cellflow.preprocessing.compute_wknn

cellflow.preprocessing.compute_wknn(ref_adata=None, query_adata=None, n_neighbors=30, ref_rep_key='X_pca', query_rep_key='X_pca', uns_key_added='wknn', query2ref=True, ref2query=False, weighting_scheme='jaccard_square', top_n=None, copy=False)

Compute the weighted k-nearest neighbors graph between the reference and query datasets

Parameters:

• ref_adata (AnnData) – An AnnData object with the reference representation to build ref-query neighbor graph

• query_adata (AnnData) – An AnnData object with the query representation to build ref-query neighbor graph

• n_neighbors (int) – Number of neighbors per cell

• ref_rep_key (str) – Key in obsm of ref_adata containing the reference representation

• query_rep_key (str) – Key in obsm of query_adata containing the query representation

• uns_key_added (str) – Key to store the weighted k-nearest neighbors graph in uns

• query2ref (bool) – Consider query-to-ref neighbors

• ref2query (bool) – Consider ref-to-query neighbors

• weighting_scheme (Optional[Literal['top_n', 'jaccard', 'jaccard_square']]) –

  How to weight edges in the ref-query neighbor graph. Options are:

  • None: No weighting

  • 'top_n': Binaries edges based on the top top_n neighbors.

  • 'jaccard': Weight edges based on the Jaccard index.

  • 'jaccard_square': Weight edges based on the square of the Jaccard index.

• top_n (int | None) – The number of top neighbors to consider

• copy (bool) – Return a copy of ref_adata instead of updating it in place

Return type:

AnnData | None

Returns:

: If copy is True, returns a new AnnData object with the weighted k-nearest neighbors stored in uns. Otherwise, updates adata in place.

Sets the following fields:

• .uns[uns_key_added]: Weighted k-nearest neighbors graph