Points2Regions: Fast, interactive clustering of imaging-based spatial transcriptomics data.

Area of Science:

• Computational Biology
• Bioinformatics
• Spatial Transcriptomics

Background:

• Spatial transcriptomics generates point-based mRNA data, requiring identification of biologically significant regions.
• Current methods for region identification are often scale-specific, require complementary data, or involve lengthy optimization.
• This limits their utility in exploratory analysis across different biological scales.

Purpose of the Study:

• To introduce Points2Regions, a novel computational tool for identifying regions with similar mRNA compositions in spatial transcriptomics data.
• To enable rapid, multi-scale region discovery without reliance on pre-segmented cells or extensive optimization.
• To provide a user-friendly tool for exploratory analysis of spatial transcriptomics data.

Main Methods:

• Feature extraction via rasterizing mRNA points onto a pyramidal grid.
• Efficient clustering using a hybrid approach of hierarchical and k-means clustering.
• Integration into TissUUmaps and as a Napari plugin for interactive visualization.

Main Results:

• Points2Regions achieves performance comparable to state-of-the-art methods on simulated datasets.
• The tool is significantly faster than existing methods and does not require segmented cells.
• Analysis of real-world datasets confirms the identification of biologically relevant regions consistent with prior studies.

Conclusions:

• Points2Regions offers a fast, efficient, and versatile solution for multi-scale region discovery in spatial transcriptomics.
• The tool enhances exploratory data analysis by enabling rapid identification of tissue structures and cell-type compositions.
• Its integration into existing platforms (TissUUmaps, Napari) improves user experience and accessibility.