical clustering and nearest neighbor analysis was performed for the FOXA1. The top 10 genes by Pearson correlation are indicated, as well as select genes identified from HER2+ cell line SE analysis. e The breast invasive carcinoma (TCGA, provisional) data set was used to identify genes most significantly co-expressed with HER3 expression in patient tumors. The top genes were XBP1 (not shown) and FOXA1. The dashed line indicates regression line. Table contains legend for mutation status, if known. f Pathway analysis of gene expression data from matched patient samples was performed and the pathway scores differences was used for marker selection comparing the strongly responsive sample pairs (blue dashed line) to all other sample pairs. The top 25 pathways (increased and decreased) are shown (10% FDR cutoff). Representative significant pathways are listed with associated reference.