Abstract Therapy for glioblastoma (GBM) includes surgical resection, radiation, and chemotherapy, which functionally reduces tumor burden to a state of minimal residual disease (MRD). However, recurrence over time is universal. Large-scale longitudinal studies of GBM patient samples did not identify selection pressure for specific DNA mutations in recurrent disease, highlighting the importance of epigenetic mechanisms during relapse. In this study, we aim to model MRD and the functional and transcriptional effects of epigenetic perturbations in GBM patient derived xenografts (PDXs). We developed an in vivo inducible genome editing system by engineering 827 GBM stem cells to express doxycycline inducible Cas9 and a pool of 14 CRISPR sgRNAs targeting 5 epigenetic factors (readers/erasers of H3K27me3 and H4 acetyltransferase KAT5) and controls. Tumor cells are implanted orthotopically and mice are followed by serial MRIs until tumor volumes reach 2mm3, after which in vivo editing is induced by administration of doxycycline. Cells are harvested for scRNA-sequencing and targeted amplification of sgRNA libraries, allowing for assignment of sgRNAs to individual cells. Genome editing efficiencies ranged from 50-90% in vivo. Knockout of core Polycomb (PcG) complex members led to dysregulated expression of genes involved in therapeutic resistance and altered cell state distributions. Current directions include optimization of sgRNA targeting and recovery of CRISPR-edited cells from tumors for scaling up screens. In addition, we have developed a “standard of care” treatment regimen in GBM PDX models whereby temozolomide and radiation therapy is dose optimized to achieve a latent state that mimics MRD followed by eventual tumor recurrence. Future experiments will combine MRD modeling in PDXs with in vivo CRISPR screening to characterize functional dependencies of MRD in GBM. This data provides proof of concept for utilizing these tools for studying genotype to phenotype connections in a complex GBM model system.