MultiCell is a geometric deep learning method that uses a dual-graph data structure to predict multicellular development. It unifies granular and foam-like models to forecast single-cell behaviors, like junction loss and invagination, with high precision during embryogenesis.

References:

• Yang H, Roy G, Nguyen A Q, et al. Multicell: geometric learning in multicellular development[J]. Nature Methods, 2025: 1-9.