3DProtDTA: a deep learning model for drug-target affinity prediction based on residue-level protein graphs

Area of Science:

• Computational chemistry
• Bioinformatics
• Drug discovery

Background:

• Accurate drug-target affinity (DTA) prediction is crucial for efficient drug discovery.
• Computational methods accelerate early-stage drug development, reducing costs.
• Machine learning, particularly deep learning and graph neural networks, shows promise for DTA assessment.

Purpose of the Study:

• To propose a novel deep learning model, 3DProtDTA, for in silico drug-target affinity prediction.
• To leverage AlphaFold-predicted protein structures and graph representations for enhanced DTA modeling.
• To evaluate the performance of 3DProtDTA against existing DTA prediction methods.

Main Methods:

• Developed a deep learning model (3DProtDTA) incorporating protein structure information.
• Utilized AlphaFold predictions for protein structure data.
• Employed graph neural networks to represent molecular structures.
• Benchmarked the model against established DTA prediction datasets.

Main Results:

• 3DProtDTA demonstrated superior performance compared to existing DTA prediction models.
• The model effectively integrates protein structure data with graph-based molecular representations.
• Achieved state-of-the-art results on common DTA benchmarking datasets.

Conclusions:

• The proposed 3DProtDTA model offers a significant advancement in computational drug-target affinity prediction.
• Integrating AlphaFold structures enhances the accuracy of deep learning-based DTA prediction.
• 3DProtDTA shows potential for further development and application in accelerating drug discovery pipelines.