Drug repositioning with metapath guidance and adaptive negative sampling enhancement

Area of Science:

• Computational biology
• Pharmacology
• Bioinformatics

Background:

• Drug repositioning accelerates drug discovery by identifying new uses for existing drugs.
• Computational methods are crucial for predicting drug-disease associations (DDAs).
• Existing DDA prediction methods face challenges with heterogeneous biological networks and negative sample quality.

Purpose of the Study:

• To introduce DRMGNE, a novel framework for DDA prediction.
• To address limitations in exploring diverse biological network relationships and negative sample quality.
• To accelerate drug discovery and repurposing.

Main Methods:

• DRMGNE utilizes an autoencoder for semantic feature extraction from similarity matrices.
• Metapath-guided learning and graph convolutional networks extract enriched node representations from subgraphs.
• Adaptive negative enhancement improves the quality and balance of negative samples for robust learning.

Main Results:

• DRMGNE demonstrated superior performance compared to state-of-the-art algorithms on three benchmark datasets.
• Case studies and molecular docking validations confirmed DRMGNE's efficacy.
• The framework shows significant potential for facilitating drug discovery and repurposing.

Conclusions:

• DRMGNE is an effective novel framework for DDA prediction.
• The combination of metapath-based guidance and adaptive negative enhancement yields superior results.
• DRMGNE holds promise for advancing drug discovery and repurposing pipelines.