Discovering disease-disease associations by fusing systems-level molecular data

Area of Science:

• Genomics
• Systems Biology
• Bioinformatics

Background:

• Traditional disease classification relies on shared genes.
• A shift towards systems-level integration of molecular data is emerging.
• Understanding complex disease relationships requires multi-modal data.

Purpose of the Study:

• To discover novel disease-disease associations by fusing molecular interaction and ontology data.
• To propose a multi-level hierarchy of disease classes integrating diverse molecular information.
• To identify the most significant data types for predicting disease relationships.

Main Methods:

• Data fusion of molecular interaction networks and biological ontologies.
• Development of a multi-level hierarchical disease classification model.
• Validation of novel disease associations using comorbidity data and literature curation.

Main Results:

• A novel multi-level disease hierarchy was established, showing significant overlap with existing classifications.
• 14 new disease-disease associations were identified, not currently present in the Disease Ontology.
• Human genetic interactions emerged as the most critical predictor of disease linkage, despite limited data.
• Omitting any data source diminished the prediction quality of disease relationships.

Conclusions:

• Systems-level data fusion is crucial for advancing disease classification and understanding.
• Integrating diverse molecular data, especially genetic interactions, enhances the discovery of disease relationships.
• The proposed approach offers a powerful framework for uncovering novel comorbidity patterns and therapeutic targets.