Integrating machine learning and metabolomics to identify PFAS-associated metabolic alterations related to lung cancer risk

Per- and polyfluoroalkyl substances (PFAS) are pervasive environmental contaminants increasingly implicated in lung cancer risk. However, the metabolic features and biological pathways associated with human PFAS exposure in relation to lung cancer remain poorly characterized. We conducted a nested case-control study within the Quzhou Environment Exposure and Human Health cohort (408 lung cancer cases and 421 controls) to examine whether serum PFAS concentrations are associated with lung cancer risk and to identify metabolic alterations associated with both PFAS exposure and lung cancer. Seven PFAS were quantified and associations assessed with multivariable logistic regression, Bayesian Kernel Machine Regression, and Weighted Quantile Sum (WQS) regression models, alongside high-resolution serum metabolomics (1,403 features) and machine learning classification. Higher serum levels of perfluorooctanoic acid (PFOA), perfluorononanoate (PFNA), perfluorohexane sulfonate (PFHxS), perfluorooctanesulfonic acid (PFOS) and 6:2 chlorinated polyfluoroether sulfonic acid (6:2Cl-PFESA) were associated with increased risk, with dose-response patterns for several chemicals. Highest-quartile odds ratios were 7.94 for PFOA, 3.82 for PFNA, 4.55 for PFHxS, 2.84 for PFOS, and 5.54 for 6:2Cl-PFESA (all trend p < 0.05). Mixture analyses supported a positive joint effect (WQS β = 1.95, 95% CI 1.61-2.38), with greatest weights for PFOA, PFHxS and PFNA. Metabolomics revealed broad metabolic alterations in cases (886 features up-regulated and 202 down-regulated). Among the evaluated metabolomics-based models, the least absolute shrinkage and selection operator (LASSO) showed the highest performance under the present analytical framework. The LASSO model identified candidate discriminative metabolites including adenosine triphosphate, citrulline, L-carnitine, 1-methylhistidine, and ethanolamine (β = 1.05-51.04, p < 0.05). Pathway enrichment analysis indicated significant perturbations in phosphatidylcholine biosynthesis, phospholipid biosynthesis, phosphatidylethanolamine biosynthesis, and aspartate metabolism. This study provides novel evidence that PFAS exposure is associated with metabolic alterations that may be relevant to lung cancer-related processes.