AOP-Guided Toxicogenomic Mapping Profiles for Mechanism-Informed Toxicity Pattern Prediction in Zebrafish

Environmental pollutants often lack specific biological targets but trigger a wide range of effects. This complexity has traditionally necessitated a battery of laborious and time-consuming bioassays in traditional screening processes. Here, we developed an adverse outcome pathway (AOP)-guided chemical toxicogenomic mapping profiles (AOP-CTMP) framework to identify key effects of environmental pollutants. We integrated all 15 recognized AOPs for aquatic vertebrates related to population growth rate to construct comprehensive AOP networks. Based on this network, a series of two-dimensional gene matrices was generated, and the chemical perturbations were further mapped onto these matrices to generate the corresponding AOP-CTMPs, thereby enabling the prediction of primary effects. This mapping leveraged an established database encompassing 566 compounds, their genomic perturbations, key events, and adverse effects. Among the modeling approaches evaluated, the multilayer perceptron model demonstrated excellent performance (AUC = 0.89-0.98) by converging effectively during the fitting process and exhibited high accuracy (0.83-0.90) in internal validation. We subsequently applied AOP-CTMP to predict the adverse outcomes of 13 representative chemicals and validated the predictions on a series of end point assays through in vitro and in vivo experiments. The individual-level effects predicted through AOP-CTMP showed strong consistency with actual observed toxicity patterns, confirming the reliability and practical potential of this approach. The results underscore that AOP-CTMP, enabling prediction of toxicity patterns from chemical perturbations, offers a novel and systematic strategy for key effects identification during toxicity screening.