Precise Characterization of Cellular States and Spatial Variable Patterns within Spatial Transcriptomics

Spatial transcriptomics enables the simultaneous characterization of gene expression and spatial organization, providing transformative insights into tissue architecture and function. While numerous tools have been developed to identify spatial variable genes as proxies for functional variation, none can directly detect spatial variable patterns, such as celluar states and biological pathways. To address this gap, we introduce SVP (Spatially Variable Pathways), a computational framework for predicting functional cell states and analyzing their spatial variation. By generalizing spatial variable features to genes, cellular states, and biological pathways, SVP facilitates the co-distribution analysis of spatial features, enhancing biological interpretation and mechanistic exploration. Integrating graph propagation, hypergeometric testing, and advanced spatial statistics, SVP identifies spatially variable functions and uncover spatially resolved interactions. Evaluations on benchmark and real-world datasets demonstrate its accuracy and scalability. SVP has broad applications, including elucidating immune evasion in pancreatic cancer, tracking cardiac developmental dynamics, and investigating neurodegeneration in Alzheimer's disease model. Overall, SVP provides a robust framework for dissecting cellular and tissue-level organization.