Discovering highly selective and diverse PPAR-delta agonists by ligand based machine learning and structural...

Area of Science:

• Pharmacology
• Computational Chemistry
• Drug Discovery

Background:

• Peroxisome proliferator-activated receptor delta (PPAR-δ) agonists improve fatty acid metabolism, glucose regulation, and physical endurance.
• PPAR-δ agonists show therapeutic potential for metabolic diseases, but none have reached clinical application.
• Developing novel, selective PPAR-δ agonists is crucial for metabolic disease treatment.

Purpose of the Study:

• To develop and apply a machine learning model for virtual screening of potential PPAR-δ agonists.
• To identify novel PPAR-δ agonist leads with high potency and selectivity.
• To explore new chemical scaffolds for PPAR-δ agonist development.

Main Methods:

• A ligand-based multi-filter ranking model, Iterative Stochastic Elimination (ISE), was developed using machine learning.
• Virtual screening of 1.56 million molecules was performed using the ISE model.
• Docking analysis and in vitro testing were employed to evaluate and validate potential PPAR-δ agonists.

Main Results:

• The ISE model successfully identified ~2500 top-ranking molecules from 1.56 million candidates.
• In vitro testing identified 13 molecules as potent PPAR-δ agonist leads with EC50 values between 4-19 nM.
• An additional 14 molecules showed activity with EC50 below 10 µM, and most nanomolar agonists exhibited high selectivity for PPAR-δ.

Conclusions:

• The machine learning-based virtual screening approach is effective in identifying novel PPAR-δ agonists.
• The identified leads possess high potency and selectivity, offering promising candidates for metabolic disease therapeutics.
• The discovery of structurally diverse agonists expands the chemical space for future PPAR-δ agonist development.