QSARs for toxicity of DDT-type analogs using neural network

Area of Science:

• Toxicology
• Computational Chemistry
• Insecticide Development

Background:

• DDT and its analogs are widely used insecticides.
• Understanding structure-toxicity relationships is crucial for developing safer and more effective alternatives.
• Previous studies have explored various analogs, but a comprehensive analysis using advanced modeling is needed.

Purpose of the Study:

• To analyze structure-toxicity relationships for 120 DDT-type insecticidal molecules.
• To compare the predictive accuracy of regression analysis (RA) and neural network (NN) models.
• To identify key molecular descriptors influencing toxicity against Musca domestica.

Main Methods:

• Regression analysis (RA) and neural network (NN) modeling were employed.
• Data from diverse literature sources on 120 DDT-type molecules were compiled.
• Analysis focused on diaryl nitropropanes (Prolan analogs), diaryl trichloroethane, and DDT isosteres.

Main Results:

• Steric factors were identified as critically important for the toxicity of all DDT-type analogs.
• Lipophilicity also plays a significant role, aiding neurotoxicant delivery to nerve sites.
• Neural network models demonstrated superior predictive accuracy compared to regression analysis.

Conclusions:

• Neural networks offer a more accurate approach for predicting DDT-type insecticide toxicity.
• Steric and lipophilic properties are key determinants of insecticidal activity.
• A novel method for analyzing descriptor roles using connection weights and residuals was proposed.