Fig. 1.

Tissue metabolite profiles distinguish cerebellar tumour types and medulloblastoma groups. a) PCA scatterplot shows tissue type-specific metabolite profiles capable of separating medulloblastoma tumours (n = 86) from other posterior fossa tumours (pilocytic astrocytoma (n = 24), ependymoma (n = 18)) and normal cerebellar tissue samples (n = 7). b) Metabolite profiles are capable of separating the four main molecular groups of medulloblastoma through unsupervised PCA analysis. c) Unsupervised hierarchical clustering further validates and illustrates clustering of metabolite profiles by group, sub-structures of the data, and variations in metabolite concentrations between groups. Black boxes show tumours positive for the feature of interest, grey boxes indicate desmoplastic/nodular pathology (LCA histology track) or data not available (all other tracks). d) Annotated representative spectra show visual differences between the 4 groups. e) Alignment of unsupervised MRS classification and RNA-seq-based classification with ‘gold standard’ DNA methylation array-based classification shows tissue metabolite profiles perform equivalently to RNA-seq. Colour code for molecular groups applies to all part of the figure. G3, Group3; G4, Group4; LCA, large-cell/anaplastic.