Drug repositioning with adaptive graph convolutional networks.

Area of Science:

• Computational biology
• Pharmacology
• Artificial intelligence

Background:

• Drug repositioning accelerates the identification of new therapeutic uses for existing drugs.
• Graph convolutional networks (GCNs) are increasingly used in drug repositioning, but often struggle to deeply integrate node features and topological structures.
• Limitations in current GCN methods can hinder their effectiveness in predicting novel drug-disease associations.

Purpose of the Study:

• To propose an adaptive GCNs approach, named AdaDR, for improved drug repositioning.
• To deeply integrate node features and topological structures for enhanced predictive capabilities.
• To identify novel drug-disease associations through exploratory analysis.

Main Methods:

• Developed AdaDR, an adaptive GCNs approach for drug repositioning.
• Employed an adaptive graph convolution operation to model interactive information between node features and topological structures.
• Utilized an attention mechanism to learn adaptive importance weights for feature and structure embeddings.

Main Results:

• AdaDR demonstrated superior performance compared to multiple baseline methods in drug repositioning tasks.
• The adaptive integration of node features and topological structures enhanced the model's expressive power.
• Exploratory case studies provided insights into potential novel drug-disease associations.

Conclusions:

• AdaDR offers a promising advancement in GCN-based drug repositioning by effectively integrating diverse data modalities.
• The adaptive approach enhances the accuracy and utility of computational methods in drug discovery.
• The study facilitates the identification of new therapeutic applications for existing drugs, potentially shortening the drug development timeline.