Sturgeon V2: a clinically applicable CNS tumor classifier trained on relabeled data
- 1Princess Maxima Center for Pediatric Oncology, Utrecht, The Netherlands; Oncode Institute, Utrecht, The Netherlands; Center for Molecular Medicine, UMC Utrecht,
Abstract
DNA methylation-based classification is increasingly a cornerstone of CNS tumor diagnostics. Recently,nanopore sequencing enables very rapid methylation profiling to support surgical decision making. Time-constrained sequencing yields sparse methylation profiles, necessitating specialized classifiers for accurate predictions. All publicly available sparse-profile classifiers rely on a single reference set of only 2,801 samples published in 2018 and its accompanying classification system. Here we present Sturgeon V2, trained on nearly 10,000 methylation profiles assembled from a range of public repositories and collaborations. To harmonize differences in tumor class labels, we developed an iterative semi-supervised relabeling strategy with a two-tiered concordance filter. We demonstrate that this filtering is essential for optimal performance. Compared to Sturgeon V1, we further redesigned the classifier architecture to provide more stable, and consistent confidence estimates while accumulating sequencing data and we introduce an updated class system including new classes, such as meningioma subtypes, PATZ-fusion neuroepithelial tumors and PLAG1/2-amplified embryonal tumors. In a retrospective intraoperative cohort of 94 pediatric cases, Sturgeon V2 was confidently correct in 85 cases (Sturgeon V1: 78) with one confident misclassification, and in simulated sequencing runs (n=1880) reduced the error rate among confident calls from 7.1% to 2.8%, outperforming publicly available alternatives. Sturgeon V2 is the first openly available, locally executable classifier for sparse methylation profiles trained beyond the 2018 reference era.
Related
- An Evaluation of the Guidelines for Premature Rupture of Membranes Based on the Appraisal of Guidelines for Research and Evaluation II, the Scientific, Transparent, and Applicable Rankings, and the Reporting Items for Practice Guidelines in Health Care Methods
- Functional Adaptations for Load-Bearing in a Dermal Bone: The Pectoral Fin Spine of the Russian Sturgeon (Huso Gueldenstaedtii).
- Nanoscale Characterization of Microscopically Layered Bone in the Russian Sturgeon Pectoral Fin Spine: A 3D FIB-SEM Study.
- Organ size in small infants (The OSSI Study): are sonographic reference intervals applicable in preterm infants small for gestational age?
- An XGBoost-SHAP classifier to distinguish asymptomatic hyperuricemia from acute primary gouty arthritis
BibTeX
@misc{vermeulen2026sturgeon,
title = {{Sturgeon V2: a clinically applicable CNS tumor classifier trained on relabeled data}},
author = {C. Vermeulen and H. van der Ent and S. L. Maas and E. M. Bos and K. Cammel and A. D. Maier and T. Mathiesen and M. E. G. Kranendonk and P. Wesseling and D. Seinstra and E.-J. Kooi and B. Stikker and H. J. Dubbink and M. Sie and O. H. J. Eelkman Rooda and E. W. Hoving and N. Verburg and M. Pages-Gallego and J. Hench and S. Frank and K. Aldape and O. Singh and M. Quezado and Z. Abdullaev and L. Kester and B. B. J. Tops and J. de Ridder},
howpublished = {medRxiv},
year = {2026},
month = oct,
doi = {10.64898/2026.10.19.26365910},
url = {https://www.medrxiv.org/content/10.64898/2026.10.19.26365910v1},
note = {CC BY-NC-ND},
}
RIS
TY - UNPB
TI - Sturgeon V2: a clinically applicable CNS tumor classifier trained on relabeled data
AU - C. Vermeulen
AU - H. van der Ent
AU - S. L. Maas
AU - E. M. Bos
AU - K. Cammel
AU - A. D. Maier
AU - T. Mathiesen
AU - M. E. G. Kranendonk
AU - P. Wesseling
AU - D. Seinstra
AU - E.-J. Kooi
AU - B. Stikker
AU - H. J. Dubbink
AU - M. Sie
AU - O. H. J. Eelkman Rooda
AU - E. W. Hoving
AU - N. Verburg
AU - M. Pages-Gallego
AU - J. Hench
AU - S. Frank
AU - K. Aldape
AU - O. Singh
AU - M. Quezado
AU - Z. Abdullaev
AU - L. Kester
AU - B. B. J. Tops
AU - J. de Ridder
PY - 2026
DA - 2026/10/19
PB - medRxiv
DO - 10.64898/2026.10.19.26365910
UR - https://www.medrxiv.org/content/10.64898/2026.10.19.26365910v1
AB - DNA methylation-based classification is increasingly a cornerstone of CNS tumor diagnostics. Recently,nanopore sequencing enables very rapid methylation profiling to support surgical decision making. Time-constrained sequencing yields sparse methylation profiles, necessitating specialized classifiers for accurate predictions. All publicly available sparse-profile classifiers rely on a single reference set of only 2,801 samples published in 2018 and its accompanying classification system. Here we present Sturgeon V2, trained on nearly 10,000 methylation profiles assembled from a range of public repositories and collaborations. To harmonize differences in tumor class labels, we developed an iterative semi-supervised relabeling strategy with a two-tiered concordance filter. We demonstrate that this filtering is essential for optimal performance. Compared to Sturgeon V1, we further redesigned the classifier architecture to provide more stable, and consistent confidence estimates while accumulating sequencing data and we introduce an updated class system including new classes, such as meningioma subtypes, PATZ-fusion neuroepithelial tumors and PLAG1/2-amplified embryonal tumors. In a retrospective intraoperative cohort of 94 pediatric cases, Sturgeon V2 was confidently correct in 85 cases (Sturgeon V1: 78) with one confident misclassification, and in simulated sequencing runs (n=1880) reduced the error rate among confident calls from 7.1% to 2.8%, outperforming publicly available alternatives. Sturgeon V2 is the first openly available, locally executable classifier for sparse methylation profiles trained beyond the 2018 reference era.
ER -