DHAG-DTA: Dynamic Hierarchical Affinity Graph Model for Drug-Target Binding Affinity Prediction.

Area of Science:

• Computational chemistry
• Bioinformatics
• Drug discovery

Background:

• Drug-target binding affinity (DTA) prediction is crucial for identifying potential therapeutics.
• Deep neural networks (DNNs) show promise for DTA prediction, especially when only sequence data is available.

Purpose of the Study:

• To propose DHAG-DTA, a novel dynamic hierarchical affinity graph DNN approach for DTA prediction.
• To leverage molecular sequence information and known drug-target interactions for improved prediction accuracy.

Main Methods:

• DHAG-DTA utilizes a two-level hierarchical graph: an affinity graph for inter-molecular interactions and embedded molecular graphs for intra-molecular interactions.
• Key innovations include a unified hierarchical graph, dynamic affinity graph structure determination, skip connections for information fusion, and robust feature embeddings for unseen molecules.

Main Results:

• DHAG-DTA demonstrated superior performance compared to existing models on two benchmark datasets.
• The model achieved state-of-the-art results across multiple evaluation metrics.

Conclusions:

• DHAG-DTA effectively integrates inter- and intra-molecular interaction information for accurate DTA prediction.
• The proposed method advances computational approaches in drug discovery and development.