SECE: accurate identification of spatial domain by incorporating global spatial proximity and local expression proximity

Abstract

AbstractIdentification of spatial domains is a critical step in spatial transcriptomics (ST) analysis. Here, we introduce SECE, a method that incorporates global spatial proximity and local expression proximity to identify spatial domains. SECE can also obtain spatial features for low-dimensional visualization and trajectory inference. In addition, SECE can identify cell types when dealing with ST data with single-cell resolution. Compared with eight spatial domain identification and five cell type identification methods, SECE consistently achieved top performance on spatial and cellular characterization, accurately providing biological insights into ST data.