Accurate inference of genome-wide spatial expression with iSpatial

Abstract

Spatially resolved transcriptomic analyses can reveal molecular insights underlying tissue structure and context-dependent cell-cell or cell-environment interaction. Due to the current technical limitation, obtaining genome-wide spatial transcriptome at single-cell resolution is challenging. Here we developed a new algorithm named iSpatial to derive spatial pattern of the entire transcriptome by integrating spatial transcriptomic and single-cell RNA-seq datasets. Compared to other existing methods, iSpatial has higher accuracy in predicting gene expression and their spatial distribution. Furthermore, it reduces false-positive and false-negative signals in the original datasets. By testing iSpatial with multiple spatial transcriptomic datasets, we demonstrate its wide applicability to datasets from different tissues and by different techniques. Thus, we innovated a computational approach to reveal spatial organization of the entire transcriptome at single cell resolution without the need of new technology development. With numerous high quality datasets available in the public domain, iSpatial provides a unique way for understanding the structure, function of complex tissues and disease processes. HIGHLIGHTSO_LIiSpatial infers genome-wide spatial gene expression pattern by integrating spatial transcriptomic and scRNA-seq data C_LIO_LIiSpatial outperforms existing approaches in inferring spatial gene expression patterns C_LIO_LIiSpatial reduces false-positive/negative signals of the original spatial transcriptome C_LIO_LIiSpatial is applicable to spatial transcriptomic datasets from different tissues and techniques C_LI