Multi-step structure-activity relationship screening efficiently predicts diverse PPARγ antagonists.

Area of Science:

• Computational Chemistry
• Pharmacology
• Drug Discovery

Background:

• Predicting antagonists for peroxisome proliferator-activated receptor gamma (PPARγ) is complex due to the wide structural variety of compounds.
• Conventional structure-activity relationship (SAR) methods struggle with this diversity.

Purpose of the Study:

• To develop and validate a multi-step SAR screening strategy for predicting PPARγ antagonists.
• To combine complementary computational methodologies for enhanced prediction accuracy.

Main Methods:

• Constructed three distinct SAR models: read-across-like SAR, docking-simulation-interpreting SAR, and deep-learning-based SAR.
• Integrated these models in a stepwise manner to classify potential PPARγ antagonists based on reliability levels.

Main Results:

• The read-across-like SAR model demonstrated the highest positive predictive value (PPV) for known antagonist scaffolds.
• Docking-simulation-interpreting SAR yielded high PPV and true-positive rate (TPR) by analyzing molecular surface features.
• Deep-learning-based SAR achieved the highest TPR in the final classification stage, capturing diverse antagonist structures.

Conclusions:

• Multi-step SAR screening effectively predicts PPARγ antagonists by integrating diverse computational approaches.
• This method enhances reliability by covering both well-defined and structurally diverse antagonistic compounds.
• The proposed multi-step SAR screening serves as a valuable tool for identifying potential PPARγ antagonists.