Summary Accurate pathological diagnosis is crucial for optimal management of cancer patients. For the ~100 known central nervous system (CNS) tumour entities, standardization of the diagnostic process has been shown to be particularly challenging - with substantial inter-observer variability in the histopathological diagnosis of many tumour types. We herein present the development of a comprehensive approach for DNA methylation-based CNS tumour classification across all entities and age groups, and demonstrate its application in a routine diagnostic setting. We show that availability of this method may have substantial impact on diagnostic precision compared with standard methods, resulting in a change of diagnosis in up to 12% of prospective cases. For broader accessibility we have designed a free online classifier tool (www.molecularneuropathology.org) requiring no additional onsite data processing. Our results provide a blueprint for the generation of machine learning-based tumour classifiers across other cancer entities, with the potential to fundamentally transform tumour pathology.

Glioblastoma (GBM) is a brain tumor that carries a dismal prognosis and displays considerable heterogeneity. We have recently identified recurrent H3F3A mutations affecting two critical amino acids (K27 and G34) of histone H3.3 in one-third of pediatric GBM. Here, we show that each H3F3A mutation defines an epigenetic subgroup of GBM with a distinct global methylation pattern, and that they are mutually exclusive with IDH1 mutations, which characterize a third mutation-defined subgroup. Three further epigenetic subgroups were enriched for hallmark genetic events of adult GBM and/or established transcriptomic signatures. We also demonstrate that the two H3F3A mutations give rise to GBMs in separate anatomic compartments, with differential regulation of transcription factors OLIG1, OLIG2, and FOXG1, possibly reflecting different cellular origins.