This paper describes a new computational model called Topological Velocity Inference (TopoVelo), designed to jointly infer the spatial and temporal dynamics of cell fate transitions using spatial transcriptomic data. TopoVelo significantly advances existing RNA velocity frameworks by integrating spatial coupling among cells, modeling the entire tissue's dynamics through spatially coupled differential equations and graph neural networks (GNNs). The researchers apply TopoVelo to various biological systems, including the developing mouse cerebral cortex and neural tube closure, demonstrating that it accurately estimates cell velocity (the rate and direction of cell differentiation or migration) and identifies influential cell niches correlated with ligand-receptor signaling. Furthermore, the model is shown to be robust and provides biologically meaningful insights into complex developmental processes both in vivo and in in vitro human embryoid body models.

References:

• Gu Y, Liu J, Lee K H, et al. Topological velocity inference from spatial transcriptomic data[J]. Nature Biotechnology, 2025: 1-12.