Novel Big Data-Driven Machine Learning Models for Drug Discovery Application.

Area of Science:

• Computational chemistry
• Structural biology
• Machine learning in drug discovery

Background:

• Drug discovery involves identifying chemicals that interact with protein targets.
• Virtual docking calculations use computational models to predict binding.
• Ensemble docking considers multiple protein conformations, but most are non-binding.

Purpose of the Study:

• To develop a machine learning approach for identifying protein conformations that bind to drug candidates.
• To address the class imbalance problem in predicting binding conformations.
• To accelerate the hit discovery phase in drug development.

Main Methods:

• Utilized machine learning to characterize properties of binding vs. non-binding protein conformations.
• Applied advanced machine learning techniques to handle class imbalance.
• Tested the approach on ADORA2A (Adenosine A2a Receptor) and OPRK1 (Opioid Receptor Kappa 1) proteins.

Main Results:

• Successfully characterized and identified properties of protein conformations selected by chemicals.
• Maximized prediction rates for potential protein molecular conformations.
• Demonstrated reduction in failure rates for virtual docking calculations.

Conclusions:

• The machine learning approach effectively distinguishes binding from non-binding protein conformations.
• This method enhances the accuracy and efficiency of the hit discovery process.
• Accelerates drug discovery by improving the prediction of potential drug candidates.