Spanve: an Statistical Method to Detect Clustering-friendly Spatially Variable Genes in Large-scale Spatial Transcriptomics Data

Depicting gene expression in a spatial context through spatial transcriptomics would be beneficial for inferring cell function mechanisms. The identification of spatially variable genes is a crucial step in leveraging the spatial transcriptome to understand intricate spatial dynamics. In this study, we developed Spanve, a nonparametric statistical method for detecting spatially variable genes in large-scale ST data by quantifying expression differences between spots and their spatial neighbours. This method offers a nonparametric approach to identifying spatial dependencies in gene expression without assuming specific distributions. Compared to traditional methods, Spanve decreases the number of false-positive outcomes, leading to more accurate identification of spatially variable genes. Furthermore, Spanve could facilitate downstream spatial transcriptomics analyses, including spatial domain detection and cell type deconvolution. These results show the broad applications of Spanve in advancing our understanding of spatial gene expression patterns within complex tissue microenvironments. Graphical abstract O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=200 SRC="FIGDIR/small/527623v3_ufig1.gif" ALT="Figure 1"> View larger version (52K): org.highwire.dtl.DTLVardef@1574cfcorg.highwire.dtl.DTLVardef@7f3f56org.highwire.dtl.DTLVardef@1732a3org.highwire.dtl.DTLVardef@fdaafa_HPS_FORMAT_FIGEXP M_FIG C_FIG