Spatial-ID: a cell typing method for spatially resolved transcriptomics via transfer learning and spatial embedding

Abstract

Spatially resolved transcriptomics provides the opportunity to investigate the gene expression profiles and the spatial context of cells in naive state, but at low transcript detection sensitivity or with limited gene throughput. Comprehensive annotating of cell types in spatially resolved transcriptomics to understand biological processes at the single cell level remains challenging. Here we propose Spatial-ID, a supervision-based cell typing method, that combines the existing knowledge of reference single-cell RNA-seq data and the spatial information of spatially resolved transcriptomics data. We present a series of benchmarking analyses on publicly available spatially resolved transcriptomics datasets, that demonstrate the superiority of Spatial-ID compared with state-of-the-art methods. Besides, we apply Spatial-ID on a self-collected mouse brain hemisphere dataset measured by Stereo-seq, that shows the scalability of Spatial-ID to three-dimensional large field tissues with subcellular spatial resolution.

Stereopy: modeling comparative and spatiotemporal cellular heterogeneity via multi-sample spatial transcriptomics

“…To assess the stability and reliability of Stereopy’s CCD, we conducted a comparison with existing algorithms for domain detection on three samples: single sample Stereo-seq mouse embryo whole brain [3], Slide-seq V2 UMOD KI kidney comparative samples [28], and Stereo-seq multi-sample adult mouse brain [64] (Supplementary Note 2.1). For single sample, we included Giotto’s Spatial Domain Identification (GSDI) [15], SpaGCN [65] and GraphST [23]for comparison.…”

CellBin: a highly accurate single-cell gene expression processing pipeline for high-resolution spatial transcriptomics

“…4E showed Bin20 slicing of approximately 90% of the cells, and only about 10% of the cell nuclei were completely covered in the nuclei-stained image, while using StereoCell, only ~2% of the cells and most of the nuclei were included. The results of Bin20 and StereoCell were annotated by Spatial-ID 25 (Fig. 4F, right) with the Adolescent mouse brain as a reference 26 .…”

LETSmix: a spatially informed and learning-based domain adaptation method for cell-type deconvolution in spatial transcriptomics

“…Techniques such as graph networks (26,27) capitalize on spatial relationships, while domain adaptation methods (28,29) bridge technical variances between scRNA-seq and ST. DSTG (30) constructs a graph network, generating synthetic data from scRNA-seq to approximate true cell-type proportions in ST. Furthermore, self-supervised training with variational graph autoencoders applied in Spatial-ID (31), tissue histological image integration applied in SpaDecon (32), and domain-adversarial learning applied in CellDART (33) are innovative deep learning strategies, improving cell-type deconvolution accuracy in ST data.…”

“…However, these existing methods either overlook spatial correlations among individual sequencing spots or inadequately leverage ancillary information available in ST datasets such as the paired high-resolution histopathological images. Additionally, the inherent lower sensitivity of ST sequencing technology in detecting genes compared to scRNA-seq results in generally diminished gene expression levels (31). This discrepancy, termed the "platform effect", poses significant challenges for current cell-type deconvolution methodologies, demanding a high level of coherence between the reference scRNA-seq and ST data (34).…”