Application of Deep Learning for Studying NMDA Receptors

Area of Science:

• Artificial Intelligence in Chemistry
• Drug Discovery and Development
• Computational Chemistry

Background:

• Artificial intelligence (AI) has advanced significantly, offering new opportunities in drug development.
• Large pretrained models show human-level performance in specific tasks.
• Limited accessibility of AI tools hinders practical application for non-experts.

Purpose of the Study:

• To present three accessible online tools for drug development against Central Nervous System (CNS) diseases.
• To leverage the Uni-Mol large pretrained model for chemistry.
• To facilitate drug development targeting the NMDA receptor.

Main Methods:

• Developed three online tools based on the Uni-Mol large pretrained model.
• Included a blood-brain barrier (BBB) permeability prediction tool.
• Integrated a quantitative structure-activity relationship (QSAR) analysis system and an AI-based molecule generation model (VD-gen).

Main Results:

• Created user-friendly interfaces for complex AI models.
• Enabled prediction of BBB permeability for drug candidates.
• Provided capabilities for QSAR analysis and de novo molecule generation.

Conclusions:

• The developed tools bridge the gap between AI technology and NMDAR experts.
• These resources accelerate rapid and rational drug development for CNS diseases.
• The accessible AI platform supports efficient discovery of novel therapeutics.