Enabling Panoramic Integration of Spatial Multi-omics and Histology with GSMO

Panoramic integration of spatial multi-omics across molecular, cellular, and morphological layers is essential for achieving a coherent, multiscale understanding of tissue architecture and function. However, existing computational frameworks are constrained by limited scalability and therefore fail to reconcile the growing diversity, dimensionality, and cross-modality heterogeneity inherent in modern spatial omics data. Here, we present GSMO (Generalist Spatial Multi-Omics), a unified Transformer-based architecture for panoramic integration across spatial multi-omics and histology. GSMO establishes a modality-agnostic spatial representation that overcomes the dimensionality bottlenecks of multi-modal fusion, enabling seamless integration of any number and combination of spatial omics modalities with robust scalability and biological fidelity. In a landmark demonstration, GSMO advanced spatial multi-omics integration in an adult EAE mouse brain from the conventional two or three modalities to a record six, encompassing transcriptomics, protein expression, chromatin accessibility, two histone modifications, and histology. The resulting integrative analysis resolved fine-grained anatomical architectures and revealed complementary molecular signals and spatial heterogeneity that remain inaccessible to existing integration approaches. Furthermore, GSMO is released as a user-friendly open-source package that automatically performs data parsing, feature standardization, multi-modal alignment, spatial fusion, and visualization. We anticipate that GSMO will serve as a generalizable and accessible platform for the spatial multi-omics community, enabling deeper exploration of tissue organization, cellular identity, and disease mechanisms across diverse biological systems.