Advancing aquatic ecological risk assessment through AI-driven prediction of chemical exposure and toxicity in freshwater and marine fish

Abstract

Increasing evidence on the massive biodiversity loss caused by chemical pollution highlights the urgent need to understand the internal exposure and resulting risk of chemicals to aquatic species. Current ecological risk assessment frameworks face challenges in addressing the diversity of physiological and biochemical characteristics of aquatic organisms. Here we pioneer the intelligent high-throughput multi-species physiologically based toxicokinetic (HM-PBTK) model by developing AI-driven models for multi-species physiological and biochemical parameters. The HM-PBTK model precisely quantifies tissue-specific internal exposure and predicts in vivo toxicity for neutral and ionizable chemicals across 151 freshwater and marine fish species. In an oestrogenic-effect case study, 85% of toxicity predictions were within fivefold of the corresponding observations. We further developed a user-friendly platform enabling comprehensive ‘exposure–toxicity’ predictions for chemicals by simply setting up exposure scenarios. Our study helps to explore the risks of chemical pollution to aquatic ecosystem, ultimately contributing to the protection of aquatic biodiversity. The ecological risks posed by chemical pollution are difficult to evaluate across diverse aquatic species using existing assessment frameworks. An AI-driven multi-species modelling framework and prediction platform can quantify chemical exposure and toxicity in freshwater and marine fish alike.