Evaluating the Vascular Risk of PFCs: An Integrated XGBoost-Driven Structure–Activity Prediction and Experimental Validation Study

Abstract

Perfluorochemicals (PFCs) are emergent and persistent organic pollutants with widespread application. Their structural similarity and detection in serum raises substantial concerns regarding their toxicological effects. While the endocrine-disrupting effects of PFCs are well-recognized, the structure–activity relationship with respect to vascular function has not been investigated yet. This study addresses this critical gap by investigating the impact of PFCs on endothelial cell function, a key determinant of cardiovascular health. Through a machine learning-based quantitative structure–activity relationship (QSAR) model, we analyzed 16 structural descriptors for 23 environmentally prevalent PFCs with respect to their binding affinities to endothelial cell receptors. The eXtreme Gradient Boosting (XGBoost) algorithm suggested short-chain PFCs with strong acid groups may as particularly detrimental to endothelial cells, a finding substantiated by subsequent cell culture experiments. We also integrated computational and experimental approaches, providing a detailed understanding of the structure–activity and dose–response relationships of PFCs. Furthermore, the population health risk assessment, linking in vitro adverse effect with in vivo exposure data, indicated differences in risks across countries due to the global shift in the fluoride industry; the entire Chinese population is at high risk, with risk varying by gender and industrialization level. This study not only elucidates the structure–activity relationship of PFCs on vascular function but also offers a strategic framework for managing toxic PFCs and proposing the development of safer alternatives.