What the study found
The study found that an unsupervised machine learning method called BERTopic, which is a topic-modelling tool based on BERT (Bidirectional Encoder Representations from Transformers), could surface known canine disease phenotypes and potential new patterns from large veterinary clinical notes data.
Why the authors say this matters
The authors say this matters because scalable and granular modelling of large clinical datasets could allow rapid review of many phenotypes in a population. They also state that it may enable early detection of temporal changes that could indicate emerging infectious or environmental diseases.
What the researchers tested
The researchers applied BERTopic to a large volume of clinical notes collected by the Small Animal Veterinary Surveillance Network (SAVSNET) from veterinary practices across the UK. The work aimed to create a representation of health-related phenotypes in one million dogs using unsupervised machine learning.
What worked and what didn't
The method was reported to surface known breed predispositions to hypoadrenocorticism, diabetes mellitus, and mitral valve disease. It also showed potential for identifying novel patterns of disease phenotypes, but the abstract does not provide detailed performance measures or say which candidate patterns were confirmed.
What to keep in mind
The abstract does not describe limitations in detail. It also does not provide quantitative accuracy results, and the mention of novel patterns is presented as potential rather than confirmed findings.
Key points
- BERTopic was used on veterinary clinical notes from SAVSNET in the UK.
- The method surfaced known breed predispositions to hypoadrenocorticism, diabetes mellitus, and mitral valve disease.
- The study reports potential novel patterns of disease phenotypes in dogs.
- The authors say the approach could support rapid interrogation of large clinical datasets and early detection of temporal changes.
Disclosure
- Research title:
- Topic modelling revealed known and potential canine disease phenotypes
- Authors:
- Peter‐John M. Noble, Sean Farrell, Noura Al-Moubayed, Alan David Radford
- Institutions:
- University of Liverpool, Durham University, Stanford University
- Publication date:
- 2026-02-24
- OpenAlex record:
- View
Get the weekly research newsletter
Stay current with peer-reviewed research without reading academic papers — one filtered digest, every Friday.