Perfluorooctane Sulfonate Exposure and Peripheral Artery Disease: Mendelian Randomization and Integrative Multi-Omics Systems Toxicology With Machine Learning.

Epidemiological and experimental studies have linked legacy and emerging PFAS to adverse cardiovascular outcomes, yet most evidence remains observational and is susceptible to confounding, measurement error, and reverse causation. This study used European-ancestry GWAS data and two-sample Mendelian randomization to assess potential causal associations between PFOA or PFOS exposure and five cardiovascular diseases. For significant MR findings, network toxicology was integrated with transcriptomic data to identify candidate targets and pathways, followed by machine learning-based prioritization. Immune cell infiltration, single-cell transcriptomic analysis, molecular docking, and molecular dynamics simulation were further applied to explore immune mechanisms and pollutant-protein interactions. Genetically predicted PFOS exposure was associated with an increased risk of peripheral artery disease. Enrichment analysis implicated PPAR and PI3K-Akt signaling pathways, while machine learning based on 129 overlapping genes identified HMOX1, KDR, MMP9, and PPARG as key targets. Immune infiltration and single-cell analyses indicated remodeling of the PAD immune microenvironment, with HMOX1 mainly enriched in macrophages. Virtual knockout highlighted APC-related MHC II antigen presentation, and docking and dynamics simulations supported stable PFOS-HMOX1 binding. These findings suggest a potential PFOS-PAD association and provide candidate biomarkers for further validation.