First report on a classification-based QSAR model for chemical toxicity to earthworm

Area of Science:

• Environmental Chemistry
• Toxicology
• Computational Chemistry

Background:

• Pesticide use in agriculture has escalated, leading to increased environmental concerns.
• Non-target organisms, such as earthworms (Eisenia foetida), are negatively impacted by pesticide toxicity.
• Understanding the structural basis of pesticide toxicity is crucial for developing safer alternatives.

Purpose of the Study:

• To develop a classification-based quantitative structure-activity relationship (QSAR) model for predicting earthworm toxicity.
• To identify specific molecular descriptors responsible for pesticide toxicity in non-target organisms.
• To provide insights for designing less toxic pesticides.

Main Methods:

• Utilized linear discriminant analysis (LDA) for classification-based QSAR modeling.
• Employed 2D molecular descriptors for model development after variable selection.
• Performed rigorous statistical validation, including sensitivity, specificity, and accuracy assessments.

Main Results:

• Developed a robust LDA model with 8 descriptors showing significant statistical validation (Wilks' λ=0.490, F=14.03).
• Achieved high accuracy on the training set (83.76%) and acceptable performance on the test set (71.74%).
• Identified lipophilicity, electron richness, and lower branching as key factors contributing to earthworm toxicity.

Conclusions:

• The developed QSAR model effectively predicts earthworm toxicity based on chemical structure.
• Molecular properties like lipophilicity and electron richness are critical determinants of earthworm toxicity.
• The model serves as a valuable tool for designing novel pesticides with reduced toxicity to earthworms.