Unveiling chemical space, scaffold diversity, critical structural features of pesticides: A comprehensive QSAR, qRASAR, machine learning studies to predict pesticides toxicity

The increasing use of pesticides in agriculture and urban areas has led to significant contamination of aquatic ecosystems, posing risks to non-target species. Fish, particularly the rainbow trout (Oncorhynchus mykiss), are highly vulnerable due to their gill permeability and ecological importance. As a sensitive and globally distributed species, rainbow trout serves as a key model in ecotoxicological studies and environmental monitoring. Here, a comprehensive cheminformatics workflow was employed to investigate the structural diversity and predictive modeling of a series of pesticides having acute aquatic toxicity to O. mykiss. The chemical space of these pesticides was explored through the Structure-Similarity Activity Trailing (SimilACTrail) map, revealing high structural uniqueness among pesticides, with several clusters exhibiting 80.0 %–90.3 % singleton ratios. A machine learning (ML) classifier model was developed using optimized hyperparameters, achieving robust predictive performance. Additionally, integrating Quantitative Structure-Activity Relationship (QSAR) and quantitative Read-Across Structure-Activity Relationship (q-RASAR) strategies enabled the construction of statistically reliable and mechanistically interpretable models, followed by toxicity data gap filling of 2000+ pesticides from external data sources. This integrated approach provides valuable insights into the structure–activity relationships (SARs) of pesticides and offers a predictive framework for future pesticide prioritization and environmental risk assessment. Developed q-RASAR model also provide an interpretable and reproducible alternative to fish testing, supporting regulatory prioritization efforts under USEPA and ECHA frameworks. At the same time, we recognize the limitations of the present work, including its focus solely on acute rainbow trout toxicity, potential uncertainty for structurally novel pesticides, and exclusion of chronic and mixture toxicity endpoints.