Defining neurovascular ecosystem states with single-cell multi-omics and machine learning: insights into cerebrovascular remodeling in the tumor context.

Objective: To define neurovascular ecosystem states within the tumor microenvironment and elucidate the regulatory basis of cerebrovascular remodeling through single-cell multi-omics and interpretable machine learning. Methods: An orthotopic murine brain tumor model was established. Single-cell RNA sequencing (scRNA-seq), single-cell assay for transposase-accessible chromatin sequencing (scATAC-seq), regional tissue validation, and integrative computational analyses were performed to characterize cellular heterogeneity, regulatory programs, and state transitions during tumor progression. Results: Tumor progression induced coordinated reprogramming across endothelial, mural, glial, and myeloid compartments. Trajectory inference modeled a progressive shift of endothelial cells from barrier-maintaining states toward angiogenic and inflammatory phenotypes, while mural cells exhibited transcriptional signatures indicative of matrix-remodeling adaptation. Extracellular matrix remodeling emerged as a shared neurovascular program with pronounced spatial heterogeneity. Epigenomic profiling further supported stable regulatory transitions, and interpretable machine learning prioritized key determinants associated with angiogenesis, barrier dysfunction, inflammation, hypoxic adaptation, and perivascular remodeling. Conclusion: Cerebrovascular remodeling in brain tumors reflects a dynamic, multicellular reorganization of the neurovascular ecosystem rather than an isolated vascular abnormality. These findings provide a systems-level framework for understanding tumor-associated vascular remodeling and its regulatory basis.