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

Abstract

MotivationAccurate identification of spatial domains is essential for analyzing spatial transcriptomics data to elucidate tissue microenvironments and biological functions. Existing methods utilize either local or global spatial relationships between spots to aid domain segmentation. A method that can concurrently capture both local and global spatial information may improve identification of spatial domains. ResultsIn this article, we propose SECE, a deep learning-based method that captures both local and global relationships among spots and aggregates their information using expression similarity and spatial similarity. We benchmarked SECE against eight state-of-the-art methods on six real spatial transcriptomics datasets spanning four different platforms. SECE consistently outperformed other methods in spatial domain identification accuracy. Moreover, SECE produced spatial embeddings that exhibited clearer patterns in low-dimensional visualizations and facilitated more accurate trajectory inference. Availability and implementationSECE is implemented and provided as a pip installable Python package which is available on GitHub https://github.com/xie-lab/SECE.