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

Area of Science:

• Environmental Toxicology
• Computational Chemistry
• Ecotoxicology

Background:

• Pesticide use poses ecological risks, necessitating efficient toxicity assessment.
• Traditional methods are costly, time-consuming, and ethically problematic due to animal testing.
• Quantitative Structure-Activity Relationship (QSAR) offers a predictive alternative.

Purpose of the Study:

• Develop robust QSAR models to predict pesticide toxicity to mayfly (Ephemera vulgata) species.
• Utilize OECD guidelines and LD50 endpoint for toxicity assessment.
• Identify key molecular descriptors indicative of pesticide toxicity.

Main Methods:

• Constructed Partial Least Squares (PLS)-based QSAR models.
• Validated models using internal and external validation parameters (R², Q²(LOO), Q²F1, Q²F2).
• Employed Intelligent Consensus Prediction (ICP) to enhance model reliability and external predictivity.

Main Results:

• Models demonstrated stability and robustness with high statistical parameter agreement.
• ICP improved external predictivity (Q²F1=0.726, Q²F2=0.722).
• Identified biomarkers: electronegative atoms, large fragments, aliphatic groups, less polar atoms, phosphate groups, and long carbon chains.

Conclusions:

• The developed QSAR framework effectively predicts pesticide toxicity to mayflies.
• Identified biomarkers can guide the design of safer, eco-friendly chemicals.
• This approach supports regulatory decision-making and sustainable development.