CO-CLUSTERING OF SPATIALLY RESOLVED TRANSCRIPTOMIC DATA.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Spatial transcriptomics measures gene activity and location within tissues.
• Understanding spatial gene variation is key to biological mechanisms like cell communication and tumor microenvironments.
• Current statistical tools lack methods to fully leverage spatial information for cell and gene clustering.

Purpose of the Study:

• To introduce SpaRTaCo, a novel statistical model for spatial transcriptomics data analysis.
• To enable coherent clustering of genes and tissue regions by integrating spatial information.
• To improve the understanding of biological processes through advanced spatial gene expression analysis.

Main Methods:

• Developed SpaRTaCo, a statistical model for co-clustering spatial gene expression profiles.
• Inferred latent block structures to simultaneously cluster genes based on expression patterns and tissue areas based on gene activity.
• Applied the model to spatial gene expression data, including human brain tissue processed with the 10X-Visium protocol.

Main Results:

• SpaRTaCo effectively clusters genes and spatial regions by leveraging expression patterns across tissue locations.
• The model's performance was validated through simulation experiments.
• Demonstrated the utility of SpaRTaCo in addressing biological questions using real-world spatial transcriptomics data.

Conclusions:

• SpaRTaCo provides a powerful new statistical approach for analyzing spatial transcriptomics data.
• The co-clustering method enhances the biological interpretation of gene expression patterns in their spatial context.
• This methodology advances the field by offering tools to better understand tissue architecture and cellular interactions.