Truace
Subscribe

An artificial intelligence framework for end-to-end rare disease phenotyping from clinical notes using large language models

Phenotyping is fundamental to rare disease diagnosis, but manually curating structured phenotypes from clinical notes is labor intensive and difficult to scale. Existing artificial intelligence (AI) approaches optimize individual components of phenotyping but generally do not operationalize the full workflow of extracting features from clinical text, standardizing them to Human Phenotype Ontology (HPO) terms, and prioritizing rare disease-relevant HPO terms. We developed RARE-PHENIX, an end-to-end AI framework f…

By the Truvace desk1 min read1 sourcePeer-reviewedTRV-2026-1594· cite & verify
An artificial intelligence framework for end-to-end rare disease phenotyping from clinical notes using large language models
Management of Cancer Pain: Adults, Quick Reference Guide for Clinicians Number 9 by U.S. Department of Health and Human Services, Public Health Service, Agency for Health Care Policy and Research. Public domain

In brief

Researchers built RARE-PHENIX, a framework that chains together three steps of rare disease phenotyping: large language model extraction of features from clinical text, mapping of those features to Human Phenotype Ontology terms, and supervised ranking of the terms most relevant to rare disease.

Phenotyping underpins rare disease diagnosis, yet manual curation is slow and hard to scale, so automating the whole workflow could matter for clinical genetics. The authors describe their results as suggestive and as supporting human review rather than replacing it, and the disease-prioritization signal is described only as potential.

Main points

  1. The framework combines LLM-based phenotype extraction, ontology-grounded HPO standardization, and supervised ranking of HPO terms.
  2. Training used data from 2671 patients across 11 Undiagnosed Diseases Network clinical sites.
  3. External validation used 16,357 real-world clinical notes from 143 patients at Vanderbilt University Medical Center.
  4. Output was highly concordant with clinician curation and showed potential downstream disease-prioritization signal.

The gain

RARE-PHENIX automates extraction, HPO standardization and ranking of phenotypes from clinical notes, producing terms highly concordant with clinician curation and reducing reliance on labor-intensive manual curation.

The rundown

The study starts from a bottleneck: manually curating structured phenotypes from clinical notes is labor intensive and difficult to scale, even though phenotyping is fundamental to diagnosing rare diseases.

Earlier AI approaches tended to optimize individual components and did not operationalize the full workflow from clinical text to standardized, prioritized HPO terms. RARE-PHENIX is presented as an end-to-end alternative that covers that whole workflow.

Testing moved from the multi-site Undiagnosed Diseases Network training cohort to a separate external set of notes at Vanderbilt University Medical Center. The authors conclude it may support human-in-the-loop phenotyping in real-world settings.

Sources

  1. Peer-reviewednpj Digital Medicine2026-10-15

The debate