Brain Tumor DNA-Methylation Classifier
Overview
The brain tumor DNA-methylation classifier (Capper et al., Nature 2018) is a random forest classifier trained on genome-wide DNA methylation profiles from a large reference cohort of ~100 distinct CNS tumor types and subtypes. It assigns a tumor’s methylation profile to a CNS tumor methylation class and returns a calibrated prediction score (0 to 1). The classifier is available at www.molecularneuropathology.org and is considered a standard clinical tool for CNS tumor classification, particularly for histologically ambiguous cases.
Used by
- Applied to a urinary bladder EWSR1::BEND2 fusion sarcoma using Infinium MethylationEPIC v2.0 data; returned a low-confidence prediction score of 0.364 for the MN1-rearranged astroblastoma class — suggesting weak epigenetic similarity to MN1::BEND2-rearranged CNS tumors but below the threshold for reliable classification PMID:28199314
Notes
- Calibrated scores >0.9 are generally considered reliable for clinical classification.
- Scores between 0.3 and 0.9 (ambiguous range) indicate uncertain class membership and require clinical/histological correlation.
- The classifier includes astroblastoma, MN1-altered and related BEND2-rearranged subclasses that share epigenetic features with some extra-CNS FET-family fusion sarcomas.
- Reference: Capper D et al., Nature 2018; PMID: 29539639.
Sources
This page was processed by crosslinker on 2026-05-14.