Exploring the pathophysiological relationship between bisphenol A exposure and ischemic stroke risk using network toxicology and machine learning

To explore the molecular mechanisms by which bisphenol A (BPA) exposure contributes to ischemic stroke (IS) risk via the induction of cellular senescence, using an integrative approach combining network toxicology, transcriptomics, and machine learning. BPA’s chemical properties and target genes were retrieved from PubChem, ChEMBL, STITCH, and SwissTargetPrediction databases, with toxicity validated via ProTox-3. Ischemic stroke (IS)-related differentially expressed genes (DEGs) were identified from merged GEO datasets (GSE16561, GSE37857) and validated in GSE58294. Cellular senescence-related genes (CSRGs) were obtained from CellAge. The intersection of BPA target genes, IS-DEGs, and CSRGs (BPA-CSRG-IS intergenes) was analyzed using protein-protein interaction (PPI) networks and machine learning algorithms (random forest, LASSO) to screen hub genes, with SHAP analysis applied to evaluate the optimal model. A risk prediction model was constructed via logistic regression and validated externally. Additionally, immune cell infiltration analysis, single-sample gene set enrichment analysis (ssGSEA), miRNA-TF-mRNA regulatory network construction, and molecular docking were performed to explore underlying mechanisms. BPA was confirmed to cross the blood-brain barrier and interact with 2,985 target genes. Merged IS datasets identified 708 DEGs, enriched in immune-related pathways. Intersection analysis yielded 17 BPA-CSRG-IS intergenes, with 2 hub genes (MMP9, HDAC1) identified via machine learning (KNN model, AUC = 0.940) and SHAP analysis (MMP9 as the top contributor). The risk model based on MMP9 and HDAC1 showed robust performance (training set AUC = 0.830; validation set AUC = 0.946). IS patients exhibited elevated immune cell infiltration (e.g., CD8 + T cells, M0 macrophages), with MMP9 and HDAC1 strongly correlated with immune cell subsets. ssGSEA revealed hub genes regulate pathways related to neuroinflammation, BBB disruption, and cellular senescence. Molecular docking confirmed strong binding affinity between BPA and MMP9 (–7.2 kcal/mol) or HDAC1 (–7.7 kcal/mol). This study identifies MMP9 and HDAC1 as key hub genes mediating BPA-induced IS via cellular senescence, highlighting their roles in regulating the “senescence-inflammation” cascade and immune cell infiltration. These findings provide mechanistic insights into BPA’s neurotoxicity and potential biomarkers/targets for IS prevention and treatment.