scVelo for RNA velocity, the velocity half of the cell-cycle replication, run downstream of the learned geometry.

scvelo

scVelo estimates RNA velocity. Its role in the cell-cycle replication is specific and worth stating precisely: the paper computes neighbours and moments on the TopoMetry spectral scaffold rather than on PCA, then overlays the velocity field on that learned geometry. So velocity estimation stays downstream of geometry learning, and swapping the two would not reproduce the analysis.

conda-forge, locked green. Bioconda’s scvelo is stuck at 0.2.5, so there is no automatic BioContainer and the fixture grades L3 rather than L4.

Data caveat. scvelo.datasets.pancreas() downloads from the mutable master branch of theislab/scvelo_notebooks, a repository with no license file. The data can be fetched and analysed locally, but it is not redistributable and must not be baked into a container or a test fixture.