omicverse.space.spatial_autocorr

omicverse.space.spatial_autocorr(adata, connectivity_key='spatial_connectivities', genes=None, mode='moran', transformation=True, n_perms=None, two_tailed=False, corr_method='fdr_bh', layer=None, seed=None, copy=False, n_jobs=1)[source]

Compute spatial autocorrelation statistics for gene expression.

Moran’s I (mode='moran') measures positive spatial autocorrelation on (-1, 1]; Geary’s C (mode='geary') measures the opposite – values near 0 indicate strong clustering. P-values are computed analytically under the normal approximation and optionally via label permutation.

Parameters:

  • adata – AnnData with a spatial connectivity matrix in adata.obsp.

  • connectivity_key (str (default: 'spatial_connectivities')) – Key of the spatial connectivity matrix in adata.obsp. Default: ‘spatial_connectivities’.

  • genes (default: None) – Gene names or indices to test. None tests all genes. Default: None.

  • mode (str (default: 'moran')) – 'moran' (Moran’s I) or 'geary' (Geary’s C). Default: ‘moran’.

  • transformation (bool (default: True)) – Row-normalise the connectivity matrix before scoring. Default: True.

  • n_perms (default: None) – Number of label-permutation iterations for empirical p-values. None uses only the analytical p-value. Default: None.

  • two_tailed (bool (default: False)) – Use two-tailed test for the normal-approximation z-score. Default: False.

  • corr_method (default: 'fdr_bh') – Multiple-testing correction method passed to statsmodels.stats.multitest.multipletests (e.g. 'fdr_bh'). Default: ‘fdr_bh’.

  • layer (default: None) – Expression layer to use. None uses adata.X. Default: None.

  • seed (default: None) – Random seed for permutation testing. Default: None.

  • copy (bool (default: False)) – Return the result DataFrame instead of (also) storing it in adata.uns. Default: False.

  • n_jobs (int (default: 1)) – Reserved for future parallel permutation support. Default: 1.

Returns:

Results with columns I / C, pval_norm, and optionally pval_sim, pval_z_sim, pval_adj. Also stored in adata.uns['moranI'] or adata.uns['gearyC'].

Return type:

DataFrame

Examples

>>> import omicverse as ov
>>> ov.space.spatial_neighbors(adata, n_neighs=6)
>>> df = ov.space.spatial_autocorr(adata, mode='moran')
>>> df.head()