An Integrative Heterogeneous Graph Neural Network-Based Method for Multi-Labeled Drug Repurposing.

Area of Science:

• Computational biology
• Pharmacology
• Artificial intelligence

Background:

• Drug repurposing accelerates the discovery of new therapeutic uses for existing medications.
• Predicting novel drug-disease associations is crucial for efficient drug development.
• Existing computational methods require enhancement for complex biological networks.

Purpose of the Study:

• To propose DR-HGNN, an integrative heterogeneous graph neural network for multi-labeled drug repurposing.
• To discover new indications for existing drugs using a novel computational approach.
• To improve the prediction accuracy of drug-disease associations.

Main Methods:

• Constructed a heterogeneous drug-protein-disease (DPD) network using the DTINet dataset.
• Developed multi-label ranking approaches using neural networks on graph-structured data.
• Employed a GraphSAGE derivative (HinSAGE) to learn node embeddings and predict disease labels.

Main Results:

• The DR-HGNN method demonstrated superior performance compared to existing approaches.
• Achieved a high Area Under the Curve (AUC) of 0.964 on the DTINet dataset.
• Successfully predicted disease labels acting as bridges between drugs and proteins.

Conclusions:

• DR-HGNN offers a powerful and accurate method for multi-labeled drug repurposing.
• The approach effectively leverages heterogeneous network structures and node features for prediction.
• This method holds significant potential for accelerating the identification of new drug indications.