Toxicological impact of benzo[a]pyrene on esophageal cancer: an integrated analysis via network toxicology, machine learning, and molecular docking
Xuyan Lan, Zuqiang Huang, Yukun Lin, Xiaoyu Sun, Binghan Guo, Genglin Li, Jintao Wang, Jinlan Lin, Lihuan Zhu, Tianxing Guo
Background To investigate the mechanisms underlying benzo[a]pyrene-induced esophageal cancer (EC), and to screen and identify the key targets and biomarkers associated with benzo[a]pyrene-related EC. Methods Potential targets of benzo[a]pyrene (BaP) were predicted using PharmMapper, SwissTargetPrediction, and ChEMBL databases, and were intersected with differentially expressed genes (DEGs) from the GEO database to screen candidate key genes. Subsequently, diagnostic models were constructed using 14 machine learning algorithms based on the identified key genes. Meanwhile, a prognostic model of key genes was constructed based on the TCGA esophageal cancer cohort, and the correlation between these key genes and tumor immune infiltration was further explored. Additional explainability was provided via SHAP analysis by determining the contributions of key features. Molecular docking was performed to verify the binding between BaP and core targets. Results A total of 82 genes were identified as potential targets of EC induced by BaP. These key genes were found to be mainly involved in core tumor-related pathways, cell cycle regulation, MAPK signaling, and immune-inflammatory pathways, covering the crucial biological processes underlying malignant transformation of EC. Subsequently, 12 core genes ( ACOT9、ACOX3、AURKA、HMGCR、INHBA、MMP3、MSR1、SHC1、SORT1、MAOB、MMP13、CDK4 ) were identified as key regulators by machine learning analysis. Among them, SORT1 and ACOX3 were significantly down-regulated, while AURKA and MMP13 were markedly up-regulated ( P < 0.05). The 12-gene prognostic model enables efficient prognostic stratification of esophageal cancer patients, and core genes are implicated in the remodeling of the esophageal cancer immunosuppressive microenvironment through the regulation of immune cell infiltration. Molecular docking revealed strong binding ability between BaP and target proteins. Conclusions Bioinformatics analysis and molecular docking results revealed significant associations between BaP and 12 core esophageal cancer-related genes. BaP could stably bind to core proteins including AURKA , CDK4 , MMP13 and INHBA , which is potentially correlated with altered cell cycle, metabolic disorders and dysregulated tumor immune microenvironment in esophageal cancer. 12 core genes were identified via machine learning, which offers new perspectives for the interdisciplinary field of environmental toxicology and precision oncology and provides a foundation for the development of individualized therapeutic strategies.