Artificial Intelligence Teaches Drugs to Target Proteins by Tackling the Induced Folding Problem

Area of Science:

• Computational Biology
• Drug Discovery
• Artificial Intelligence

Background:

• Proteins are not rigid; they adapt their structure in response to drug binding.
• Predicting these adaptive protein conformations is crucial for successful drug design.
• Current methods often overlook the dynamic nature of protein-ligand interactions.

Purpose of the Study:

• To develop a deep learning (DL) platform that guides drug design by inducing specific protein conformations.
• To address the challenge of predicting drug-induced protein structural changes.
• To integrate predictions of protein structural disorder into the drug design process.

Main Methods:

• A novel DL system was developed to predict protein conformations.
• The system integrates signals of structural disorder in flexible protein regions.
• It models how drugs stabilize specific protein structures, addressing the drug-induced folding problem.

Main Results:

• The DL platform can infer the ensemble of protein conformations induced by drug binding.
• It demonstrates the possibility of selecting a therapeutically relevant conformation through DL-guided drug design.
• Preliminary tests show the feasibility of AI-empowered prediction of drug-induced protein folding.

Conclusions:

• Deep learning offers a powerful approach to designing drugs that target adaptive protein structures.
• The developed platform represents a significant step towards AI-driven drug discovery by solving the drug-induced folding problem.
• This method enhances the prediction of drug-target interactions by considering protein conformational flexibility.