Phenotype-driven gene prioritization for rare diseases using graph convolution on heterogeneous networks.

Area of Science:

• Genomic Medicine
• Bioinformatics
• Computational Biology

Background:

• Identifying causal genomic variants and their link to clinical phenotypes is a key goal in genomic medicine.
• Prioritizing variants from genotype data alone yields hundreds of candidates, complicating rare disease gene discovery.
• Relating genomic variants to clinical phenotypes for rare diseases remains a significant challenge.

Purpose of the Study:

• To develop a phenotype-driven gene prioritization approach for rare diseases.
• To leverage heterogeneous networks for improved causal gene identification.
• To create an integrated network, HANRD (Heterogeneous Association Network for Rare Diseases), for rare disease research.

Main Methods:

• Constructed a heterogeneous network integrating ontological and curated associations (genes, diseases, phenotypes, pathways).
• Employed graph convolution techniques to infer novel phenotype-gene associations.
• Integrated inferred associations into the network, creating HANRD.
• Validated HANRD on 230 rare disease clinical cases using patient phenotypes.

Main Results:

• HANRD identified causal genes within the Top-50 ranked genes for over 31% of cases.
• Causal genes were found within the Top-200 for more than 56% of cases.
• The approach demonstrated superior performance compared to existing state-of-the-art tools.

Conclusions:

• The heterogeneous network HANRD, incorporating curated, ontological, and inferred associations, enhances causal gene identification in rare diseases.
• HANRD's architecture supports future expansion with new data types and sources.
• This method offers a promising avenue for improving the diagnosis and understanding of rare genetic disorders.