SGCD: High-Resolution Spatial Domain Characterization via Data Interpolation and Cell-Type Deconvolution.

Area of Science:

• Genomics
• Computational Biology
• Bioinformatics

Background:

• Spatial transcriptomics enables high-resolution tissue characterization.
• Traditional methods for spatial domain identification have limitations in utilizing inter-spot information and integrating prior cell type knowledge.
• Existing approaches often overlook crucial data between sampled spots in low-resolution spatial transcriptomic datasets.

Purpose of the Study:

• To introduce SGCD, a novel method for tissue spatial domain identification.
• To address limitations in traditional methods by incorporating data interpolation and cell type deconvolution.
• To achieve accurate delineation of complex spatial domains by integrating gene expression, cell type, and spatial location data.

Main Methods:

• SGCD employs data interpolation to estimate gene expression in regions between sampled spots.
• Cell type information is extracted using deconvolution from both spots and interstitial regions.
• Graph contrastive learning is utilized to integrate diverse data types for domain identification.

Main Results:

• SGCD significantly outperforms existing methods in accuracy and detail across multiple datasets.
• The method effectively delineates complex spatial domains in human and mouse tissues, as well as in cancer samples.
• Evaluations were conducted on human dorsolateral prefrontal cortex, mouse brain, pancreatic ductal adenocarcinoma, and breast cancer datasets.

Conclusions:

• SGCD offers a more comprehensive approach to spatial domain identification in transcriptomics.
• The method's ability to integrate inter-spot data and cell type information enhances understanding of tissue architecture.
• SGCD provides strong support for advancing the study of tissue functions and disease mechanisms through improved spatial analysis.