From Transcripts to Cells: Dissecting Sensitivity, Signal Contamination, and Specificity in Xenium Spatial Transcriptomics

Abstract

Spatial transcriptomics has transformed our ability to map gene expression within intact tissues at cellular and subcellular resolution. Among current platforms, Xenium is widely adopted for its reliability, accessibility, and high data quality. Yet, the properties and limitations of Xenium-derived data remain poorly characterized. Here, we present one of the most comprehensive Xenium datasets to date, encompassing over 40 breast and lung tumor sections profiled using a diverse set of gene panels. Leveraging this resource, we systematically dissect technical noise--including transcript diffusion--alongside assay specificity, panel performance, and segmentation strategies. Our comparison of targeted panels with the newer 5K panel reveals that although the latter captures more transcripts overall, it suffers from reduced per-gene sensitivity and persistent diffusion, even with enhanced chemistry. We demonstrate that single-nucleus RNA-seq (snRNA-seq) markedly improves cell type annotation and enables more precise quantification of diffusion. Building on this, we introduce SPLIT (Spatial Purification of Layered Intracellular Transcripts), a novel method that integrates snRNA-seq with RCTD deconvolution to enhance signal purity. SPLIT effectively resolves mixed transcriptomic signals, improving background correction and cell-type resolution. Together, our findings provide a critical benchmark for Xenium performance and introduce a scalable strategy for signal refinement--advancing the accuracy and utility of spatial transcriptomics.