Identification of key toxicological targets for aristolochic acid-induced lung adenocarcinoma using multiple machine learning and molecular docking techniques

Aristolochic acid (AA) is a compound derived from plants of the Aristolochiaceae family that exhibits nephrotoxicity and carcinogenicity. Although its association with clear cell renal cell carcinoma (ccRCC) has been established, the correlation between AA and lung adenocarcinoma (LUAD), along with the key molecular targets and underlying mechanisms, remains unclear. This study aims to address this research gap and explore the potential molecular mechanisms by which AA may induce LUAD. Five lung adenocarcinoma transcriptome datasets were selected from the GEO database. Differentially expressed genes (DEGs) associated with LUAD were identified through differential expression analysis, while potential targets of AA were obtained by integrating multiple databases. Key module genes were screened using Weighted Gene Co-expression Network Analysis (WGCNA), followed by the identification of core targets through 113 machine learning models and SHAP (SHapley Additive exPlanations) analysis. Molecular docking was used to predict binding between AA and the core targets, survival analysis was used to assess prognostic associations, and 200 ns molecular dynamics simulations were used to evaluate the conformational stability of the representative AA-EDNRB and AA-FABP5 complexes. Through comprehensive analysis, a total of 33 potential targets involved in AA-induced lung adenocarcinoma were identified. Further screening yielded 5 core genes: EDNRB, DTYMK, FABP5, MME, and CPB2. Among these core genes, EDNRB, FABP5, MME, and CPB2 showed significant downregulation in lung adenocarcinoma tissues, while DTYMK exhibited significant upregulation (all with P < 0.05). Molecular docking results demonstrated that AA had strong potential binding affinity for all 5 core targets (binding energy < − 5 kcal/mol), and subsequent 200 ns molecular dynamics simulations supported the conformational stability of the AA-EDNRB and AA-FABP5 complexes. Integrated transcriptomic, machine-learning, and structural analyses identified five candidate core genes associated with the putative relationship between AA exposure and LUAD. Molecular dynamics simulations provided dynamic support for the predicted AA-EDNRB and AA-FABP5 interactions. These results are hypothesis-generating and require experimental validation before biological or clinical conclusions can be drawn.