MMADPE: drug repositioning based on multi-hop graph Mamba aggregation with dual-modality graph positional encoding.

Area of Science:

• Computational Biology
• Pharmacology
• Bioinformatics

Background:

• Drug repositioning (DR) is crucial for identifying new uses for existing drugs.
• Current computational DR methods struggle with high-order relationships, combined network properties, and biological interpretability.

Purpose of the Study:

• To propose a novel drug repositioning framework, MMADPE, addressing limitations of existing computational methods.
• To enhance the modeling of long-range drug-disease interactions and integrate topological structures with biological semantics.

Main Methods:

• MMADPE constructs a similarity network with multi-hop aggregation using Graph Mamba for efficient integration of multi-order neighborhood information.
• A dual-modal graph positional encoding captures global topology and local node strengths.
• A GraphGPS hybrid architecture fuses gated graph convolution and Transformer attention for feature extraction.

Main Results:

• MMADPE consistently outperforms state-of-the-art methods on three benchmark datasets for drug repositioning.
• Case studies and molecular docking experiments validate MMADPE's effectiveness.
• The framework provides mechanistic insights into novel drug-disease associations.

Conclusions:

• MMADPE offers a significant advancement in computational drug repositioning by integrating network structure and biological meaning.
• The framework demonstrates potential for discovering previously unrecognized therapeutic applications of drugs.
• MMADPE provides a more interpretable and effective approach to drug discovery.