GliomaDeconv maps glioma-specific cellular programs from bulk transcriptomic profiles and reveals spatiotemporal associations with progression and treatment response.

Glioma recurrence and resistance to therapy remain major challenges in neuro-oncology, driven in part by profound cellular and spatial heterogeneity within the tumor microenvironment. Malignant and immune populations dynamically coexist and transition during disease progression, yet their coordinated organization remains poorly understood, and standard bulk RNA sequencing cannot resolve cell-type composition. To address these limitations, we established GliomaDeconv, a glioma-specific reference and interpretation framework for bulk RNA-seq deconvolution, derived from multi-subtype scRNA-seq datasets and implemented using established probabilistic deconvolution architecture. Application of GliomaDeconv to spatially annotated IVY GAP datasets and large bulk RNA-seq cohorts enabled systematic mapping of cellular composition, spatial organization, and dynamic evolution across glioma subtypes and disease stages. Single-cell analyses identified distinct malignant cell states, including a neural progenitor-like (NPC-like) population with strong tumor-propagating capacity, as well as multiple functionally specialized tumor-associated macrophage subsets enriched in grade- and subtype-specific patterns. GliomaDeconv-derived estimates showed biological concordance with single-cell-derived molecular programs in bulk datasets and suggested that histologically defined tumor regions are associated with distinct malignant and immune cell-state enrichments. Large-cohort analyses further identified subtype-specific cellular programs and therapy-associated phenotypic transitions. Functional validation in an orthotopic model showed that SNAP25-high/NPC-like glioma cells exhibited enhanced tumor-propagating capacity, supporting an association between this cell state and glioma progression. Collectively, GliomaDeconv bridges single-cell resolution and large-cohort transcriptomics, enabling precise dissection of glioma microenvironmental composition across subtypes and disease stages. The tool provides a practical framework for biological interpretation and is freely available at http://www.szflab.site/gliomadeconv/.