[!IMPORTANT] scvelo-modern is a minimal compatibility distribution of scVelo 0.3.4. It preserves the scvelo import package and scientific implementation while fixing the stochastic velocity regression caused by NumPy 2 scalar-assignment rules. Use the official scVelo project unless you need this compatibility fix. Do not install scvelo and scvelo-modern in the same environment because both distributions provide the same import package.

The compatibility patch is intentionally narrow: it converts the one-element least-squares result to a scalar before assignment. Regression tests cover all four fit_offset/fit_offset2 combinations on NumPy 1.26 and NumPy 2, compare against the unmodified scVelo 0.3.4 numerical baseline, and exercise the complete stochastic velocity path. The fork can be retired after upstream publishes an equivalent fix.

This distribution is based on upstream scVelo v0.3.4 at commit 9b6e946. The NumPy 2 regression is tracked upstream in #1341.

scVelo - RNA velocity generalized through dynamical modeling

scVelo is a scalable toolkit for RNA velocity analysis in single cells; RNA velocity enables the recovery of directed dynamic information by leveraging splicing kinetics ¹. scVelo collects different methods for inferring RNA velocity using an expectation-maximization framework ², deep generative modeling ³, or metabolically labeled transcripts⁴.

scVelo's key applications

• estimate RNA velocity to study cellular dynamics.
• identify putative driver genes and regimes of regulatory changes.
• infer a latent time to reconstruct the temporal sequence of transcriptomic events.
• estimate reaction rates of transcription, splicing and degradation.
• use statistical tests, e.g., to detect different kinetics regimes.

Citing scVelo

If you include or rely on scVelo when publishing research, please adhere to the following citation guide:

EM and steady-state model

If you use the EM (dynamical) or steady-state model, cite

@article{Bergen2020,
  title = {Generalizing RNA velocity to transient cell states through dynamical modeling},
  volume = {38},
  ISSN = {1546-1696},
  url = {http://dx.doi.org/10.1038/s41587-020-0591-3},
  DOI = {10.1038/s41587-020-0591-3},
  number = {12},
  journal = {Nature Biotechnology},
  publisher = {Springer Science and Business Media LLC},
  author = {Bergen, Volker and Lange, Marius and Peidli, Stefan and Wolf, F. Alexander and Theis, Fabian J.},
  year = {2020},
  month = aug,
  pages = {1408–1414}
}

RNA velocity inference through metabolic labeling information

If you use the implemented method for estimating RNA velocity from metabolic labeling information, cite

@article{Weiler2024,
  author = {Weiler, Philipp and Lange, Marius and Klein, Michal and Pe'er, Dana and Theis, Fabian},
  publisher = {Springer Science and Business Media LLC},
  url = {http://dx.doi.org/10.1038/s41592-024-02303-9},
  doi = {10.1038/s41592-024-02303-9},
  issn = {1548-7105},
  journal = {Nature Methods},
  month = jun,
  number = {7},
  pages = {1196--1205},
  title = {CellRank 2: unified fate mapping in multiview single-cell data},
  volume = {21},
  year = {2024},
}

Support

Found a bug or would like to see a feature implemented? Feel free to submit an issue. Have a question or would like to start a new discussion? Head over to GitHub discussions. Your help to improve scVelo is highly appreciated. For further information visit scvelo.org.