A Fast and Robust Tool for Identifying Spatially Variable Features in High-Resolution Spatial Omics

Spatially resolved transcriptomics and emerging spatial omics technologies enable comprehensive exploration of biological systems in their native tissue context, capturing molecular features with spatial coordinates in two- and three-dimensional space. Identifying spatially variable genes and features is critical for linking molecular functions to tissue phenotypes, yet existing computational methods often lack reliability, scalability, and the ability to handle high-resolution, three-dimensional data. Here we present a spatial granularity-based approach and a corresponding open-source, user-friendly tool, scBSP, for identifying spatially variable features in high-resolution spatial omics data. Leveraging sparse matrix operations, scBSP achieves exceptional computational efficiency, processing datasets with millions of cells or spots in seconds on standard desktops. Its non-parametric approach ensures robustness and accuracy across diverse sequencing platforms and resolutions. Validated through simulations and case studies in cancer, neuroscience, rheumatoid arthritis, and kidney disease, scBSP reveals biologically meaningful spatial patterns, offering insights into critical pathological mechanisms and advancing the potential of spatial omics research.