Fig. 1

The flowchart of RNA velocity implementation in scRNA-seq data. The graph reflects scVelo. a A k-NN graph constructed from a PCA of the spliced counts is used to smooth (impute) the spliced and unspliced count matrices, resulting in the Ms and Mu matrix. This is followed by gene-specific velocity estimation using either the dynamical or the steady-state model. b Visualization of the estimated velocities on a low-dimensional embedding using velocity transition probabilities. First, transition probabilities are computed by considering which neighbors have a difference between the expression of the neighbor and the expression of the cell in question most similar to the estimated velocities. These transition probabilities are used to compute a vector as a linear combination of existing displacements. Finally, the resulting vector field can be visualized using streamline plots or a gridding approach (abbreviations: PCs, principal components; Diff. matrix, difference matrix between one cell to other cells)