Drug-target affinity prediction using graph neural network and contact maps.

Area of Science:

• Computational chemistry
• Bioinformatics
• Machine learning

Background:

• Computer-aided drug design (CADD) accelerates pharmaceutical research.
• Drug-target affinity (DTA) prediction is crucial for efficient drug development.
• Deep learning advancements are enhancing DTA prediction accuracy.

Purpose of the Study:

• To propose a novel deep learning method, DGraphDTA, for accurate drug-target affinity prediction.
• To leverage molecular and protein structural information for improved DTA modeling.
• To enhance the efficiency and reduce the cost of drug discovery.

Main Methods:

• Constructing molecular and protein graphs using structural information.
• Employing graph neural networks (GNNs) for feature representation learning.
• Building protein graphs from predicted contact maps based on amino acid sequences.

Main Results:

• DGraphDTA demonstrates strong robustness and generalizability across benchmark datasets.
• The method effectively utilizes structural information for DTA prediction.
• Achieved high accuracy in predicting drug-target interactions.

Conclusions:

• DGraphDTA offers a promising approach for computer-aided drug design.
• The integration of GNNs with structural data improves DTA prediction.
• This method can significantly speed up the drug development pipeline.