Pan-cancer bimodal coupling between E-cadherin protein expression and genome-wide transcripts defines an EMT signature, predicting EMT- and MET-inducing small-molecules.

<p>(<b>A</b>) Top-25 transcripts in CCLE with the strongest positive and top 24 negative bimodal coupling coefficients (r_b) to E-cadherin protein expression. Red squares indicate previous EMT signature genes in non-small cell lung carcinoma published by Byers <i>et al</i>. [<a href="http://www.ploscompbiol.org/article/info:doi/10.1371/journal.pcbi.1005911#pcbi.1005911.ref051" target="_blank">51</a>]. To define an EMT signature, we considered transcripts with |r_b| > 0.5, resulting in 215 epithelial and 24 mesenchymal markers. (<b>B</b>) Distribution of bimodal coupling coefficients, showing that E-cadherin coupling coefficients are shifted towards negative values compared to all measured proteins. (<b>C</b>) Overlap of EMT signature with previously published transcriptomic EMT signatures. The ‘mesenchymal’ bar plot is for the inversely correlated (coupled) genes and the ‘epithelial’ for the positively correlated genes. (<b>D</b>-<b>E</b>) Gene set enrichment analysis of epithelial part of the EMT signature. The TF enrichment analysis used ChIP-seq data to predict TFs involved in the regulation of the epithelial genes. The pie charts indicate the fraction of the signature genes associated with significantly enriched terms or TFs. (<b>F</b>) Small-molecule perturbations predicted to induce EMT and MET based on L1000 cell line data and the L1000CDS² method. The top-50 signatures are shown with results from multiple cell lines or concentrations aggregated by boxplots. PK: protein kinase.