Linking bisphenol a exposure to pancreatic cancer: Evidence from population epidemiology, multi-omics screening, and cellular validation.

Abstract

Bisphenol A (BPA) is a ubiquitous environmental endocrine disruptor; its association with pancreatic cancer and its potential mechanisms of action remain unclear. This study established a comprehensive framework integrating population epidemiology, computational toxicology, machine learning, molecular dynamics simulations, and experimental validation to conduct a systematic investigation. Analysis of NHANES 2003-2016 data (N = 11,894 adults) revealed that high urinary BPA exposure (quartile 4 vs. quartiles 1-3) was significantly associated with increased cancer mortality (OR = 1.43, 95% CI: 1.04-1.96, P = 0.027; PAF = 9.51%). Intersection analysis of 597 BPA target genes and 3933 pancreatic cancer-associated genes identified 246 overlapping candidates, from which 14 core genes were prioritized through machine learning screening across 15 algorithms and 175 combinations. A diagnostic nomogram constructed based on these 14 genes achieved AUCs of 0.984 and 0.997 in the GSE15471 and TCGA-PAAD + GTEx datasets, respectively. Molecular dynamics simulations (200 ns × 3 replicates) combined with MM/GBSA binding free energy calculations indicated that the binding affinity of AHR for BPA (-57.55 ± 7.20 kcal/mol) was approximately twice that of GPRC5A (-28.87 ± 12.96) and comparable to that of known AHR ligands. Functionally, BPA exhibited a cell-type-dependent biphasic effect, inhibiting normal pancreatic ductal epithelial cell (hTERT-HPNE) viability while enhancing colony formation in AsPC-1 and MiaPaCa-2 pancreatic cancer cells, with AHR knockdown attenuating this proliferative advantage. Collectively, these findings provide population-level, computational, biophysical, and functional evidence supporting an association between BPA and pancreatic cancer, with AHR identified as a potential key molecular mediator.