Spatial domains identification in spatial transcriptomics by domain knowledge-aware and subspace-enhanced graph contrastive learning

Yang Gui,Chao Li,Yan Xu

DOI: https://doi.org/10.1101/2024.05.09.593192

2024-05-11

Abstract:Spatial transcriptomics (ST) technologies have emerged as an effective tool to identify the spatial architecture of the tissue, facilitating a comprehensive understanding of organ function and tissue microenvironment. Spatial domain identification is the first and most critical step in ST data analysis, which requires thoughtful utilization of tissue microenvironment and morphological priors. To this end, we propose a graph contrastive learning framework, GRAS4T, which combines contrastive learning and subspace module to accurately distinguish different spatial domains by capturing tissue microenvironment through self-expressiveness of spots within the same domain. To uncover the pertinent features for spatial domain identification, GRAS4T employs a graph augmentation based on histological images prior, preserving information crucial for the clustering task. Experimental results on 8 ST datasets from 5 different platforms show that GRAS4T outperforms five state-of-the-art competing methods in spatial domain identification. Significantly, GRAS4T excels at separating distinct tissue structures and unveiling more detailed spatial domains. GRAS4T combines the advantages of subspace analysis and graph representation learning with extensibility, making it an ideal framework for ST domain identification.

Bioinformatics

What problem does this paper attempt to address?

### Problems the Paper Aims to Solve

This paper aims to address the issue of spatial domain identification in Spatial Transcriptomics (ST) data. Specifically:

- **Spatial Domain Identification**: The paper proposes a method based on a graph contrastive learning framework (GRAS4T), which combines contrastive learning and subspace modules to accurately distinguish different spatial domains by capturing the self-expressiveness of spots within the same domain.
- **Utilizing Tissue Microenvironment and Morphological Prior Information**: To better identify spatial domains, GRAS4T leverages tissue microenvironment and morphological prior information, thereby improving the accuracy of identification.
- **Improving Existing Methods**: Existing methods have shortcomings in identifying nearest neighbor relationships, leading to inaccurate spatial domain partitioning. GRAS4T addresses these issues by combining graph augmentation techniques and subspace analysis.

Overall, the goal of GRAS4T is to more accurately identify and distinguish different spatial domains in spatial transcriptomics data analysis, and it is applicable to various platforms and technologies.