DIPG-57. A systems biology approach to defining and targeting master regulator dependencies from bulk and single-Cell RNA-seq in diffuse midline glioma (DMG)

Abstract Diffuse midline glioma (DMG) are fatal pediatric brain tumors with no effective systemic therapies. Molecular profiling demonstrates epigenetic dysregulation and heterogeneity, and novel approaches are needed to identify promising drugs and drug combinations. We used network-based computational analysis of RNA-seq to discover Master Regulator (MR) proteins that represent targetable, mechanistic determinants of distinct DMG cell states. We reverse-engineered the first DMG-specific regulatory network from 122 publicly available DMG RNA-seq profiles with ARACNe. Using this network, we measured sample-specific protein activity based on differential expression of their targets via VIPER. Activity-based clustering identified two clusters showing a trend in survival differences (>1 year, by χ2). The most activated MRs (i.e., TOP2A, CENPF, BUB1B, FOXM1, GTSE1, MKI67, E2F8), relative to normal caudate tissue, were enriched in cell cycle regulation members. The cluster with worse outcomes had significantly higher activity. Targetable MRs activated in subsets of patients () included TOP2A, CHEK1, CDK2, and EZH2. RNA-seq profiles were generated in two DMG cell lines following perturbation with ~400 oncology drugs, and used to identify drugs that invert MR activity profiles with the OncoTreat algorithm. We identified four pharmacotypes, and predicted sensitivity to HDAC, proteosome, EGFR, MEK, and PI3K inhibitors (), amongst others, consistent with published DMG high-throughput drug screens. To dissect intra-tumor heterogeneity, we measured protein activity from published single-cell RNA-seq profiles of 6 DMG patients, using single-cell based regulatory networks. Preliminary activity-based analysis identified 6 cell states representing distinct stages of differentiation/proliferation, including an oligodendrocyte precursor cell-like proliferative state whose MRs overlapped with bulk data (i.e. TOP2A, CENPF, FOXM1, E2F8, ZWINT, CCNA2), suggesting this as a key regulatory module. We are identifying cell state-specific MR-inverter drugs with OncoTreat to ultimately suggest drugs and drug combinations that collapse the transcriptional program and induce tumor kill for validation in MR-matched preclinical DMG models.