Predicting gene disease associations with knowledge graph embeddings for diseases with curtailed information

Area of Science:

• Biomedical informatics
• Computational biology
• Genomics

Background:

• Knowledge graph embeddings (KGE) represent complex biological knowledge in low-dimensional spaces.
• Understanding KGE applications for prioritizing genes in complex diseases with limited genetic data is crucial.

Purpose of the Study:

• To develop and evaluate novel KGE algorithms for prioritizing disease-associated genes.
• To assess the impact of data quality and integration on prediction accuracy.
• To apply KGE for identifying genes linked to Intervertebral Disc Degeneration (IDD).

Main Methods:

• Constructed a biomedical knowledge graph (KG) integrating heterogeneous data.
• Generated KGE using state-of-the-art methods and two novel algorithms: Dlemb and BioKG2vec.
• Validated embeddings using unsupervised clustering and supervised prediction methods.

Main Results:

• KGE successfully predicts genes associated with complex diseases.
• The novel Dlemb and BioKG2vec algorithms outperformed existing methods in gene prioritization.
• Data quality, preprocessing, and integration significantly impact prediction accuracy.
• Prioritized genes for IDD showed enrichment for disease-relevant functions.

Conclusions:

• KGE is a valuable tool for gene prioritization in biomedical research.
• The developed Dlemb and BioKG2vec algorithms offer improved performance for disease gene discovery.
• Emphasis on data quality and integration is essential for robust KGE applications in genomics.