Unraveling the hepatocarcinogenic mechanisms of organochlorine pesticides: an integrated in silico framework combining network toxicology, machine learning, and molecular docking

Organochlorine pesticides (OCPs), such as dichlorodiphenyltrichloroethane (DDT) and Lindane, are persistent environmental pollutants that have been epidemiologically linked to hepatocellular carcinoma (HCC). Nevertheless, the molecular mechanisms through which these agents drive hepatocarcinogenesis remain poorly defined. An integrative computational framework was developed that combines network toxicology, seven machine learning algorithms, weighted gene co-expression network analysis (WGCNA), transcriptomics, single-cell omics, molecular docking, and molecular dynamics simulations. Putative targets of DDT and Lindane were retrieved from public databases, and HCC-associated genes were identified using GeneCards and OMIM. Functional enrichment analyses were performed to elucidate the underlying biological pathways. Hub genes were screened by WGCNA and machine learning-based feature selection. The expression levels, prognostic significance, and diagnostic performance of the candidate genes were further evaluated using bulk transcriptomic data. Immune infiltration and single-cell analyses were conducted to characterize the tumor microenvironment and cell type specific expression patterns. Finally, molecular docking and 100 ns molecular dynamics simulations were employed to assess the binding stability between the primary target and the pesticides. Enrichment analyses revealed that the MAPK and PI3K-Akt signaling pathways play pivotal roles in OCP-induced hepatocarcinogenesis. WGCNA and machine learning identified three hub genes (EZH2, HSP90AB1, and CDK1), among which EZH2 emerged as the most critical. EZH2 was markedly overexpressed in HCC, correlated with poor prognosis, and showed excellent diagnostic accuracy (AUC = 0.977). Immune infiltration analysis indicated that EZH2 contributes to remodeling of the hepatic immune microenvironment. Single-cell analysis demonstrated upregulation of EZH2 in proliferating tumor cells; differentially expressed genes within this population were enriched in DNA replication and cell cycle progression, and these cells exhibited elevated stemness scores. In silico knockout of EZH2 predicted CCL5, CD96, and CD3E as the most significantly affected downstream targets, which were co-expressed in both proliferating tumor cells and cytotoxic CD4⁺ T cells. Molecular docking and 100 ns molecular dynamics simulations confirmed that EZH2 displayed binding energies of − 7.056 kcal/mol for DDT, − 4.989 kcal/mol for lindane, and − 10.37 kcal/mol for the positive control GSK126, corresponding to moderate, weak, and strong binding interactions, respectively. In summary, by integrating network toxicology, machine learning, and molecular docking, this study identifies EZH2 as a computationally prioritized candidate for OCP-associated HCC and generates testable hypotheses concerning its potential roles in oncogenic signaling and immune modulation. These findings establish a rationale for future experimental validation and risk assessment.