Multiscale topology classifies cells in subcellular spatial transcriptomics

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Spatial transcriptomics offers in situ gene expression but faces trade-offs in resolution, depth, and sample size.
• Current methods integrating image segmentation are limited by imaging quality and tissue heterogeneity.
• Array-based technologies provide whole-transcriptome data at subcellular resolution, but lack cell annotation tools.

Purpose of the Study:

• To develop a novel multiscale computational framework for automated cell type classification using spatial transcriptomics.
• To leverage both transcriptomic and spatial context for subcellular-level cell identification.
• To create a generalizable pipeline applicable to diverse spatial transcriptomics platforms.

Main Methods:

• A multiscale approach combining transcriptomic data and spatial context for cell classification.
• Application to targeted and whole-transcriptome spatial platforms.
• Integration with multiparameter persistent homology for topological analysis of cell spatial relationships.

Main Results:

• Improved cell classification and morphology analysis in human kidney tissue.
• Accurate pinpointing of sparsely distributed mouse renal immune cells without image data.
• Identification of cell spatial relationships in a mouse model of lupus nephritis, validated experimentally.

Conclusions:

• The proposed framework enables automated, high-resolution cell type identification from spatial transcriptomics data.
• It overcomes limitations of image-based methods and generalizes across different technologies.
• This pipeline bridges gene expression and tissue-level organization for biological discovery.