Virtual dynamics of spatiotemporal single-cell interactions

CytoBridge reconstructs continuous virtual cell dynamics from discrete spatial transcriptomic snapshots by learning stochastic trajectories of cell states, positions, and population sizes.

• Why it matters: Spatial omics captures only sparse time points, leaving the evolution of interacting cell populations in space and time difficult to infer for virtual tissue building.
• What they did: CytoBridge uses an unbalanced mean-field Schrodinger bridge to learn trajectories, with a time-varying interaction graph from spatial proximity and ligand-receptor priors that decomposes velocity into interaction and intrinsic-context components.
• The result: Across five datasets on three spatial transcriptomics platforms, it generates dynamics at unmeasured time points, tracks ligand-receptor signaling through measured and unmeasured stages, and enables hypothesis-generating in silico perturbations.

AI summary of the abstract · may contain errors · not medical advice · read the paper.