Poster #72: Minhe Liu

Effect of RNA Diffusion on Spatial Transcriptomics Data

Minhe Liu, Yang Lab, Department of Genetics, School of Arts and Sciences, Rutgers University-New Brunswick, Piscataway, NJ, Chuhanwen Sun, Yang Lab, Department of Genetics, School of Arts and Sciences, Rutgers University-New Brunswick, Human Genetics Institute of New Jersey, Rutgers University-New Brunswick, Jiekun Yang, Department of Genetics, School of Arts and Sciences, Rutgers University-New Brunswick, Human Genetics Institute of New Jersey, Rutgers University-New Brunswick, Computer Science and Artificial Intelligence Laboratory, Massachusetts Institute of Technology, Broad Institute of MIT and Harvard

Spatial transcriptomics can profile gene expression in its native spatial context at a sub-cellular resolution, but transcripts originating from one cell or tissue region may bleed into neighbors through RNA diffusion. The extent of this RNA displacement is difficult to quantify because the true origin of individual transcripts is generally unknown. Characterizing the spatial range of RNA diffusion is therefore important for distinguishing local expression from contamination and improving the interpretation of high-resolution spatial transcriptomics data. To quantify this problem, we analyze a Visium HD human-mouse xenograft dataset in which species-specific transcripts provide a marker of RNA origin. We calculate human- and mouse-specific transcript counts across 2 µm spatial bins and map these signals onto H&E histology. After manually identifying the human–mouse tissue interface, we measure cross-species RNA signal as a function of distance into the opposite tissue. Cross-boundary signal decays sharply with distance and is largely confined within approximately the first 100 µm from the interface, providing an empirical estimate of the spatial range over which RNA diffusion contributes to noise in Visium HD data. We then apply this distance-based characterization to a Xenium–Visium HD human dataset to examine its generalizability in single-species tissue. By creating a direct approach to analyze the spatial extent of cross-boundary RNA signal and how RNA diffusion influences spatial transcriptomics data, we are able to better understand and adjust for RNA diffusion. Incorporating this spatial effect may improve downstream analyses that depend on accurate localization of transcripts, such as cell segmentation, cell-type assignment, and spatial gene-expression analysis.