Inferring spatial single-cell-level interactions through interpreting cell state and niche correlations learned by...

Area of Science:

• Computational Biology
• Genomics
• Bioinformatics

Background:

• Cell-cell interactions (CCI) are crucial for tissue development and organ function.
• Spatial transcriptomics (ST) enables studying CCI at single-cell resolution but faces challenges in data analysis.
• Existing methods struggle with limited ligand-receptor data, spatial encoding, and interpretability.

Purpose of the Study:

• To introduce GITIII, a self-supervised graph transformer model for enhanced CCI analysis.
• To address limitations in current ST data analysis for inferring cell communication.
• To provide an interpretable framework for understanding spatial CCI patterns.

Main Methods:

• Developed GITIII, a lightweight, interpretable, self-supervised graph transformer model.
• Conceptualized cells as words and their neighborhoods as context for gene expression analysis.
• Inferred CCI by correlating cell states with their microenvironment (niche).

Main Results:

• GITIII successfully inferred and visualized spatial CCI patterns across diverse ST datasets.
• The model enabled statistically interpretable analysis of how sender cells influence receiver cells.
• GITIII facilitated CCI-informed cell clustering and the construction of cell communication networks.

Conclusions:

• GITIII offers a powerful and interpretable approach to unraveling cell-cell interactions from spatial transcriptomics data.
• The model effectively identified key CCI patterns in complex biological systems like brain and tumor microenvironments.
• GITIII advances the analysis of spatial omics data for understanding tissue function and disease.