Spatial-Spectral Fusion Enables Drug Repositioning by Capturing Indirect and Long-Range Associations in Biological...

Area of Science:

• Biomedical Informatics
• Computational Biology
• Network Science

Background:

• Drug repositioning aims to find new uses for existing drugs, accelerating clinical translation.
• Biomolecular networks often feature indirect therapeutic associations, challenging traditional detection methods.
• Existing shallow and deep learning methods struggle with indirect links and highly connected networks.

Purpose of the Study:

• To develop a novel framework for identifying indirect therapeutic associations in biomolecular networks.
• To overcome limitations of existing methods in capturing long-range and transitive drug-target-disease relationships.
• To improve the accuracy and scope of drug repositioning strategies.

Main Methods:

• A spatial-spectral collaborative framework integrating wave-evolution for spatial similarity propagation and spectral transforms for global connectivity.
• Contrastive learning to align spatial and spectral representations, capturing multi-hop transitive associations and long-range dependencies.
• Application to homogeneous similarity, and heterogeneous drug-protein-disease networks.

Main Results:

• The proposed method outperforms state-of-the-art baselines across multiple evaluation metrics on benchmark datasets.
• Successfully recovered non-explicit therapeutic associations in case studies for Alzheimer's and Parkinson's diseases.
• Validated findings through molecular docking, confirming the framework's efficacy.

Conclusions:

• The spatial-spectral collaborative framework effectively identifies indirect therapeutic associations, advancing drug repositioning.
• This approach offers a powerful tool for discovering novel drug indications by leveraging complex network structures.
• The method demonstrates significant potential for accelerating drug discovery and development pipelines.