Deciphering the role of per- and polyfluoroalkyl substances in prostate cancer: a multi-omics and computational toxicology approach.

Background: Per- and polyfluoroalkyl substances (PFAS), persistent environmental contaminants, are associated with increased Prostate cancer (PCa) risk. However, their molecular mechanisms are poorly defined. Methods: We employed a comprehensive computational and experimental framework. The toxicological profiles of PFOA and PFOS were predicted. Shared molecular targets between PFAS and PCa were identified by integrating toxicogenomic and transcriptomic data, followed by protein-protein interaction network and enrichment analyses. A robust prognostic model was built and validated using multiple machine-learning algorithms. Core targets were further investigated via single-cell/spatial transcriptomics and molecular docking. Key findings were functionally validated in DU145 PCa cells using qPCR, Western blotting, and assays for proliferation, migration, and invasion. Results: Computational analysis confirmed the carcinogenic and endocrine-disrupting potential of PFAS. We identified 219 common targets significantly enriched in inflammation, oxidative stress, and cancer-related pathways like PPAR and p53 signaling. Network topology highlighted key hub genes, including ALB and PPARG. A 10-gene machine-learning model demonstrated strong prognostic performance (average C-index: 0.710). Cross-omics analyses pinpointed CDC20 as a pivotal core gene within the PFAS-PCa network. Molecular docking indicated stable binding of PFAS to core targets like CDC20. Conclusion: This work systematically reveals that PFAS exposure is associated with PCa progression, potentially involving dysregulation of core genes such as CDC20 and perturbing critical cancer pathways. The developed prognostic model holds clinical relevance, and the identified natural products offer a foundation for designing interventions to potentially mitigate PFAS-associated carcinogenic effects, advancing both mechanistic understanding and preventive strategies. However, the findings are primarily based on computational predictions and a single cell line; further validation in multiple models and experimental systems is required.