NHGNN-DTA: a node-adaptive hybrid graph neural network for interpretable drug-target binding affinity prediction

Area of Science:

• Computational chemistry
• Bioinformatics
• Machine learning

Background:

• Drug-target affinity (DTA) prediction is crucial for drug discovery.
• Current machine learning methods for DTA prediction either use sequence or structural information, with limitations in feature extraction and information interaction.

Purpose of the Study:

• To propose NHGNN-DTA, a node-adaptive hybrid neural network for interpretable DTA prediction.
• To combine the advantages of sequence-based and graph-based approaches for improved DTA prediction.

Main Methods:

• Developed a node-adaptive hybrid neural network (NHGNN-DTA).
• Employed a multi-head self-attention mechanism for model interpretability.
• Integrated adaptive feature representation and graph-level information interaction.

Main Results:

• Achieved state-of-the-art performance on Davis (MSE 0.196) and KIBA (0.124) datasets.
• Demonstrated superior robustness and effectiveness in cold-start scenarios compared to baseline methods.
• Provided new exploratory insights for drug discovery through model interpretability.

Conclusions:

• NHGNN-DTA offers a powerful and interpretable approach for DTA prediction.
• The model's ability to handle unseen inputs makes it valuable for drug discovery.
• Case study on Omicron variants highlights potential for drug repurposing in infectious diseases.