Abstract

We investigated the carcinogenic effects of four endocrine-disrupting chemicals—bisphenol A (BPA), diethyl phthalate (DEP), dimethyl phthalate (DMP), and dioctyl phthalate (DOP)—in nasopharyngeal carcinoma (NPC) and thyroid carcinoma (THCA) using an integrated toxicogenomic–machine learning–docking–experimental pipeline. Intersection analysis identified 31 NPC-related overlapping genes and 39 THCA-related overlapping genes, with 19 shared core targets across both malignancies. These shared targets were enriched in oncogenic signaling pathways including Mitogen-activated protein kinase （MAPK）, Phosphoinositide 3-Kinase-Protein Kinase B (PI3K–AKT), and Janus kinase/signal transducers and activators of transcription (JAK/STAT). A multi-algorithm machine learning framework constructed 113 predictive models and prioritized six diagnostic genes (CCNA2, CDK2, MET, F2, TYMS, PPARG). High expression of CCNA2 (HR=1.43, p = 0.016), CDK2 (HR=1.66, p = 0.002), MET (HR=1.58, p = 0.002), and PPARG (HR=1.45, p = 0.0072) was associated with worse overall survival, whereas TYMS and F2 were not significant. Molecular docking showed stable ligand–protein binding with energies from –5.2 to –8.1 kcal·mol⁻¹ , with the strongest affinities observed for BPA–CDK2 (–8.1) and BPA–PPARG (–8.1); DEP also showed strong binding to CDK2 (–7.0). In vitro, BPA and DEP (but not DMP/DOP) increased colony formation (p < 0.01), accelerated wound closure, upregulated oncogenic genes (e.g., CDK2/MET/CCNA2; p < 0.05), and elevated p-MEK without changing total MEK in 5–8 F and TPC-1 cells. Collectively, BPA and DEP promote head and neck tumor progression through MEK pathway activation and cell-cycle dysregulation.