geneEX: An Integrated Phenotype-Driven Algorithm for Rapid Identification of Causative Variants in Monogenic...

Area of Science:

• Genetics
• Bioinformatics
• Computational Biology

Background:

• Accurate identification of pathogenic variants is critical for diagnosing monogenic genetic disorders.
• Next-Generation Sequencing (NGS) has improved diagnostic efficiency but challenges remain in pinpointing causative variants due to data complexity.
• Current diagnostic methods face limitations in speed and accuracy for complex genetic data analysis.

Purpose of the Study:

• To develop an innovative phenotype-driven algorithm, geneEX, for enhanced identification of pathogenic variants in rare genetic disorders.
• To improve the efficiency and accuracy of diagnosing monogenic diseases through automated data analysis and interpretation.
• To leverage large language model technology for precise phenotype extraction and gene association.

Main Methods:

• Developed geneEX, a phenotype-driven algorithm integrating large language model technology.
• Implemented semantic vector representation for automatic Human Phenotype Ontology (HPO) acquisition and HPO-associated gene identification.
• Enabled semantic matching between patient free-text phenotypes and disease phenotypes for improved pathogenic gene discovery.
• Algorithm ranks candidate causative variants for rapid identification in rare genetic disorders.

Main Results:

• geneEX demonstrated strong performance in ranking pathogenic variants on both virtual and clinical datasets.
• Supplementary matching of free-text phenotypes significantly improved the precision of candidate variant prioritization.
• The algorithm effectively automates the process from clinical samples to pathogenic variant identification.

Conclusions:

• geneEX achieves automated HPO acquisition and full-process identification of pathogenic variants.
• Integrating free-text phenotypic descriptions enhances the accuracy of pathogenic gene identification.
• This approach significantly boosts the precision and efficiency of identifying pathogenic variants in rare genetic disorders, supporting monogenic disease diagnosis.