Fig. 2

Gene expression studies on lung cancer subtypes. a A schematic description of the unsupervised consensus clustering approach is provided on the left panel. The clustering results are displayed on the right panel as a heatmap, in which tumor samples are arranged in columns, grouped according to their expression clustering class, annotated for the histological subtype and for the somatic alteration status. Expression values of genes identified by ClaNC (Methods section) are represented as a heatmap; red and blue indicate high and low expression, respectively. Selected candidate genes are shown on the right. b Significant enrichment of differentially expressed genes in signaling pathways is displayed for all clustering classes (P < 0.0001, Methods section). c Expression values for key neuroendocrine differentiation markers are plotted for each clustering class as box-plots (median and interquartile range, whiskers: min–max values). Dashed black lines indicate the threshold for low expression (Methods section). Q < 0.05 (#), significance determined by SAM (Supplementary Dataset 12); P < 0.001 (***) Mann–Whitney U-test. d The correlation of each sample to the centroid of its clustering class was calculated and displayed as box-plot (median and interquartile range, whiskers 5–95 percentile)