Medulloblastoma (MB) is the most common malignant pediatric brain tumor. Current clinical trials for recurrent MB patients based on genomic profiles of primary, treatment-naïve tumors, provide limited clinical benefit since recurrent metastatic MBs are highly genetically divergent from their primary tumors. By adapting the existing Children’s Oncology Group treatment protocol for children with newly diagnosed high-risk MB for treatment of mice intracranially engrafted with human MB cells, we have characterized the rare treatment-refractory cell population in Group 3 MBs. MB cell populations recovered separately from brains and spines during the course of tumor development and therapy were comprehensively profiled for gene expression analysis, stem cell and molecular features to generate a global, comparative profile of MB cells through therapy to relapse. One of the most intriguing observations from our gene expression data was consistent over-expression of proteins belonging to Inhibitor of DNA-binding/differentiation (ID) family and a longevity associated protein bactericidal/permeability-increasing fold-containing-family-B-member-4 (BPIFB4) in our refractory population. The persistent upregulation of genes preserving undifferentiated state and cellular longevity further strengthens the hypothesis of stem-cell like cells driving tumor relapse in MB. Targeting BPIFB4 using both knockdown (KD) and knockout (KO) strategies have resulted in decreased proliferation and self-renewal of both primary and recurrent MB cells, further highlighting its potential as a novel therapeutic target in MB. Our differential genomic and gene expression profiles of the “treatment-responsive” tumors against those that fail therapy have successfully contributed to discovery and characterization of novel therapeutic targets for the most aggressive subgroup of MB.