Abstract Medulloblastoma, the most common malignant pediatric brain tumor, includes four distinct subtypes: Wnt, Sonic Hedgehog (Shh), Group 3, and Group 4. In contrast, ependymomas, which account for 10% of pediatric brain tumors and 4% of adult brain tumors, are categorized into three types: supratentorial (ST), posterior fossa (PF), and spinal (SP). While molecular subgrouping has significantly advanced the classification of these diseases, the extent of heterogeneity within these subgroups remains unknown. To address this, we collected bulk RNA sequencing (RNASeq) data from 888 medulloblastoma and 370 ependymoma from various institutions in North America and Europe. Our goal was to establish a comprehensive reference landscape for both medulloblastoma and ependymoma. We also included 42 forebrain and 52 hindbrain fetal samples to contrast against the tumor samples. Our analysis revealed that each disease subtype forms distinct clusters. Previously, subtyping was primarily achieved using methylation data; however, we demonstrate that bulk RNASeq data can also accurately identify disease subtypes. By examining gene expression patterns, gene fusions, and copy number aberrations, we validated existing molecular features and discovered new ones for each subtype. Additionally, we developed an innovative pipeline to identify specific biological patterns for a given region within the landscape. This resource coupled with mapping of new patients’ bulk RNASeq data onto the reference landscape, facilitating accurate disease subtype identification for new patients. This extensive dataset, available through the online tool Oncoscape, will assist the scientific community in uncovering novel therapeutic targets and advancing research in pediatric brain tumors.
Read full abstract