Research
Two routes to the same answer: DNA methylation, and now histology.
Our research
Everything we build predicts one thing: the tumour’s methylation class
A methylation class is a reproducible molecular identity: a group of tumours that share the same epigenetic fingerprint, and with it the same biology, behaviour and treatment path. Since the first CNS classifier in 2018, methylation classes have moved from a research finding into guidance from the WHO, EANO and NCCN.
Our research runs along two routes to that same answer. DNA methylation profiling measures the fingerprint directly, and remains our flagship. Histology AI predicts it from a slide the laboratory has already cut.
Two programmes
Methylation classification and histology AI
One classification framework, two kinds of input. The methylation classifiers are established and in routine research use; Hetairos, our histology programme, is newer and in early access.
Methylation classification
Tumour classification from the DNA methylation profile, using calibrated machine-learning models trained on reference cohorts of known entities.
- Platform-agnostic: input from any methylation platform, including sequencing
- 184 subclasses across CNS, sarcoma and skin
- Calibrated scores, CNV profiles and QC metrics with every run
- Reflected in WHO, EANO and NCCN guidance
Histology AI
Models that read a standard H&E slide and predict the methylation class directly, so that molecular testing can be targeted rather than ordered blind.
- Hetairos CNS CNS subtypes from FFPE H&E whole-slide images
- Molecularly defined classification for laboratories without molecular infrastructure
- Research use only: Hetairos available in early access
Where the research runs
Our full classifier portfolio is hosted for research use
Every classifier we develop is available on app.epignostix.com, our research-use-only platform. Upload a methylation profile from any supported platform and receive a calibrated classification, a CNV profile and QC metrics. Registration is free and access is usually approved within a few days.
Research use: hosted
The complete portfolio, including new and beta classifiers, on our platform.
Routine use: licensed locally
Validated classifiers installed inside your own infrastructure, fully offline.
Classifiers on app.epignostix.com are for research use only and are not intended for diagnostic procedures.
Collaborate with us
Working on a tumour type or a data type we don’t cover yet?
We develop classifiers together with clinical and research partners: new entities, new platforms, new modalities. If you have a cohort, a question or an idea, we would like to hear about it.
Read further
Where to next
Publications
The peer-reviewed work behind the classifiers, from Capper et al. 2018 onwards.
