CoCo-ST detects global and local biological structures in spatial transcriptomics datasets.

Area of Science:

• Computational Biology
• Genomics
• Bioinformatics

Background:

• Current spatial domain detection methods often miss subtle gene expression changes in low-variance regions.
• This limitation hinders the identification of early-stage tissue alterations.

Purpose of the Study:

• To introduce a novel framework, compare and contrast spatial transcriptomics (CoCo-ST), for enhanced spatial domain detection.
• To address the limitations of existing methods in identifying low-variance spatial structures.

Main Methods:

• CoCo-ST utilizes a graph contrastive feature representation framework.
• It compares target and background samples to detect both high-variance and low-variance structures.
• The method supports multisample integration, batch-effect correction, and scalability across various spatial transcriptomics technologies (e.g., Visium, Xenium).

Main Results:

• CoCo-ST demonstrated superior performance in identifying low-variance spatial structures compared to ten state-of-the-art algorithms in mouse lung precancerous samples.
• The framework effectively distinguished cell clusters and niche structures in Visium HD and Xenium Prime 5K data.
• CoCo-ST successfully detected subtle gene expression shifts overlooked by other methods.

Conclusions:

• CoCo-ST offers a significant advancement in spatial domain detection, particularly for identifying subtle biological features.
• The framework's technical advantages and broad applicability make it a valuable tool for spatial transcriptomics research.
• CoCo-ST enhances the ability to discover and analyze complex spatial patterns in tissues.