Publications
The peer-reviewed research behind our classifiers.
Peer-reviewed research
The science behind our classifiers
Our classifiers are built on published, peer-reviewed research carried out with clinical and academic partners. These are the papers that define the reference cohorts, the class structures and the methods behind the products and research tools on this site.
Nature Cancer · 2026
Hetairos is a histology-based artificial intelligence model for predicting central nervous system tumor methylation subtypes
Jin et al. A deep learning model that predicts 102 methylation-based CNS tumour subtypes directly from digital H&E slides. Trained on 6,115 slides from 4,961 patients and externally validated on 5,498 slides from ten centres across four continents.
Cancer Cell · 2026
Advancing CNS tumor diagnostics with expanded DNA methylation-based classification
Sill et al. Describes the expanded CNS reference set and classifier generation behind the current release, covering 184 subclasses.
Nature Medicine · 2025
Prospective, multicenter validation of a platform for rapid molecular profiling of central nervous system tumors
Patel et al. Introduces the rapid profiling platform and the platform-agnostic MNP-Flex classifier that lets one model read array and sequencing data alike.
medRxiv · 2025 · preprint
Advancing sarcoma diagnostics with expanded DNA methylation-based classification
Jäger et al. The expanded sarcoma reference cohort and class structure behind the current sarcoma classifier.
Nature Communications · 2021
Sarcoma classification by DNA methylation profiling
Koelsche et al. Extends methylation classification beyond the CNS and establishes the sarcoma reference cohort.
Nature Protocols · 2020
Machine learning workflows to estimate class probabilities for precision cancer diagnostics on DNA methylation microarray data
Maros et al. The methodology behind the calibrated score. Benchmarks classification and calibration approaches on 2,801 brain tumour samples across 91 classes, so that a raw classifier output can be reported as a probability a pathologist can act on.
Nature · 2018
DNA methylation-based classification of central nervous system tumours
Capper et al. The paper that established methylation-based tumour classification and now underpins WHO, EANO and NCCN guidance.
The Lancet Oncology · 2017
DNA methylation-based classification and grading system for meningioma: a multicentre, retrospective analysis
Sahm et al. Shows that methylation classes carry prognostic information beyond histological grade, in a multicentre meningioma cohort.
Validation
How the classifiers have been tested
These studies set out to measure what methylation classification changes in diagnostic practice: how often it confirms the histological diagnosis, how often it refines it, and how often it changes it outright.
Neuropathology and Applied Neurobiology · 2025
Use of DNA methylation profiling as a molecular classification tool for paediatric central nervous system tumours: a middle-income country population-based study
Euzébio et al. Brazil. 182 paediatric patients at a reference oncology centre in Campinas; 74.2% of profiled samples reached a calibrated score of 0.9 or above, and profiling changed the diagnosis in 20.9% of classified cases.
Scientific Reports · 2025
Diagnostic impact of DNA methylation classification in adult and pediatric CNS tumors
Lebrun et al. Hôpital Universitaire de Bruxelles. In 70 tumours, 36 adult and 34 paediatric, classification confirmed the morphological diagnosis in 63% of adult and 23% of paediatric cases and refined it in 21% and 65% respectively; at high confidence scores it modified the diagnosis in 8% and 9%.
The American Journal of Surgical Pathology · 2024
Assessment of the utility of the sarcoma DNA methylation classifier in surgical pathology
Miettinen et al. National Cancer Institute. 619 well-characterised soft tissue and bone tumours across 62 entities. High sensitivity and specificity for fusion-driven sarcomas; lower sensitivity for desmoid fibromatosis, neurofibroma and schwannoma.
The Journal of Molecular Diagnostics · 2023
Implementation of DNA methylation array profiling in pediatric central nervous system tumors: the AIM BRAIN project
White et al. Eleven paediatric cancer centres across Australia and New Zealand. Of 269 patients profiled prospectively, 79.2% were classified at a score of 0.90 or above with classifier v12.5, and 74% of classified cases yielded significant diagnostic information.
Nature Medicine · 2023
Multiomic neuropathology improves diagnostic accuracy in pediatric neuro-oncology
Sturm et al. Prospective, population-based study of more than 1,200 newly diagnosed paediatric CNS tumours across 65 centres in Germany, Australia, New Zealand and Switzerland. Methylation class refined the diagnosis in 50% of cases, and clinically significant disagreement with histology was found in around 5% of patients, mostly among histological high-grade gliomas.
Neuropathology and Applied Neurobiology · 2022
DNA methylation profiling improves routine diagnosis of paediatric central nervous system tumours: a prospective population-based study
Schepke et al. Sweden. Every paediatric CNS tumour diagnosed nationally between 2017 and 2020, 240 tumours in total. 78% reached a high-confidence score; the histological diagnosis was confirmed in 69% and incongruent in 6%, and the change would have altered clinical management in 5% of all patients.
Neuro-Oncology · 2022
Impact of the methylation classifier and ancillary methods on CNS tumor diagnostics
Wu et al. National Cancer Institute, United States. Across 1,258 surgical neuropathology samples, 1,045 of them referred from outside institutions, the classifier affected the diagnosis in 46.7% of high-confidence cases and produced a substantially new diagnosis in 26.9%.
Neuropathology and Applied Neurobiology · 2020
Brain tumour diagnostics using a DNA methylation-based classifier as a diagnostic support tool
Priesterbach-Ackley et al. Centres across the Netherlands, Denmark, Finland, Sweden and Norway. Of 502 CNS tumour samples, the methylation class matched the initial pathological diagnosis in 54.4%, and refined it in a quarter of those. Profiling led to a new diagnosis in 9.8%, with an altered WHO grade in 71.4% of those cases.
The Lancet Child & Adolescent Health · 2020
DNA methylation-based profiling for paediatric CNS tumour diagnosis and treatment: a population-based study
Pickles et al. United Kingdom. In a diagnostic cohort of 306 patients under 20, methylation profiling made a unique contribution to the clinical diagnosis in 35% of cases and was estimated to change conventional management in 4%.
Acta Neuropathologica Communications · 2019
Methylation array profiling of adult brain tumours: diagnostic outcomes in a large, single centre
Jaunmuktane et al. National Hospital for Neurology and Neurosurgery, University College London. Of 325 referred cases, 179 reached a calibrated score of 0.84 or above; among those the diagnosis was changed in 25%, refined in 48% and confirmed in 25%.
Clinical Epigenetics · 2019
The central nervous system tumor methylation classifier changes neuro-oncology practice for challenging brain tumor diagnoses and directly impacts patient care
Karimi et al. University Health Network, Toronto. In 55 diagnostically challenging CNS tumours, 84% received a clinically relevant diagnostic alteration and patient care changed directly in 15% of cases.
Most of these studies were designed and published by academic centres without our involvement. They describe research use of the DNA methylation classifier in each centre’s own diagnostic setting rather than the performance of a specific Heidelberg Epignostix product release.
Reprints and requests
Looking for something not listed here?
If you need a reference that is not on this page, a reprint, or the citation for a specific classifier version, write to us and we will point you to it.
