Spatial transition tensor of single cells

Area of Science:

• Computational Biology
• Genomics
• Developmental Biology

Background:

• Spatial transcriptomics and mRNA splicing provide spatiotemporal information on cell states and transitions.
• Existing lineage-inference methods often lack spatial dynamics or cannot capture multiple cell state transition paths.

Purpose of the Study:

• To develop a novel method, spatial transition tensor (STT), for characterizing multistability in space using spatial transcriptomes and mRNA splicing.
• To reconstruct cell-state-specific dynamics and spatial state transitions across multiple spatiotemporal scales.

Main Methods:

• Utilized a multiscale dynamical model incorporating mRNA splicing and spatial transcriptomes.
• Employed a four-dimensional transition tensor and spatial-constrained random walk for analysis.
• Reconstructed cell-state-specific dynamics and spatial transitions through short-time local tensor streamlines and long-time transition paths.

Main Results:

• Benchmarking and applications on diverse transcriptome datasets demonstrated STT's capability.
• Successfully recovered cell-state-specific dynamics and associated genes missed by existing methods.
• Validated STT on epithelial-mesenchymal transitions, blood development, mouse brain, and chicken heart development.

Conclusions:

• STT offers a robust method for analyzing spatiotemporal dynamics in single-cell transcriptomics.
• Provides a consistent multiscale description of single-cell transcriptome data.
• Enhances understanding of cell state transitions and their underlying gene regulation across biological systems.