Artificial intelligence-augmented mixture toxicology: reframing unresolved risk at Camp Lejeune

Persistent uncertainty regarding health risks associated with contaminated drinking water at Camp Lejeune reflects a broader limitation in toxicology: the difficulty of evaluating complex chemical mixtures in genetically heterogeneous populations. Historical assessments have relied on single-chemical paradigms, assumptions of dose additivity, and population-average susceptibility. These approaches are not well aligned with exposures involving multiple volatile organic compounds (VOCs) that share metabolic pathways and generate reactive intermediates. This Forum article advances the position that artificial intelligence (AI)-enabled computational toxicology can strengthen mixture risk assessment by integrating toxicokinetic, toxicodynamic, and toxicogenomic data into mechanistically coherent, testable models. Using the Camp Lejeune VOC mixture-trichloroethylene, tetrachloroethylene, benzene, and vinyl chloride-as a case study, we illustrate how AI-augmented approaches can identify plausible interaction mechanisms, quantify genotype-dependent variability in internal dose, and generate probabilistic estimates of risk. We argue that the principal unresolved issue at Camp Lejeune is whether chronic low-level exposure to interacting mixtures disproportionately affected susceptible subpopulations. AI-driven approaches provide a tractable framework to address this question and to clarify the interface between mechanistic toxicology and public health decision-making.