BACKGROUND: According to the current WHO classification of CNS tumors, childhood CNS primitive neuro-ectodermal tumors (CNS-PNETs; WHO °IV) are poorly differentiated embryonal tumors with early onset and aggressive clinical behavior. Histological diagnosis can be complicated by morphological heterogeneity and divergent differentiation. Recent studies suggest the existence of molecular subgroups of CNS-PNETs sharing biological characteristics with other childhood CNS tumors. Here, we aimed at a comprehensive molecular characterization of CNS-PNETs and compared our results to profiles of other brain tumor entities in order to define the biological nature of tumors diagnosed as CNS-PNETs. METHODS: A collection of 211 fresh-frozen or paraffin-embedded tumor samples with an institutional diagnosis “CNS-PNET” was profiled for genome-wide DNA methylation patterns and copy-number alterations, complemented by transcriptomic profiling of a subset (n = 71). (Epi-)genetic profiles of CNS-PNETs were compared to those of >3000 other childhood brain tumors including embryonal, astrocytic, and ependymal entities, and their respective molecular subgroups. We screened selected groups of tumors for recurrent mutations and expression of established molecular markers. RESULTS: DNA methylation and gene expression profiles showed a clear segregation of pediatric brain tumors by histological entities and molecular subgroups. Interestingly, CNS-PNET profiles showed a significant overlap with various well-defined entities, including AT/RT, ETMR, high-grade glioma, medulloblastoma, and ependymoma. When screening CNS-PNETs with profiles highly resembling other entities, hallmark genetic alterations of these, such as amplification of 19q13.42, mutations in IDH1 or H3F3A, or mutations/deletions of the SMARCB1 locus, were frequently detected. Also, established protein markers, such as INI-1, LIN28A, and OLIG2, confirmed the reclassification of these CNS-PNETs. Strikingly, a subset (∼25%) of CNS-PNETs which could not be reclassified segregated into 3-4 distinct molecular subgroups, each with its own characteristic pattern of copy-number aberrations, DNA-methylation and gene expression. Currently, whole genome and RNA-sequencing of these distinct subgroups of CNS-PNETs is ongoing to reveal their underlying genetics. CONCLUSIONS: The correct classification of CNS-PNETs remains challenging. Based on the detection of recurrent genetic aberrations, many cases can be reliably re-classified, indicating that a significant proportion of CNS-PNETs may comprise a variety of other tumor subtypes. These findings suggest that the use of established and novel subgroups markers is needed in order to assist the histopathological evaluation of these tumors. In addition, we have identified a number of true CNS-PNET subtypes and are currently analyzing them in more detail in order to elucidate the genetics of these distinct groups. SECONDARY CATEGORY: Tumor Biology.