Structure-based drug design with geometric deep learning.

Area of Science:

• Bioorganic chemistry
• Medicinal chemistry
• Computational chemistry

Background:

• Structure-based drug design (SBDD) relies on 3D macromolecular structures to find ligands.
• Geometric deep learning (GDL) is an emerging machine learning technique applied to macromolecular structures.

Purpose of the Study:

• To review recent applications of GDL in bioorganic and medicinal chemistry.
• To highlight GDL's potential in SBDD.
• To forecast the future of GDL in drug discovery.

Main Methods:

• Review of recent literature on GDL applications in drug discovery.
• Focus on GDL for molecular property prediction.
• Emphasis on GDL for ligand binding site and pose prediction.
• Exploration of GDL for de novo molecular design.

Main Results:

• GDL shows significant potential across various SBDD tasks.
• Applications include predicting molecular properties and binding sites.
• GDL facilitates structure-based de novo molecular design.

Conclusions:

• GDL is a powerful tool for advancing SBDD.
• It offers opportunities for more efficient and effective drug discovery.
• Future research should focus on overcoming current challenges and exploring new GDL methodologies.