The dynamic evolution of cell fate is a fundamental scientific question in developmental biology, disease modeling, and regenerative medicine. While single-cell RNA-sequencing provides high-resolution snapshots of cellular states, the destructive nature of the technology means that longitudinal tracking of individual cells is often impossible, resulting in unpaired temporal data. This poses a significant hurdle for reconstructing developmental trajectories. Here, we introduce SCOPE, a novel generative model that learns the evolutionary dynamics from these unpaired snapshots by formulating it in a generalized version of Schrödinger Bridge framework. We validate SCOPE on multiple datasets to show that SCOPE can generate complete and high-resolution trajectories, interpolate missing time points, and predict shifts in cell fate through gene perturbation. Our results establish SCOPE as a powerful and flexible framework for generating continuous and reliable cellular trajectories from discrete scRNA-seq data, opening new avenues for dissecting the mechanisms of cell fate determination.

海报.pdf