Extended Data Fig. 2: CRISPR validation of VIPER-inferred tumorigenic master regulators in diffuse midline glioma cell line models profiled by single-cell RNA sequencing.

a, One-sided Gene Set Enrichment Analysis (GSEA) showing enrichment of CRISPR-essential genes (defined by log₂FC ≤ 0 and FDR ≤ 0.05; red ticks) within the VIPER-inferred synthetic bulk tumorigenic protein activity signature of human diffuse midline glioma (DMG) malignant cells (GTEx caudate reference) across three DMG cell lines (DIPG17, DIPG4, SF8628). Normalized enrichment scores (NES) and p-values are shown for each line: DIPG17 (p = 5×10⁻³), DIPG4 (p = .027), and SF8628 (p = 0.01). b, FDR-adjusted p-values from one-sided GSEA assessing enrichment of CRISPR-essential genes identified in each DMG line and in the integrated set across all three lines, evaluated against synthetic bulk tumorigenic protein activity signatures for each of the seven DMG cell states. Oligodendrocyte precursor cell (OPC)-like tumorigenic master regulators (MRs) showed the most consistent enrichment, while SF8628-specific essential genes uniquely showed enrichment for astrocyte (AC)-like tumorigenic MRs. c, FDR-adjusted p-values from one-sided GSEA testing enrichment of CRISPR-essential genes within synthetic bulk tumorigenic protein activity signatures derived from malignant cells in each of the six patient samples profiled by single-cell RNA sequencing (scRNA-seq). DIPG4-specific essential genes uniquely enriched in the MUV5 patient sample, which shares the HIST1H3B mutation. d, Heatmap of patient-specific synthetic bulk tumorigenic protein activity for 37 CRISPR-validated essential MRs (rows) comprising the leading edge of enrichment in at least one DMG patient sample (columns). The annotation bar indicates patient-specific enrichment based on one-sided GSEA (black, FDR ≤ 0.05; gray, not significant). e, One-sided GSEA p-values for enrichment of essential genes from each CRISPR-screened cell line within tumorigenic protein activity signatures of six DMG cell lines profiled by bulk RNA-seq, supporting VIPER’s ability to nominate sample-specific essential MRs. f, UMAP projection of VIPER-inferred protein activity across 11,325 single cells from six DMG patient-derived cell lines profiled by scRNA-seq. Cell numbers and genotypes (histone and TP53 mutation status) are summarized below. g, UMAP from (f) colored by assigned human DMG cell state based on strongest statistically significant enrichment of state-specific MRs per cell (one-sided GSEA, FDR-adjusted). Cells not meeting significance (FDR > 0.05) are labeled unassigned, reflecting cell line-specific dependencies driven by ex vivo culture conditions.