Network toxicology and machine learning analysis of MC-LR-associated hepatocellular carcinoma.

Abstract

Microcystin-LR (MC-LR), a hepatotoxic aquatic contaminant classified by IARC as a Group 2B carcinogen, was investigated using an integrated network toxicology and machine learning framework to explore potential molecular mechanisms underlying MC-LR-related hepatocarcinogenesis. By integrating GEO transcriptomic datasets, differential expression analysis, WGCNA, and predicted MC-LR targets, we identified 24 candidate genes potentially associated with MC-LR-related HCC. Functional enrichment analyses implicated these genes mainly in complement and coagulation cascades and metabolic processes. Machine learning consensus analysis prioritized five core genes-AKR1C3, FABP5, CA2, ADH1B, and EPHX2-each showing an AUC greater than 0.79 and suggesting moderate discriminatory potential. Single-gene GSEA linked these genes mainly to ribosome-related pathways, while single-cell transcriptomic analysis revealed cell-type-specific expression heterogeneity within the HCC microenvironment. Molecular docking provided exploratory structural evidence for possible in silico compatibility between MC-LR and the encoded proteins. However, further experimental validation is required to confirm this hypothesis. Overall, these findings establish a multidimensional framework for investigating MC-LR-associated hepatotoxicity and nominate candidate molecular nodes for future biomarker assessment, functional validation, and environment-related HCC risk evaluation.