grassp.tl.merge_small_clusters

merge_small_clusters(adata, cluster_key='leiden', min_n=3, key_added=None, verbose=True)

Iteratively merge small clusters into their most-connected neighbor.

Each round, clusters with fewer than min_n members are merged into the neighbor with the highest PAGA connectivity. Ties are broken by choosing the smallest candidate. Rounds repeat until no small clusters remain.

Parameters:

adata AnnData

AnnData object with a precomputed KNN graph in adata.obsp.

cluster_key str (default: 'leiden')

adata.obs column with cluster labels.

min_n int (default: 3)

Minimum cluster size. Clusters smaller than this are merged.

key_added str | None (default: None)

adata.obs column to write merged labels to. Defaults to f"{cluster_key}_merged".

verbose bool (default: True)

Print per-round merge decisions.

Return type:

None