A Hypergraph powered approach to Phenotype-driven Gene Prioritization and Rare Disease Prediction

Area of Science:

• Computational biology
• Medical informatics
• Genomics

Background:

• Medical misdiagnosis remains a critical challenge in healthcare.
• Advancing medical technology necessitates improved diagnostic accuracy.
• Genomic data and phenotype information are crucial for disease identification.

Purpose of the Study:

• To reduce disease misdiagnosis rates through enhanced diagnostic accuracy.
• To develop a computational framework for phenotype-driven disease prediction.
• To leverage hypergraphs and genomic data for precise medical diagnoses.

Main Methods:

• Implemented a computational framework using hypergraphs and genomic data.
• Utilized robust ranking algorithms on a dataset of 2130 diseases, 4655 genes, and 9541 phenotypes.
• Sourced data from Orphanet and the Human Phenotype Ontology (HPO) database.

Main Results:

• Achieved a high accuracy rate of 98.09% for the top-ranked gene prediction.
• Captured 50% of causal genes within the top 10 predictions and 85% within the top 100.
• Outperformed existing tools like Phenomizer and GCN in accuracy and speed.

Conclusions:

• The hypergraph framework and ranking algorithms effectively improve disease diagnosis accuracy.
• Algorithmic hypergraph-based approaches show significant potential for advancing medical diagnostics.
• Future research should explore larger datasets and refined algorithms for broader application.