Clinical Scorecard: Employing GPT-4 for the Assessment of Severity in Phenotypic Abnormalities within the Human Phenotype Ontology

At a Glance

Key Highlights

• GPT-4 achieved true positive recall rates of 89% to 100% for phenotypic severity annotation.
• A novel severity scoring system was developed integrating clinical characteristics and their frequency.
• The HPO contains over 17,500 phenotypic abnormalities linked to more than 8,600 rare diseases.
• Automated curation significantly reduces the burden of manual expert annotation.
• Severity metrics facilitate prioritization of therapeutic targets in rare diseases.

Guideline-Based Recommendations

Diagnosis

• Utilize the HPO framework for comprehensive phenotypic descriptions.
• Incorporate severity metrics in differential diagnosis processes.

Management

• Prioritize treatment based on severity scores of phenotypic abnormalities.
• Use severity metrics to guide funding and research priorities.

Monitoring & Follow-up

• Regularly assess phenotypic severity to inform clinical decisions and patient management.

Risks

• Inadequate severity annotation may lead to misprioritization in treatment and research.

Patient & Prescribing Data

Patients with rare diseases exhibiting phenotypic abnormalities.

Severity metrics can guide therapeutic interventions and resource allocation.

Clinical Best Practices

• Employ automated tools like GPT-4 for efficient curation of clinical metadata.
• Integrate severity assessments into clinical workflows for improved patient outcomes.
• Engage healthcare professionals in the development of severity classification systems.

Related Resources & Content

• Human Phenotype Ontology
• GPT-4 Performance in Medical Contexts