Fig. 4: Analysis of mouse embryonic reprogramming data (n = 85, 010 cells).

a Streamline plots from SDEvelo and scVelo (dyn) on projected PCA space, with each cell colored according to reprogramming days, ranging from 0 to 28. b Heatmap of the estimated latent time by SDEvelo on projected PCA space, showing consistency with reprogramming days. c Bar plots comparing SDEvelo with other methods (n = 5 cross-validation replicates). Upper panel: Pearson correlation coefficients between estimated latent time and reprogramming days. Bar plots show mean ± standard deviation (SD) with individual points. Lower panel: mean and standard deviations of AUC values quantifying the correctness of inferred transitions. Source data are provided as a Source Data file. d Comparative alignment heatmap of time estimates scaled from 0 to 1: latent time estimated by SDEvelo, true reprogramming days, Monocle’s estimated pseudotime, and latent time estimated by scVelo (dyn). Pearson correlation coefficients are calculated between each method’s time estimate and the true reprogramming days. e Middle left panel: overall streamline plot from SDEvelo in PCA space, visualized using a subset of 5000 cells for improved clarity. Top left panel: subset streamline plot illustrating SDEvelo’s identification of a dead-end trajectory spanning from Cluster 8 to 4 to 3, marked by the expression of the MEF marker gene Col1a2. The heatmap of estimated latent times on the PCA manifold illustrates the temporal progression within identified trajectories. Bottom left panel: subset streamline plot focusing on clusters 1, 2, and 6 within the PCA space and showcasing SDEvelo’s tracing of a reprogramming trajectory from Cluster 2 to 6 to 1, characterized by the expression of the iEP marker gene Apoa1. The corresponding heatmap of estimated latent time provides a visualization of temporal dynamics of the PCA manifold for the specified cluster subset. Right panel: After integrating the latent time and velocity estimated by SDEvelo with gene expression analyzed by CellRank, clusters 3 and 1 were identified as terminal states. The gene expression heatmap for top genes associated with the dead-end and reprogramming trajectories. The color scheme for latent time is the same as that used in (b).