snapatac2.tl.leiden#

snapatac2.tl.leiden(adata, resolution=1, objective_function='modularity', min_cluster_size=5, n_iterations=-1, random_state=0, key_added='leiden', weighted=False, inplace=True)[source]#

Cluster cells with the Leiden community detection algorithm.

Use this function after building a nearest-neighbor graph with snapatac2.pp.knn, or pass a sparse adjacency matrix directly.

Anti-Patterns#

Do NOT pass raw embeddings as adata; pass an AnnData object with adata.obsp["distances"] or a sparse graph matrix.
Do NOT expect clusters smaller than min_cluster_size to keep their original labels; they are relabeled as "-1".

param adata:: Annotated data object containing adata.obsp["distances"], or a sparse graph adjacency matrix.
type adata:: AnnData | AnnDataSet | spmatrix
param resolution:: Clustering resolution. Larger values usually produce more clusters.
type resolution:: float
param objective_function:: Leiden objective function.
type objective_function:: Literal['CPM', 'modularity']
param min_cluster_size:: Minimum retained cluster size. Smaller clusters are labeled "-1".
type min_cluster_size:: int
param n_iterations:: Number of Leiden optimization iterations. Use -1 to run until convergence.
type n_iterations:: int
param random_state:: Seed for Leiden initialization.
type random_state:: int
param key_added:: Key in adata.obs used to store cluster labels.
type key_added:: str
param weighted:: If True, transform graph distances into edge weights before clustering.
type weighted:: bool
param inplace:: If True, store labels in adata.obs[key_added]; if False, return them.
type inplace:: bool
returns:: If inplace=True, stores categorical labels in adata.obs[key_added] and returns None. If inplace=False, returns a string array of labels.
rtype:: ndarray | None

Examples

>>> import snapatac2 as snap
>>> adata = snap.datasets.pbmc5k(type="annotated_h5ad")
>>> snap.pp.knn(adata, use_rep="X_spectral")
>>> snap.tl.leiden(adata, resolution=1.0)
>>> "leiden" in adata.obs
True