Abstract BACKGROUND Medulloblastoma is the commonest malignant brain tumour in childhood, with progression-free survival (PFS) <65% in high-risk disease. Novel biomarkers, independent of current clinicopathological risk factors, could enhance understanding of biology and support improved treatment approaches. METHODS Using high-depth RNA sequencing of 217 patients, we identified novel multi-exonic medulloblastoma transcripts using a de novo analysis. We evaluated these alongside known/curated gene transcripts, including correlative survival analysis for 170 patients (aged 3-16 years and treated with curative intent). Multivariable survival analyses of the overall cohort (n=145) and a MBGroup3/MBGroup4 (Group3/Group4)-specific subcohort (n=95) was undertaken to identify prognostic transcripts, adjusting for current risk criteria and correcting for multiple testing. RESULTS From 3,150 novel transcripts identified, expression of 217 varied significantly and were further assessed. One was prognostic for PFS independent of subgroup (located on chromosome 20) while three were independently prognostic within Group3/Group4 (located on chromosome 1, 4 and 21 respectively). Transcript presence was validated by qPCR. From >60,000 known transcripts identified, 9693 were assessed. Four were independently prognostic for PFS in the whole-cohort (MRPS12, SLC25A34, MLYCD and ROGD1). In Group3/Group4, the top upregulated canonical pathways associated with PFS were: polo-like kinase, cyclin/cell cycle regulation, BRCA1/ DNA damage response, and glutamate receptor signalling. MELK and NQO2 were among the top 10 transcripts most highly associated with Group3/Group4 PFS. The prognostic significance of six transcripts (MRPS12, MLYCD, MELK, CDK4, NQO2, ROGD1) was validated in an independent Group3/Group4 dataset (n=470) while ROGD1 was prognostic independent of subgroup. Immunohistochemical validation showed low CDK4 expression, low MLYCD expression and high CHK1 expression were independent prognosticators in Group3/Group4. CONCLUSION De novo transcripts represent a significant component of the medulloblastoma transcriptome, providing prognostic information, and enhancing biological understanding of Group3/Group4 disease. CHK1 and MELK represent promising validated candidates for further evaluation as prognostic biomarkers in Group3/Group4 medulloblastoma.