Abstract
Alzheimer's disease (AD) is a multifactorial neurodegenerative disorder in which exposure to environmental endocrine-disrupting chemicals may contribute to disease susceptibility. Here, we integrated network toxicology, bulk transcriptomics, machine learning, immune deconvolution, single-nucleus transcriptomics, competing endogenous RNA (ceRNA) analysis, molecular docking, and 100-ns molecular dynamics (MD) simulations to investigate the potential molecular links between bisphenol A (BPA) exposure and AD. Cross-database integration identified 248 shared BPA-AD targets, from which 47 network-central genes and 19 differentially expressed genes were prioritized. Five genes-GSK3B, BCL2, EGFR, APP, and PPARD-were subsequently retained as a candidate gene signature. Although the resulting model showed strong discriminatory performance in the discovery cohort, substantially lower performance in validation cohorts indicated limited generalizability and potential overfitting. Functional analyses implicated neuroinflammatory signaling, mitochondrial apoptosis, amyloid-related processes, and lipid and metabolic dysregulation in the molecular association between BPA and AD. LM22-based CIBERSORT analysis revealed alterations in peripheral leukocyte-like immune signatures, which were interpreted cautiously because this reference matrix does not directly represent resident brain immune populations. Single-nucleus transcriptomic analysis of GSE163577 further localized key genes across biologically plausible cerebrovascular and neuroimmune cell populations. CeRNA analysis suggested potential post-transcriptional regulatory relationships associated with the identified hub genes, while molecular docking identified putative interactions between BPA and candidate target proteins. Subsequent 100-ns MD simulations further characterized the dynamic behavior and conformational stability of the selected BPA-protein complexes, providing complementary evidence for the docking-derived structural hypotheses. Collectively, these findings suggest that BPA-related molecular targets may converge on immune, vascular, metabolic, and neurodegenerative processes relevant to AD. However, the present findings remain primarily computational and hypothesis-generating, and experimental studies are required to establish causal and mechanistic relationships.