A novel approach to predict aquatic toxicity from molecular structure

Area of Science:

• Environmental Chemistry
• Toxicology
• Computational Chemistry

Background:

• Aquatic toxicity assessment is crucial for environmental protection.
• Existing experimental methods for toxicity testing are often time-consuming and expensive.
• Quantitative Structure-Activity Relationship (QSAR) modeling offers a predictive alternative.

Purpose of the Study:

• To develop and validate QSAR models for predicting aquatic toxicity.
• To utilize atom-based non-stochastic and stochastic linear indices for toxicity prediction.
• To compare the developed models with existing computational approaches.

Main Methods:

• Dataset of 392 benzene derivatives with toxicity data for Tetrahymena pyriformis.
• Development of QSAR models using multiple linear regression.
• Application of non-stochastic and stochastic linear indices.
• Validation using leave-one-out cross-validation and an external test set.

Main Results:

• Two statistically significant QSAR models were developed (R2=0.791, R2=0.799).
• Cross-validation yielded high predictive power (q2=0.781, q2=0.786).
• External validation confirmed model robustness (Rpred2=0.762, Rpred2=0.797).
• The developed models outperformed other methods in Dragon software.

Conclusions:

• Atom-based non-stochastic and stochastic linear indices are effective for QSAR modeling of aquatic toxicity.
• The developed QSAR models provide reliable predictions for benzene derivatives.
• This approach offers a cost-effective and time-efficient alternative to experimental toxicity testing.