Identification of Spatially Variable Genes with Weighted Leverage Scores

Abstract

Spatially variable (SV) gene detection plays a crucial role in analyzing spatial transcriptomicsdata for dissecting complex tissue microenvironments. Recently, a novel weighted leverage score (WLS) variable screening method was proposed to efficiently identify important features. Herein, this research introduced a new approach applying the WLS method together with a clustering algorithm, such as BayesSpace, to effectively identify SV genes. Through rigorous testing on simulated and rea-world datasets against established methods, the combination of the WLS method and BayesSpace consistently demonstrated superior accuracy in SV gene detection. As a robust, efficient, and flexible tool, the WLS method can be used to significantly improves the results in studying complex relationships between molecular cell functions and tissue phenotypes in spatial transcriptomics analysis.