A network toxicology and machine learning approach to uncover the molecular machinery of bisphenol A-induced colorectal cancer

Abstract

The carcinogenic relevance of environmental contaminant bisphenol A (BPA) to colorectal cancer (CRC) has gained growing attention, yet the molecular networks potentially linking BPA exposure to CRC remain incompletely characterized. This study integrated network toxicology and machine learning to predict candidate molecular targets and putative signaling networks associated with BPA-correlated colorectal carcinogenesis. The study’s methodology involved an initial differential expression screening across several CRC transcriptomic datasets to establish a disease-specific gene signature. Subsequently, a multi-tiered computational strategy was employed: network toxicology was used to map potential BPA-protein interactions, machine learning algorithms were applied to distill a minimal set of high-impact targets, and molecular docking simulations provided atomic-level validation of the proposed binding events. We identified 53 overlapping genes between predicted BPA-interacting proteins and CRC-related transcripts. Machine learning screening further filtered a 12-gene panel with favorable predictive performance for CRC status, including MET, SORD, DPEP1, KIT, RIPK2, SET, HSP90AB1, DBF4, MMP1, MMP12, ANPEP, and GLA. Molecular docking simulations predicted stable binding interactions between BPA and the protein products encoded by these 12 genes. This study delineates a specific gene network potentially targeted by BPA to promote CRC pathogenesis. The machine learning-derived 12-gene signature, interpreted as CRC-associated genes overlapping with predicted BPA targets and supported by in silico molecular docking, offers valuable insights into the molecular basis of BPA-associated colorectal carcinogenesis and presents candidate targets for subsequent experimental validation.