Drug-Target Affinity Prediction Based on Topological Enhanced Graph Neural Networks

Area of Science:

• Computational chemistry
• Bioinformatics
• Machine learning

Background:

• Graph neural networks (GNNs) excel at drug-target affinity (DTA) prediction by analyzing molecular structures.
• Current GNN models often overlook crucial protein cavity information, limiting predictive accuracy.
• Drug development requires efficient and accurate methods for predicting interactions between drugs and protein targets.

Purpose of the Study:

• To develop a novel topology-enhanced GNN for improved DTA prediction.
• To integrate protein pocket data into GNNs for enhanced feature representation.
• To optimize GNN training and message-passing for better performance.

Main Methods:

• A novel topology-enhanced graph neural network (GNN) architecture was designed.
• Protein pocket structural data was incorporated into the GNN model.
• Training and message-passing strategies were optimized for feature learning.

Main Results:

• The proposed GNN model demonstrated superior performance in DTA prediction.
• Integration of protein pocket data significantly improved prediction accuracy.
• The model effectively captured complex drug-target interactions.

Conclusions:

• The topology-enhanced GNN offers a powerful approach for DTA prediction.
• Incorporating protein cavity information is vital for accurate drug-target interaction modeling.
• This method can accelerate drug discovery by improving prediction efficiency.