Systematic design of combination therapy by targeting master regulators of coexisting diffuse midline glioma cell states.

Intratumor heterogeneity fundamentally challenges cancer treatment, as coexisting, molecularly distinct cell states with non-overlapping drug sensitivities can drive therapeutic resistance. We establish and validate a generalizable, network-based framework to systematically identify combination therapies targeting complementary tumor cell states. Applied to diffuse midline glioma (DMG)-a universally fatal pediatric malignancy-this approach identified master regulator protein dependencies in seven coexisting cell states, confirmed by pooled CRISPR-Cas9 assays. Perturbational transcriptional profiles for 372 clinically relevant drugs prioritized candidates predicted to invert state-specific master regulator activity. State-selective drug sensitivity was validated for eight out of nine (89%) drugs in vivo, including avapritinib, ruxolitinib and larotrectinib. Compared with monotherapy, co-administering drugs targeting complementary states significantly prolonged survival across virtually all combinations, with avapritinib plus ruxolitinib extending median survival nearly threefold versus vehicle and 1.5-fold versus avapritinib alone. These findings establish clinically actionable DMG combinations and a tumor-agnostic and mutation-agnostic framework for rational combination therapy design.