Fig. 2
LN-signature predicts chemotherapy response of chemo-naïve lung metastatic tumor tissue collected by mediastinoscopy. A Hierarchical clustering analysis of the LN-signature in MED samples. Data are log2-ratio. LUAD, lung adenocarcinoma; LUSC, lung squamous cell carcinoma; NSCLC, other non-small cell lung subtypes; NA, no available data. B Heat map showing gene expression of the indicated marker analyzed by qRT-PCR in LNmets (EBUS samples, N = 5; and MED samples, N = 5). NCI-H2023 and NCI-H1993 lung cancer cells (LC, yellow) were used as positive controls for the expression of epithelial marker, while HUVEC (EN, orange), WI38 (FI, red) and HL-60 cells (LK, magenta) were used as positive control for endothelial, fibroblast and immune-like markers expression, respectively. Data are log2-ratio. C Pie chart showing the number of miRNAs of LN-signature (N = 16) that were found differentially expressed between pN0 and pN2 samples in MED cohort. D Schematic representation of strategy adopted to derive miRNA-based NACT-predictive models. E–G Upper panels: receiver operating characteristic (ROC) curves of the 16-miRNA model (E), 14-miRNA model (F) and 4-miRNA model (G) in the validation set (MED samples, red). Lower panels: box plot of the predicted probability of being a responder according to the 16-miRNA model (E), 14-miRNA model (F) and 4-miRNA model (G)