scatlastb_utils.metrics.compare_distances

scatlastb_utils.metrics.compare_distances#

scatlastb_utils.metrics.compare_distances(adata, obsp_key_1, obsm_key_1, obsp_key_2, obsm_key_2, scale_distances=True, quantile=0.9, k_max=50, log_scale_diffs=False, **kwargs)#

Compare distances of same edges but from different representations.

This function computes:

  1. “average_distance_1”: the average distances per k-nearest neighborhood for obsp_key_1

  2. “average_distance_2”: the average distances per k-nearest neighborhood for obsp_key_2

  3. “average_difference”: the difference of 1. and 2.

  4. “average_distance_diff”: the average of the differences between the two embeddings

  5. “spearman_correlation”: the Spearman correlation of the distance differences (of the k-nearest neighbors only)

Parameters:

  • adata – AnnData object containing the distances in obsp

  • obsp_key_1 – slot for pair-wise distances from embedding 1

  • obsm_key_1 – slot for embedding 1 used for distance computation

  • obsp_key_2 – slot for pair-wise distances from embedding 2

  • obsm_key_2 – slot for embedding 2 used for distance computation

  • scale_distances (default: True) – if True, scale distances by the quantile of the distances

  • quantile (default: 0.9) – quantile to scale distances by, default is 0.9

  • k_max (default: 50) – maximum number of neighbors to consider, default is 50

  • log_scale_diffs (default: False) – if True, log scale the differences

  • kwargs – additional keyword arguments for distance computation, e.g. metric=’euclidean’

Returns:

DataFrame with average distances and differences