A Comprehensive Review of Data Analysis Methods for Spatially Transcriptomics

Spatially Transcriptomics (ST) has revolutionized biological research by capturing gene expression profiles while preserving spatial context, enabling in-depth analysis of tissue architecture and cellular interactions[1]. However, ST data’s multi-modal nature and complex spatial dependencies pose significant challenges for data analysis. This paper provides a comprehensive review of seven representative ST data analysis methods, including conST, DeepST, GraphST, HEARTSVG, STAMP, MuCoST, and SEDR. We systematically analyze each method’s core design principles, targeted challenges, and applicable scenarios. Quantitative evaluations on benchmark datasets demonstrate that these methods achieve superior performance in key tasks such as spatial clustering, spatially variable gene identification, multi-sample integration, and cell-type deconvolution. Finally, we discuss current limitations of ST analysis methods and outline future resear ch directions, including 3D ST data modeling, multi-omics integration, and clinical translation.