Abstract We’ve curated a comprehensive multi-omics database of medulloblastoma (MB) serving as a resource to understand spatiotemporal organization and guide therapeutic strategies. Herein, our analysis encompasses single-nucleus transcriptome (n = 51), chromatin accessibility (n = 46), and spatial transcriptomic profiles (n = 31) from human MB samples and patient-derived xenograft lines spanning four subgroups, augmented by bulk RNA (n = 322), whole-genome short-sequencing (n = 279), and bisulfite sequencing (n = 300) from matched samples. We delineate the malignant cellular hierarchy, identifying MB cancer stem-like cells (MB-CSCs), cycling cells, and more differentiated populations. Gene signatures of different cell subpopulations strongly correlate to clinical outcomes, while spatial character of patient-derived materials emphasizes the geographically heterogeneous nature of MB. Examination of somatic copy number variants (CNVs) and transcriptional signatures at single-nucleus resolution reveals geographically defined subclones harboring distinct CNVs, suggesting a competitive agglomeration of co-existing multi-clones for MB. Alongside the trajectorial transition, we observe distinct chromosomal alterations in apical MB-CSCs compared to more differentiated cells, indicative of extensive subclone-primed spatiotemporal evolution. Further analysis of reagent-gene interactions of subclonal lineages contributes to predicting the druggable candidates for treatment optimization. Integration of spatial architecture data highlights the recapitulation of subclone-determined niches through colocation with identified malignant lineages and their own inflammatory or fibrotic adaptions. Taken together, this comprehensive database, comprising five or six kinds of datatypes on the same MB cases, facilitates combinatorial analyses across different genomic modalities, offering genetic and geographic insights into therapeutic advancements.