PAN, J.; ZHANG, Y.; YANG, A.; JIANG, L.; SHEN, Y.; SUN, Y.; ZHU, J.; FAN, M.; SHI, J.

Background: Per- and polyfluoroalkyl substances (PFAS), particularly perfluorooctane sulfonate (PFOS) and perfluorooctanoic acid (PFOA), are persistent organic pollutants ubiquitous in the environment. Epidemiological evidence has closely linked them to an elevated risk of prostate cancer (PCa). However, the precise molecular mechanisms by which combined PFOS/PFOA exposure promotes prostate cancer and their dynamic effects on the tumor microenvironment remain unclear. Methods: This study constructed a multi-module analytical framework integrating network pharmacology and computational biology: (1) Through ADMET toxicity prediction, multi-database target collection (three-way Venn analysis), panoramic GO/KEGG enrichment, focused androgen receptor (AR) axis analysis, GWAS genetic association validation, protein-protein interaction (PPI) network construction, machine learning-based independent screening, and a relaxed intersection strategy, we systematically identified PFOS/PFOA-prostate cancer core targets. (2) Subsequently, a PFAS-PTS score weighted purely by Cox coefficients was employed to drive gene set variation analysis (GSVA)-based pathway enrichment, tumor microenvironment (TME) deconvolution, ordinary differential equation (ODE)-based kinetic modeling, and drug intervention prediction. Results: Target collection identified 100 shared PFOS/PFOA-prostate cancer targets, from which 18 core targets were determined after multi-module screening. These targets were significantly enriched in the AR signaling axis, the PI3K-AKT pathway, and cell cycle regulation. Molecular docking confirmed strong binding affinities of PFOS/PFOA with AR (-9.49/-8.56 kcal/mol), AKT1 (-7.56/-6.93 kcal/mol), and PTEN (-6.36/-6.08 kcal/mol). GSVA revealed that the G2M checkpoint and E2F target gene pathways were significantly upregulated in the high-risk group (padj < 0.001), whereas the androgen response pathway was downregulated (padj = 4.8e-4). TME deconvolution (GSE141445, NNLS) revealed a significantly increased proportion of tumor cells (PCa) (p = 2.4e-4) and markedly reduced CD8+ T cell infiltration (p = 5.7e-4) in the high-risk group, indicating immunosuppressive microenvironment remodeling. ODE-based kinetic modeling confirmed that PFAS promoted tumor cell proliferation and suppressed immune surveillance in a dose-dependent manner. Drug intervention simulation demonstrated that the combination of enzalutamide and Alpelisib achieved optimal tumor cell inhibition (33.9% predicted by the ODE model). Conclusion: PFOS/PFOA promote prostate cancer progression primarily through multi-target synergy involving AR axis disruption, PI3K-AKT pathway activation, and cell cycle dysregulation, while reshaping an immunosuppressive tumor microenvironment. The integrative computational framework established in this study provides systematic computational evidence for risk assessment and therapeutic intervention in PFAS-associated prostate cancer. Keywords: PFOS; PFOA; prostate cancer; network pharmacology; machine learning; molecular docking; tumor microenvironment; ODE kinetic modeling