Binding affinity prediction for protein-ligand complex using deep attention mechanism based on intermolecular...

Area of Science:

• Computational chemistry
• Structural biology
• Drug discovery

Background:

• Accurate prediction of protein-ligand binding affinity is crucial for cost-effective drug discovery.
• Existing scoring functions and machine learning methods have limitations in capturing complex protein-ligand interactions.
• Deep learning offers potential but requires efficient architectures and representations for protein-ligand complexes.

Purpose of the Study:

• To develop a deep neural network model for improved prediction of protein-ligand binding affinity.
• To address limitations of current methods by incorporating detailed interaction information.

Main Methods:

• Proposed a deep neural network model incorporating descriptor embeddings that capture local structural information.
• Implemented an attention mechanism to identify and emphasize key descriptors relevant to binding affinity.

Main Results:

• The proposed model demonstrated superior performance compared to existing binding affinity prediction models on benchmark datasets.
• The attention mechanism effectively highlighted important descriptors, contributing to enhanced prediction accuracy.

Conclusions:

• An attention mechanism can successfully identify binding sites within protein-ligand complexes, leading to improved prediction performance.
• The developed deep learning model offers a promising approach for accurate binding affinity prediction.
• The study's code is publicly available for further research and application.