Development and evaluation of a deep learning model for protein-ligand binding affinity prediction

Area of Science:

• Computational chemistry
• Drug discovery
• Machine learning

Background:

• Structure-based ligand discovery is crucial for drug development.
• Machine learning (ML) and deep learning are increasingly used to enhance drug discovery pipelines.
• Deep learning models can automatically learn relevant features for complex tasks.

Purpose of the Study:

• To develop a novel deep neural network for estimating ligand-receptor binding affinity.
• To represent ligand-receptor complexes in a 3D grid format for analysis.
• To apply 3D convolution techniques for feature extraction from atomic representations.

Main Methods:

• Developed a deep neural network architecture.
• Utilized 3D grid representation for ligand-receptor complexes.
• Employed 3D convolution for feature mapping.
• Treated protein and ligand atoms uniformly within the model.

Main Results:

• The developed network accurately estimates binding affinity.
• Performance was evaluated on the CASF-2013 'scoring power' benchmark and Astex Diverse Set.
• The deep learning model demonstrated superior performance compared to classical scoring functions.

Conclusions:

• The novel deep neural network is effective for binding affinity estimation.
• This approach advances structure-based ligand discovery using deep learning.
• The model and associated resources are publicly available for use.