Exploring novel disease-disease associations based on multi-view fusion network.

Area of Science:

• Computational biology
• Bioinformatics
• Systems biology

Background:

• Traditional disease classification relies on phenotypic observations, lacking molecular insights.
• There is a need to integrate multi-omics data for a deeper understanding of disease structure.

Purpose of the Study:

• To develop a flexible disease classification integrating biological process, gene expression, and symptom phenotype.
• To propose a multi-view fusion approach for constructing a disease-disease association network.

Main Methods:

• Applied a multi-view fusion approach to 223 diseases, creating 24 disease clusters.
• Analyzed internal and external cluster edges and compared results with Medical Subject Headings (MeSH).

Main Results:

• The proposed approach outperforms other integration methods in model performance.
• The generated clusters revealed novel and interesting disease-disease associations.
• The multi-view network captures disease relationships across multiple molecular levels.

Conclusions:

• This multi-view human disease association network redefines disease classification beyond tissue/organ limitations.
• The approach provides a powerful perspective for understanding disease molecular structure and exploring new therapeutic strategies.
• It extends existing disease taxonomy and motivates novel clinical and research directions.