tomodrgn.analysis.run_umap

run_umap(z: ndarray, random_state: int | RandomState | None = None, **kwargs: Any) → tuple[ndarray, UMAP]

Run UMAP dimensionality reduction on latent embeddings.

Parameters:

• z – array of latent embeddings, shape (nptcls, zdim)

• random_state – random state for reproducible runs, passed to umap.UMAP

• kwargs – additional key word arguments passed to umap.UMAP

Returns:

array of UMAP embeddings, shape (len(z), 2), and fit reducer object