omicverse-single-cell-rna-velocity

Installation
SKILL.md

OmicVerse Single-Cell RNA Velocity

Goal

Run reusable RNA velocity analysis on a velocity-ready AnnData: map spliced/unspliced layers if needed, fit moments and dynamics, estimate velocity, build the velocity graph, compute UMAP/Leiden, and render the stream plot. Use this same skill for method-comparison notebooks that branch across scvelo, dynamo, latentvelo, and graphvelo. If you still need FASTQ -> h5ad generation, hand off to the kb alignment skill first.

Quick Workflow

  1. Start from a velocity-ready AnnData.
  2. Map source-specific layers such as mature/nascent to spliced/unspliced when needed.
  3. Pick the branch trio up front: recipe, backend, and method.
  4. Instantiate ov.single.Velo(adata) and run filter_genes, preprocess, moments, dynamics, and cal_velocity.
  5. Build the velocity graph, neighbors, UMAP, Leiden, and velocity embedding.
  6. If the notebook includes a GraphVelo refinement stage, run it after an initial velocity layer exists.
  7. Plot the embedding and stream plot.
  8. Validate the stored keys before treating the result as finished.

Interface Summary

Installs
2
GitHub Stars
13
First Seen
Jun 29, 2026
omicverse-single-cell-rna-velocity — omicverse/omicverse-skills