CADA

CADA prioritizes candidate disease-causing genes by integrating disease-level and clinical case-level Human Phenotype Ontology (HPO) annotations into a gene–phenotype knowledge-graph and applying network representation learning for phenotype-driven gene prioritization in rare syndromes.

Key Features:

• HPO annotation integration: Integrates disease-level and clinical case-level Human Phenotype Ontology (HPO) annotations.
• Knowledge-graph construction: Builds a gene–phenotype association knowledge-graph from disorder and case annotations.
• Network representation learning: Applies network representation learning methods to embed graph nodes.
• Link prediction-based prioritization: Performs link prediction on learned embeddings to prioritize potential disease-causing genes.
• Bias mitigation via case annotations: Uses case-level annotations to reduce overrepresentation of atypical cases in literature-based disorder annotations.
• Pathognomonic focus: Emphasizes identification of genes associated with pathognomonic or hallmark symptoms highly specific to a disease.
• Phenotypic data integration: Incorporates phenotypic information from diagnostic laboratories and databases such as ClinVar.
• Reference support for differential diagnostics: Maintains an up-to-date reference for differential diagnostics in rare disorders through integrated case annotations.

Scientific Applications:

• Phenotype-driven gene prioritization: Prioritizes candidate genes for patients with rare syndromes using integrated HPO annotations.
• Differential diagnosis of rare genetic disorders: Supports differential diagnostics by ranking genes based on phenotypic match.
• Interpretation of pathognomonic presentations: Improves identification of causal genes for cases presenting hallmark pathognomonic findings.
• Gene–phenotype association research: Enables study of gene-phenotype associations derived from combined disorder and case annotations.

Methodology:

Integrates disease-level and clinical case-level HPO annotations, constructs a gene–phenotype knowledge-graph, applies network representation learning methods to obtain node embeddings, and performs link prediction on the embeddings to prioritize candidate disease-causing genes while incorporating phenotypic data from diagnostic labs and ClinVar.