Rigorous 3-dimensional spectral data activity relationship approach modeling strategy for ToxCast estrogen receptor...

Area of Science:

• Environmental Chemistry
• Toxicology
• Computational Chemistry

Background:

• Estrogenic compounds pose risks to environmental and human health.
• Predictive modeling is crucial for screening large chemical libraries for estrogenic activity.

Purpose of the Study:

• To model the estrogenic potential of chemical compounds using a 3D-SDAR approach.
• To identify structural determinants of estrogenicity.
• To evaluate the predictive performance of the developed model.

Main Methods:

• Modeled 1528 compounds from the ToxCast database using 3D-SDAR.
• Augmented fingerprints with indicator variables for oxygen-containing functional groups.
• Validated the model with a blind test set of 2008 compounds provided by the EPA.

Main Results:

• Achieved predictive accuracy of 0.62, sensitivity of 0.71, and specificity of 0.53 on the blind test set.
• Identified key structural features for estrogenicity: phenolic OH or cyclohexenone, a second aromatic/phenolic ring, a methyl group near a phenol ring, and a proximate carbonyl group.
• The 3D-SDAR model showed minimal performance reduction between modeling and prediction sets.

Conclusions:

• The 3D-SDAR model effectively predicts estrogenic potential.
• Specific structural features are critical for estrogenic activity.
• This approach aids in identifying potentially harmful chemicals.