stEDGE enables edge-guided multiscale reconstruction of hierarchical spatial domains and transition interfaces in spatial transcriptomics

Spatial transcriptomic technologies enable high-resolution mapping of tissue architecture, yet most computational methods still represent tissues as flat spatial partitions, limiting their ability to resolve spatial domain boundaries, gradual transition regions and nested hierarchies. Here we present stEDGE, an edge-guided and interpretable framework for multiscale reconstruction of spatial domain hierarchies and transition-associated states in spatial transcriptomics. Unlike conventional domain-first approaches, stEDGE first estimates local boundary structure and uses it to guide the reconstruction of fine-grained spatial domains. We further introduce a domain transition index (DTI) to quantify domain-level transition propensity and integrate DTI with inter-domain similarity and boundary strength to organize fine domains into a coherent multilevel hierarchy. A tree-guided gene attribution strategy then distinguishes shared parent-level programs from branch-specific specialization across hierarchical spatial states. Across 15 benchmarked tissue sections spanning diverse tissues, technologies and spatial resolutions, stEDGE reconstructs developmental, inflammatory, tumor and brain architectures beyond conventional domain partitioning. Applied to cell-resolved Xenium profiling of fibrotic human lung, stEDGE further reveals lesion-associated remodeling as a coherent multiscale hierarchy in which remodeling interfaces coexist with a stable macrophage-associated airway/lumen state and spatially organized TLS-like lymphoid niches. Together, stEDGE provides a unified framework for modeling how stable compartments, domain boundaries, remodeling interfaces, localized niches and multiscale gene programs are jointly organized across spatial scales.