LAUSR

Medulloblastoma (MB) is the most common primary brain tumor in children, with group 3 and group 4 (G3/4) being the most aggressive. Methylation studies have identified eight distinct subgroups within G3/4 MB, with observed differences in survival patterns. However, the contribution of specific methylation-based epigenetic changes to the ontogeny of the disease remains poorly understood. We analyzed publicly available methylation profiles of 1, 370 G3/4 MB samples spanning all subgroups. To identify methylation probes important for classifying G3/4 MB into subgroups, we employed a wrapper-based feature selection machine learning (ML) algorithm on 400 samples. Subsequently, we trained a k-nearest neighbor (kNN)-based ML model to classify the remaining 970 samples, evaluating the sensitivity and specificity of the selected methylation probes. The kNN model achieved an accuracy of 90.41% using a panel of 541 methylation probes. We further refined this panel to 179 probes based on sensitivity and specificity metrics and observed a sequential alteration in beta values, suggesting a common developmental lineage for these subgroups. To identify specific genes driving the differentiation of the cells of origin or their transition to a cancerous state via methylation mechanisms, we selected a cohort of 460 samples that best represented the subgroups. Matched expression profiles for 103 samples were integrated with methylation data to infer candidate genes. To determine the spatiotemporal origins of these subgroups, we performed gene set enrichment analysis (GSEA) using published spatiotemporal gene sets of the developing human cerebellum. Our findings suggest that distinct subgroups originate from the Rhombic Lip Ventricular Zone, Intermediate Zone, Sub-ventricular Zone, and Choroid Plexus Epithelium. Further GSEA traced the cells of origin to Rhombic Lip - Progenitor, Unipolar Brush Cell - Progenitor, Differentiating and Defined, and Excitatory Cerebellar Neuron - Defined. In summary, we identified a prioritized methylation signature capable of accurately classifying G3/4 MB into subgroups and developed an R package, PriorityMeth, for further analysis. We also propose a developmental model outlining the origins of the subgroups of G3/4 MB and provide insights about the probable methylation-based epigenetic “switches” that drive the ontogeny of this group. Muhammad Awais Khan, Arsalan Riaz, Faisal F. Khan. Methylation-based epigenetic "switches" in group 3/4 medulloblastoma [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2025; Part 1 (Regular Abstracts); 2025 Apr 25-30; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2025;85(8_Suppl_1):Abstract nr 2618.