Deciphering the microplastic toxicity cascade in freshwater fish: Integrating machine learning and multi-level biological responses.

This study evaluated the effects of polyamide microplastics (PA-MPs) on juvenile grass carp by integrating histopathology, oxidative stress, neurotoxicity, immune-related gene expression, swimming behaviour and growth performance through a multi-layered statistical and machine-learning framework. Fish were exposed to 0, 10, 100 and 1000 μg/g PA-MPs for 30 days. Lipid peroxidation and superoxide dismutase increased dose-dependently, indicating oxidative stress, while brain acetylcholinesterase (AChE) activity was inhibited by 27.2% and responded significantly even at the lowest exposure, identifying it as an early and sensitive neurotoxic marker. All four immune-related genes were up-regulated, with the anti-inflammatory cytokine IL-10 showing the largest relative increase (+75.0%). These changes were accompanied by significant declines in weight gain (-56.2%) and feed efficiency, with benchmark dose modelling placing a 10% reduction in weight gain below the medium exposure level and Bayesian estimation assigning a 96.3% probability of a reduction exceeding 0.25 g at the highest dose. Overall, dietary PA-MP exposure produced coordinated disturbance across oxidative, neurotoxic, immune, histopathological and behavioral endpoints together with impaired growth, and the relationships among these levels are best interpreted as statistically supported associations rather than as a demonstrated causal sequence.