Partner-Specific Drug Repositioning Approach Based on Graph Convolutional Network.

Area of Science:

• Computational biology
• Pharmacology
• Bioinformatics

Background:

• Drug repositioning accelerates drug development by repurposing existing drugs, reducing time and cost.
• Current computational methods generate static drug representations, failing to account for disease-specific mechanisms of action.
• Differentiated drug representations are needed to accurately target various diseases.

Purpose of the Study:

• To propose an end-to-end approach for partner-specific drug repositioning using graph convolutional networks (GCNs).
• To develop a method that learns distinct drug representations tailored to specific drug-disease associations.
• To enhance the accuracy of predicting novel therapeutic potentials for existing drugs.

Main Methods:

• Developed PSGCN, a graph convolutional network-based model for partner-specific drug repositioning.
• Extracted disease-specific context information from drug-disease association graphs.
• Implemented a layer self-attention mechanism to capture multi-scale information and a sortpool strategy for graph embedding.
• Formulated drug-disease association prediction as a graph classification task using a fully-connected module.

Main Results:

• PSGCN demonstrated superior performance over state-of-the-art methods on three benchmark datasets.
• Case studies on lung and breast cancers validated PSGCN's ability to identify actual drug-disease associations.
• The model effectively distinguished different disease contexts for the same drug, outperforming static approaches.

Conclusions:

• Partner-specific representation learning significantly improves drug repositioning outcomes.
• PSGCN offers a novel and effective computational strategy for identifying new therapeutic uses of existing drugs.
• The approach advances the field by enabling context-aware drug-disease association prediction.