Symptom co-occurrence modeling for phenotype-driven diagnosis in genetic diseases
Tool / method
Symptom co-occurrence model (similarity graph and non-negative matrix factorisation) for HPO-based gene prioritisation
Summary
The authors analysed HPO phenotyping in 1,686 patients with a confirmed genetic diagnosis from four international cohorts, and had 12 clinical geneticists encode the same three clinical reports. Only 43% of the 11,526 gene-phenotype associations in the cohort were present in HPO-structured databases; among patients sharing the same causal gene, only 22.7% of annotations were reused, and 47.3% of one physician's annotations had no ontological counterpart in another physician's rendering of the same report. PhenoGenius, a symptom co-occurrence model built from HPO databases and text mining of OMIM, MedGen and PubMed, summarises 16,600 HPO terms into 390 symptom groups and covers 98.4% of the cohort associations. Benchmarked against seven gene prioritisation tools and two reasoning language models (OpenAI o3, Gemini 2.5 Pro), it lowered the median rank of the causal gene from 65 (CADA, best comparator) to 35, with the advantage reproduced in each of the four cohorts, and placed the causal gene in the top 500 for 87.0% of patients versus 81.6%. On 4,716 curated case reports from the GA4GH Phenopacket Store, the causal gene ranked in the top 10 in 48.6% of cases.
Synthesis written by Geno'X. For the full original abstract, please refer to the source publication.
Analysis
The starting observation will resonate with any geneticist: two clinicians do not encode the same patient in the same way, and this variability, more than a lack of tools, limits phenotype-driven prioritisation; a strong argument for the model is that 34.8% of the associations added to HPO between 2022 and 2025 had already been predicted by the factorisation in 2022. A median rank of 35 nonetheless remains far from the top ranks actually reviewed in practice, and the tool works as a supporting filter rather than an answer.
Analysis by Dr Thibaut Benquey
Why this score?
Clinical impact: 2/3 · Evidence strength: 2/3 · Novelty: 2/2 · Sample size: 1/1 · Publication status: 1/1 → Total: 8/10
Keywords
Every Wednesday · Annotated selection · Free · Unsubscribe anytime