QSAR meets ecology: Predictive framework for assessing pesticide toxicity against mayfly using consensus modelling.

The extensive use of pesticides in agriculture holds serious toxicological risks to aquatic and terrestrial ecosystems. Traditional toxicity assessment methods are costly, time-consuming, and ethically constrained due to animal testing requirements. In this context, quantitative structure-activity relationship (QSAR) offers a distinct benefit to forecast the toxicity of substances. The current study builds partial least squares (PLS)-based QSAR models for evaluating the detrimental impacts of chemicals against mayfly (Ephemera vulgata) species, following OECD rules and using LD50 as a specified endpoint. Mayfly is useful in providing food for fish and is designated as a standard species that preserves freshwater habitats, indicating heavy metal contamination. The results from internal and external validation parameters (R2 = 0.684-0.689, Q2(LOO) = 0.648-0.661, R2pred or Q2F1 = 0.665-0.694, and Q2F2 = 0.661-0.690) confirmed that the produced model is stable and robust, as all the statistical parameters are very close to each other. Intelligent consensus prediction (ICP) is used to enhance the external predictivity and reliability of the model (R2pred or Q2F1 = 0.726 and Q2F2 = 0.722). In accordance with the information obtained, the presence of electronegative atoms, large fragments, aliphatic (CHR3) groups, less polar atoms, poisonous phosphate groups, and high number of long carbon chains are the principal biomarkers for pesticide toxicity. This framework was further used to screen the Pesticides Properties Database (PPDB) to evaluate the reliability of established models. This helps design eco-friendly chemicals, streamlines regulatory decision-making, confirms the external predictability of the established model, and advances the development of a sustainable earth.