STAT3D as a new image-tailored Xenium-based spatial transcriptomics tool for accurately dissecting the tumor microenvironment

Abstract

Abstract Spatial transcriptomics enables researchers to study gene expression within tumor microenvironments at single-cell resolution while preserving tissue architecture. However, analyzing spatial omics data is challenging due to the complexity of the outputs, including large-scale images, millions of transcripts, and numerous spatial coordinates. Moreover, standard volumetric segmentation is often hampered by optical blurring and signal bleed-through in out-of-focus z-planes, which built-in tools in current platforms (e.g. Xenium Explorer) cannot solve. To overcome these challenges, we introduce STAT3D, a transparent, time-efficient, and customizable pipeline for analyzing whole-tumor biopsy slides in a user-tailored manner, including a 3D cell segmentation approach focused on the images’ sharpest z-layers. STAT3D addresses current limitations, such as generic image analysis and uniform downstream processing, by introducing an “image-tailored” strategy that selectively combines only the sharpest optical z-planes to calculate single-cell features and spatial distances. By integrating standard community tools (e.g. Seurat), the pipeline facilitates accurate multiplane cellular segmentation, tailored cell-to-transcript mapping, and fully automated singleR annotation. This versatile tool automatically streamlines critical steps in spatial omics analysis, which facilitates exploration of infiltrating cell types, interactions and tissue architectures by the scientific community.