Fig. 4
From: DeepVelo: deep learning extends RNA velocity to multi-lineage systems with cell-specific kinetics
Velocity, trajectory, and driver gene estimation of developing mouse hindbrain cells. a The putative developmental order for six cell-types in early mouse hindbrain development. b The velocity projected onto the t-distributed stochastic neighbor embedding (tsne) plot of gene expression. DeepVelo’s RNA velocity reveals the temporal order in the developing mouse hindbrain, including cells from early progenitors, GABAergic, and gliogenic lineages. c Velocity-based PAGA trajectory inference using DeepVelo’s velocity estimates. The predicted trajectory correctly reflects the developmental relations shown in a. d The top 60 driver genes with highest correlation to the GABAergic lineage computed using DeepVelo’s velocity estimation. The horizontal coordinates represent the pseudotime estimates. e Gene phase portrait, velocity, and gene expression plots of selected driver genes. Known functional genes in the GABAergic lineage - Tfap2b, Tfap2a, Lhx5, and Neurod6 - are computed among the top driver genes. f The velocity plot and trjectory inference using the scVelo dynamical model