SUMMARY While molecular subgrouping has revolutionized medulloblastoma classification, the extent of heterogeneity within subgroups is unknown. Similarity network fusion (SNF) applied to genome-wide DNA methylation and gene expression data across 763 primary samples identifies very homogeneous clusters of patients, supporting the presence of medulloblastoma subtypes. After integration of somatic copy-number alterations, and clinical features specific to each cluster, we identify 12 different subtypes of medulloblastoma. Integrative analysis using SNF further delineates group 3 from group 4 medulloblastoma, which is not as readily apparent through analyses of individual data types. Two clear subtypes of infants with Sonic Hedgehog medulloblastoma with disparate outcomes and biology are identified. Medulloblastoma subtypes identified through integrative clustering have important implications for stratification of future clinical trials.

Medulloblastoma (MB) is a pediatric malignant brain tumor composed of four different subgroups (WNT, SHH, Group 3, Group 4), each of which are a unique biological entity with distinct clinico-pathological, molecular, and prognostic characteristics. Although risk stratification of patients with MB based on molecular features may offer personalized therapies, conventional subgroup identification methods take too long and are unable to deliver subgroup information intraoperatively. This limitation prevents subgroup-specific adjustment of the extent or the aggressiveness of the tumor resection by the neurosurgeon. In this study, we investigated the potential of rapid tumor characterization with Picosecond infrared laser desorption mass spectrometry (PIRL-MS) for MB subgroup classification based on small molecule signatures. One hundred and thirteen ex vivo MB tumors from a local tissue bank were subjected to 10-to 15-second PIRL-MS data collection and principal component analysis with linear discriminant analysis (PCA-LDA). The MB subgroup model was established from 72 independent tumors; the remaining 41 de-identified unknown tumors were subjected to multiple, 10-second PIRL-MS samplings and real-time PCA-LDA analysis using the above model. The resultant 124 PIRL-MS spectra from each sampling event, after the application of a 95% PCA-LDA prediction probability threshold, yielded a 98.9% correct classification rate. Post-ablation histopathologic analysis suggested that intratumoral heterogeneity or sample damage prior to PIRL-MS sampling at the site of laser ablation was able to explain failed classifications. Therefore, upon translation, 10-seconds of PIRL-MS sampling is sufficient to allow personalized, subgroup-specific treatment of MB during surgery.Significance: This study demonstrates that laser-extracted lipids allow immediate grading of medulloblastoma tumors into prognostically important subgroups in 10 seconds, providing medulloblastoma pathology in an actionable manner during surgery.

No abstract