Φ-Space ST: A platform-agnostic method to identify cell states in spatial transcriptomics studies

Area of Science:

• Spatial transcriptomics
• Computational biology
• Genomics

Background:

• Spatial transcriptomics (ST) enables mapping gene expression within intact tissues.
• Analyzing ST data often requires integrating multiple single-cell RNA sequencing (scRNA-seq) references for accurate cell state annotation.
• Existing methods face challenges in scalability, computational efficiency, and harmonizing diverse reference datasets.

Purpose of the Study:

• To introduce Φ-Space ST, a novel platform-agnostic computational method for analyzing spatial transcriptomics data.
• To enable accurate cell-type deconvolution and cell state annotation using multiple scRNA-seq references.
• To facilitate the identification of spatial niches and disease-specific cell states in complex tissues.

Main Methods:

• Φ-Space ST utilizes multiple scRNA-seq references to identify continuous cell states in ST data.
• The method is platform-agnostic, applicable to various ST technologies (CosMx, Visium, Xenium, Stereo-seq).
• It achieves interpretable cell-type deconvolution for supercellular resolution and annotates cell states without segmentation for subcellular resolution.

Main Results:

• Φ-Space ST provides significantly faster computation for supercellular ST data analysis.
• For subcellular resolution, it enables insightful spatial niche identification without requiring cell segmentation.
• The method successfully harmonizes annotations from multiple scRNA-seq references and characterizes disease cell states using healthy references.
• Case studies on various cancer tissues revealed niche-specific cell types and distinct co-presence patterns differentiating tumor from non-tumor regions.

Conclusions:

• Φ-Space ST is a robust and scalable tool for spatial transcriptomics data analysis.
• It enhances the understanding of complex tissue architectures and pathological processes.
• The method offers interpretable deconvolution and spatial niche identification across diverse ST platforms.