Extended Data Figure 7: EMT-associated miRNAs and their relationship to miRNA clusters and TGFβR2 somatic alterations.

a, Normalized miR-200a-3p abundance (RPM) across RPPA clusters for all 112 (top) and 92 squamous (bottom) samples of the core set for which RPPA data are available. P values presented are from two-sided Kolmogorov–Smirnov tests for RPPA-based EMT cluster versus non-EMT cluster samples. For n = 112 samples, median miR-200a-3p RPM = 296.4 within the EMT cluster (n = 29) and 410.0 (n = 83) in non-EMT cluster samples. For squamous samples, median miR-200a-3p RPM = 296.4 (n = 29) within the EMT cluster and 393.4 (n = 63) in non-EMT cluster samples. EK-A2R7, which is in the hormone RPPA cluster, has an RPM value of 4,267 and is not shown. Results are not presented for adenocarcinoma samples separately owing to limited sample numbers (n = 18 from the core set with RPPA data available). b, Negative and positive Spearman correlation coefficients (FDR < 0.05) between EMT mRNA score and normalized abundance (RPM) for miRNA mature strands (n = 178). miRNAs that have been reported as associated with EMT (see Methods) are highlighted by blue bars. c, Normalized abundance heatmap of miRNAs most strongly negatively and positively correlated with EMT mRNA scores, with samples grouped by miRNA cluster and sorted by EMT score within each cluster. Somatic mutations (MUT) and deletions (HOMDEL) are shown for TGFBR2, CREBBP, EP300 and SMAD4. Methylation and concomitant downregulated expression alterations (ALT) as defined in Methods for miR-200a/b are also shown. miRNAs in blue represent those highlighted by blue bars in b. d, e, Same as b, c, for the n = 144 squamous tumour samples.