Diagnosis of a Single-Nucleotide Variant in Whole-Exome Sequencing Data for Patients With Inherited Diseases: Machine...

Area of Science:

• Genomics
• Bioinformatics
• Machine Learning

Background:

• Next-generation sequencing (NGS) enables rapid whole-genome sequencing, driving its adoption in clinical diagnostics for hereditary disorders.
• Processing NGS data, particularly single-nucleotide variants (SNVs), involves complex bioinformatics pipelines and manual interpretation challenges.

Purpose of the Study:

• To develop an automated machine learning model for rapid interpretation of genetic variations identified through NGS.
• To assist physicians in efficiently identifying disease-causing variants from patient exome data, reducing manual review time.

Main Methods:

• Constructed a machine learning model using whole-exome sequencing (WES) and gene panel data, integrating variant annotations from multiple genetic databases.
• Trained and tested the model with WES data from 108 rare genetic disorder patients, incorporating phenotypic information extracted from electronic medical records.
• Utilized a keyword extraction tool to automatically process free-text phenotypic information for model input.

Main Results:

• The AI model successfully identified causative variants within the top 10 ranked candidates for 92.5% of patients.
• AI Variant Prioritizer achieved top-rank identification for 61.1% of patients, outperforming other prioritization tools.
• Incorporating Human Phenotype Ontology (HPO) terms improved the top 10 ranking accuracy to 93.5%.

Conclusions:

• Successfully integrated WES data and automatically extracted phenotypic information for variant interpretation model training and testing.
• The developed model demonstrates performance comparable to manual analysis, aiding in genetic diagnosis.
• The AI Variant Prioritizer has been implemented at National Taiwan University Hospital to support genetic diagnostic workflows.