Mechanism-informed prioritization of drinking water mixtures for metabolic liver disease: Integrating AOP-anchored bioactivity and toxicokinetics.

Mixture-aware drinking-water assessment increasingly relies on broad-scope high-resolution mass spectrometry (HRMS), yet translating non-target screening inventories into endpoint-relevant priorities remains challenging. Here we propose a Mechanism-Informed Modular Prioritization System (MIMPS) that integrates occurrence profiling, ToxCast-based gene scoring, gene set enrichment analysis (GSEA), adverse outcome pathway (AOP)-informed mechanistic anchoring, and toxicokinetic refinement to prioritize candidate mixture drivers. As a proof-of-concept, MIMPS was configured for metabolic dysfunction-associated steatotic liver disease (MASLD) and applied to the Shanghai drinking-water supply system. Quarterly-pooled composites from 18 sites yielded 96 manually reviewed organic chemical annotations across confidence Levels 1-3, including 31 Level-1, 28 Level-2 and 37 tentative Level-3 annotations, with details provided in the supplementary tables. Detected chemicals were projected onto liver-relevant ToxCast assays and aggregated into ranked gene scores; the resulting signature was condensed by GSEA and mapped to a 12-pathway AOP-informed MASLD scaffold spanning receptor-level perturbation, lipid metabolic dysregulation, oxidative/inflammatory injury, fibrotic remodeling, and steatotic liver injury. Candidate drivers were then prioritized using a five-dimensional index integrating finished-water occurrence, pathway-based toxicity, predicted absorption, distribution, and elimination persistence. Toxicokinetic refinement shifted prioritization toward bio-persistent PFAS, particularly PFOA, PFOS and PFHxS, while a concentration-based sensitivity analysis produced broadly consistent leading candidates. Overall, MIMPS provides a transparent, modular workflow for linking complex drinking-water mixtures to endpoint-focused mechanistic hypotheses and follow-up chemical priorities, while remaining a screening-level approach rather than a quantitative prediction of MASLD risk.