Exploring the toxicological mechanisms of Benzo[a]anthracene (BaA) exposure in lung adenocarcinoma (LUAD) via network toxicology, machine learning, and multi-dimensional bioinformatics analysis

Background: Lung adenocarcinoma (LUAD) is a major lung cancer subtype influenced by environmental factors. Benzo[a]anthracene (BaA), a common Group 2B carcinogen found in pollutants, smoke, and food, shows genotoxic and oncogenic activity; however, its specific mechanisms in LUAD pathogenesis remain unclear and warrant systematic investigation. Objective: This study aims to elucidate the mechanisms of BaA-induced LUAD, identify core targets, validate their expression, immunorelevance and clinical significance, and construct a hypothesis framework for AOP in BaA-exposed LUAD. Methods: We integrated network toxicology, multi-machine learning algorithms (LASSO, SVM-RFE, and Random Forest) and multidimensional bioinformatics analysis. Potential BaA-LUAD intersection targets were collected from public databases and subjected to functional enrichment analysis. Core targets were screened and validated using GEO and TCGA-LUAD (via UALCAN) datasets for differential expression, immune infiltration and prognostic value. Molecular docking and 100 ns molecular dynamics (MD) simulations were applied to evaluate the binding stability between BaA and core targets. Results: A total of 248 intersection targets were identified, with significant enrichment in chemokine signaling, ErbB signaling, and viral protein–cytokine receptor interaction pathways. Machine learning prioritized five core targets: TNNC1, ABCC3, CRABP2, CXCL12, and OLR1. These genes were consistently dysregulated in LUAD samples across cohorts (p 0.05). Molecular docking confirmed stable binding between BaA and all core targets, with the strongest affinity for CRABP2 (–8.4 kcal/mol). MD simulations further supported complex stability. Conclusion: BaA promotes LUAD progression via multi-target regulation and tumor immune microenvironment remodeling. This study offers an integrated computational framework and an AOP-based theoretical foundation for assessing pollutant health risks and informing targeted LUAD interventions.