Epithelial cell adhesion molecule-centered, bioinformatics and machine learning-based meta-analysis for the identification of pan-cancer epithelial-mesenchymal markers for circulating tumor cells

Advances in bioinformatics have greatly contributed to the discovery of epithelial–mesenchymal transition (EMT) markers, such as epithelial cell adhesion molecule (EPCAM). This study aimed to conduct an EPCAM-centered meta-analyses of previously RNA-sequencing data for identifying pan-cancer EMT markers in circulating tumor cells (CTCs) utilizing bioinformatics- and machine learning (ML)-based approaches. In this study, the RNA sequencing data of seven different cancer types from two datasets, namely GSE273023 and GSE274442, were analyzed. Gene–gene correlation among included cancer samples and EPCAM-centered gene–gene correlation analysis were performed. The data were subjected to ML-based pathway and gene clustering analysis. Notably, the results showed that most of the cancers presented similar gene expression profile, albeit with some differences, which were primarily attributed to differences in mitochondrial gene expression. Furthermore, gene–gene correlation analysis revealed multiple genes with significantly altered expression, including CBWD2, MED23, QRSL1, ZNF568, and INTU. Similarly, TRPS1 was found to be significantly correlated with EPCAM. Overall, the findings of this study reveal the association between EPCAM–TRPS1 and CBWD2-associated MED23–QRSL1–ZNF568–INTU axes, thereby showing their potential as co-markers and for the development of multiplexed immunoassay for a robust pan-cancer CTC detection approach.