CADA: phenotype-driven gene prioritization based on a case-enriched knowledge graph.

Area of Science:

• Genomics and Bioinformatics
• Rare Disease Diagnostics
• Computational Biology

Background:

• Rare syndromes are identified by characteristic symptoms, often documented using Human Phenotype Ontology (HPO) terms.
• Existing HPO-based gene prioritization algorithms struggle with atypical or less frequent disease features.
• Electronic Health Records (EHRs) increasingly incorporate HPO terms, creating opportunities for data-driven diagnostics.

Purpose of the Study:

• To develop an improved gene prioritization tool, Cada, that addresses limitations of current algorithms.
• To enhance diagnostic accuracy for rare diseases, particularly for patients presenting with pathognomonic findings.
• To integrate phenotype-genotype information from clinical databases for more robust differential diagnostics.

Main Methods:

• Construction of a knowledge graph integrating both case and disorder annotations.
• Application of network representation learning for gene prioritization via link prediction.
• Incorporation of feature frequency information into the prioritization model.

Main Results:

• Cada demonstrates superior performance in gene prioritization, especially for patients with pathognomonic disease features.
• The tool effectively handles and incorporates information on the frequency of phenotypic features.
• Leverages phenotype-genotype data from databases like ClinVar for enhanced accuracy.

Conclusions:

• Cada offers a significant advancement in gene prioritization for rare disease diagnostics.
• The knowledge graph and network learning approach improves accuracy, particularly with characteristic symptoms.
• Cada serves as an ideal, updatable reference tool for differential diagnosis in rare genetic disorders.