Development, validation, and utility of a clinically applicable methylation classifier for recurrence risk prediction in meningiomas

Meningiomas are common intracranial tumors with complex behavior that can be difficult to predict. Historically, morphology has been used to predict tumor aggressiveness and risk of recurrence, but this strategy has limitations as a prognostic tool. DNA methylation, transcriptomics, and copy number data are valuable for identifying groups of tumors with distinct biological signatures, thereby predicting recurrence risk. Multiple risk-stratifying classifiers which incorporate methylation data are available, but to date, a clinically validated risk-predicting classifier which exclusively uses methylation data has not been created.Using samples from 217 patients, we developed, validated, and implemented a clinically applicable methylation classifier for prognostic stratification of meningiomas based on k-means clustering of methylation data.Our classifier is 96% accurate, with 91% of samples receiving high confidence scores in the validation cohort (n = 76). This classifier is unique in that it includes de novo identification of risk groups by DNA methylation, confidence score calculation, internal clinical validation, and public model availability.Our newly validated classifier has the potential to aid diagnostic workup, improve recurrence risk prediction, and enhance clinical management of meningiomas.