Diffuse Midline Glioma (DMG) are universally fatal, primarily pediatric malignancies affecting the midline structures (i.e., pons, thalamus, and spinal cord) of the central nervous system. Despite decades of clinical trials, no drugs have emerged as effective against this disease, and treatment remains limited to palliative radiation therapy. Primary treatment challenges include: A) Well-stablished, yet non-actionable, genetic alterations; B) significant intratumoral heterogeneity, and C) blood-brain barrier (BBB) drug permeability. Here, we address the former two challenges by leveraging network-based methodologies to dissect the heterogeneity of DMG tumors and to discover Master Regulators (MR) proteins representing pharmacologically accessible, mechanistic determinants of molecularly distinct DMG cell states. We reverse engineered the first DMG gene regulatory network from 122 publicly available DMG RNA-seq profiles with ARACNe and inferred sample-specific MR protein activity with VIPER based on the differential expression of their targets. Nine of the top 25 most active MRs (i.e., FOXM1, CENPF, TOP2A, ASF1B, E2F2, TIMELESS, MYBL2, CENPK, TRIP13) comprise a well-characterized MR block (MRB2), frequently activated across aggressive tumors, and found to be enriched in DMG patient MR signatures (Fisher’s Exact Test p = 3.96x10-16). A pooled CRISPR/Cas9-mediated knockout (KO) screen across three DMG patient cell lines targeting 1,433 genes identified a set of 73 essential genes that were enriched in the MR signature of 80% of patient samples (GSEA p = 0.000034). FOXM1 emerged as a highly essential MR, significantly activated across virtually all patients. We then generated drug-induced differential protein activity from RNA-seq profiles following perturbation with 372 oncology drugs in two DMG cell lines that together recapitulate DMG patient MR and used this to identify drugs that invert patient MR activity profiles using the NYS/CA Department of Health approved OncoTreat algorithm OncoTreat predicted sensitivity to HDAC, MEK, CDK, PI3K, and tyrosine kinase inhibitors in subsets of patients, overlapping with published DMG drug screens. Importantly, 80% of OncoTreat-predicted drugs (p < 10-5) from three DMG patient tumor biopsies showed in vitro sensitivity in cultured tumor cells from the respective patients, with overall 68% accuracy among 223 drugs evaluated by both OncoTreat and in vitro drug screen (Fisher’s Exact Test p = 0.0449). Given known resistance in DMG to single-agent therapy, we further interrogated single-cell DMG regulatory networks generated by ARACNe with gene expression signatures from 3,039 tumor cells previously published across six patients using VIPER to infer single-cell regulatory protein activity. Unsupervised clustering of cells by protein activity defined 7 patient-independent cell states with distinct MR profiles reflecting known glial lineage markers (OPC-like-S1, OPC-like-S2, OC-like-S1, OC-like-S2, Cycling, AC-like, and AC/OPC-like). We identified drugs that invert the MR activity profiles of the individual cell states by using OncoTarget (inhibitors of individual MRs) or OncoTreat using the drug-induced differential protein activity we previously generated. Predicted drugs were distinct across the previously defined cell states with bulk RNA-seq recapitulating predictions seen in the more prevalent OPC-like stated, but failing to recapitulate the MRs and drug predictions for the smaller AC-like stated. We selected five drugs targeting the OPC/cycling-like cells (Trametinib, Dinaciclib, Avapritinib, Mocetinostat, and Etoposide), and four drugs targeting the AC-like cells (Ruxolitinib, Venetoclax, Napabucasin, Larotrectinib) for further validation as these states comprised most tumor cells across patients. We then generated single-cell RNA-seq for 95,687 cells after 5 days of treatment with either vehicle control (n = 4) or candidate drug (n = 2-3/drug) in subcutaneous SU-DIPG-XVII patient cell line-derived mouse models. We show this model recapitulates DMG cell states seen in patients, and confirm reduction in tumor growth and significant depletion of either OPC/cycling-like cells or AC-like cells in line with our drug predictions for 8/9 candidate drugs (Chi-square p<0.01). We further treated a syngeneic (DIPG4423) orthotopic DMG model with each drug and demonstrate significant differences in survival with Avapritinib, Dinaciclib, and Trametinib. Notably, the combination of drugs targeting OPC/cycling-like and AC-like cells (i.e. Trametinib+Ruxolitinib, Dinaciclib+Ruxolitinib, Avapritinib+Venetoclax, etc.) showed significantly lower tumor volumes after 2 weeks of treatment as compared to vehicles or each drug alone, and significant survival differences for some of the combinations. This work provides a precision medicine platform to nominate much-needed novel drug combinations addressing DMG tumor heterogeneity for further study to improve outcomes in this devastating disease.