DTGHAT: multi-molecule heterogeneous graph transformer based on multi-molecule graph for drug-target identification

Area of Science:

• Computational biology
• Pharmacology
• Bioinformatics

Background:

• Drug target identification is crucial for developing new therapies.
• Current computational methods often overlook the intricate relationships between drugs, targets, and other biomolecules.
• A comprehensive approach is needed to model these complex biological systems.

Purpose of the Study:

• To propose a novel prediction model, DTGHAT (Drug and Target Association Prediction using Heterogeneous Graph Attention Transformer), for identifying drug targets.
• To address the limitations of existing methods by incorporating heterogeneous biological network data.
• To enhance the accuracy and scope of drug target prediction.

Main Methods:

• DTGHAT employs a graph attention transformer architecture.
• The model analyzes 15 heterogeneous drug-gene-disease networks, integrating chemical, genomic, phenotypic, and cellular data.
• A 5-fold cross-validation was used to evaluate the model's performance.

Main Results:

• DTGHAT achieved an Area Under the Receiver Operating Characteristic Curve (AUC) of 0.9634, surpassing state-of-the-art methods by at least 4%.
• Ablation experiments confirmed the importance of integrating multi-source biomolecular data.
• A case study on cancer drugs demonstrated DTGHAT's efficacy in predicting novel drug targets.

Conclusions:

• DTGHAT represents a significant advancement in computational drug target identification.
• The model's ability to integrate diverse biological data enhances the prediction of drug-target interactions.
• DTGHAT offers a valuable, freely available tool for accelerating drug discovery and development.