Improved predictive modeling of toxicity for organic compounds, including toxic organophosphates and polychlorinated cyclic compounds, in Labeo rohita using interpretable structural parameters.

Fish in aquatic ecosystems frequently encounter a range of biocides and contaminants due to ongoing chemical applications. Labeo rohita (rohu) is an important freshwater carp species, contributing over 15% to global carp production, which has seen significant growth in recent years. Existing predictive models, including quantitative structure-activity relationship (QSAR) and quantitative read-across structure-activity relationship (q-RASAR), depend on complex computer-based descriptors to evaluate the toxicity of organic compounds, including toxic organophosphates and polychlorinated cyclic organic compounds, on L. rohita. This study introduces interpretable structural parameters that significantly impact the pLC values (- log LC, where LC indicates the lethal concentration 50) for L. rohita. The new model shows improved predictive accuracy compared to traditional QSAR and q-RASAR models, achieving the high coefficient of determination (R2) and leave-one-out cross-validation score (Q) values that validate its reliability for future ecotoxicological applications. Additionally, applicability domain (AD) analysis demonstrates that the model can reliably predict toxicity for 297 compounds, reaching a 93% reliability rate, in contrast to the 81% reliability of conventional QSAR and q-RASAR models. This advancement makes it a valuable resource for filling toxicity data gaps in regulatory frameworks. Overall, this research enhances our understanding of the links between molecular structure and toxicity, facilitating the development of safer chemical alternatives and more effective predictive models for environmental risk assessment.