Background: Medulloblastoma is one the most malignant paediatric brain tumours. Molecular subgrouping these medulloblastomas will not only help identify specific cohorts for certain treatment but also improve confidence in prognostic prediction.
Objective: Currently, there is a consensus of the existences of four distinct subtypes of medulloblastoma. We proposed a novel bioinformatics method, clustering of self-organizing map, to determine the subgroups and their molecular diversity.
Methods: Microarray expression profiles of 46 medulloblastoma samples were analysed and five clusters with distinct demographics, clinical outcome and transcriptional profiles were identified.
Results: The previously reported Wnt subgroup was identified as expected. Three other novel subgroups were proposed for later investigation.
Conclusions: Our findings underscore the value of SOM clustering for discovering the medulloblastoma subgroups. When the suggested subdivision has been confirmed in large cohorts, this method should serve as a part of routine classification of clinical samples.
Keywords: Medulloblastoma; gene expression; self-organizing map; subgroup.