Volume learning algorithm significantly improved PLS model for predicting the estrogenic activity of xenoestrogens

Area of Science:

• Computational chemistry
• Toxicology
• Artificial intelligence in drug discovery

Background:

• Xenoestrogenic compounds pose environmental risks.
• Accurate prediction of endocrine disruption is crucial for environmental safety.
• Existing methods for predicting chemical potency have limitations.

Purpose of the Study:

• To compare the predictive performance of Volume Learning Algorithm (VLA) artificial neural networks and Partial Least Squares (PLS) methods.
• To evaluate the models' ability to predict the relative potency of xenoestrogenic compounds.
• To assess the utility of CoMFA/VLA models for environmental pollutant testing.

Main Methods:

• Leave-one-out cross-validation procedure was employed.
• Comparison of VLA artificial neural network and PLS methods.
• Wilcoxon signed rank test was used for statistical analysis.

Main Results:

• VLA models significantly outperformed PLS models.
• VLA produced statistically superior predictive models.
• Models were validated on a diverse set of compounds across eight chemical families.

Conclusions:

• CoMFA/VLA models are effective for predicting endocrine-disrupting potential.
• These models can be applied to test environmental pollutants.
• VLA offers a robust approach for evaluating prospective chemicals before environmental release.