BAITS.st.gr.spatial_neighbors

BAITS.st.gr.spatial_neighbors#

BAITS.st.gr.spatial_neighbors(adata, sample_key, coord_type, n_neighs=10, radius=None, delaunay=True, spatial_key='spatial', n_rings=1, percentile=None, set_diag=False, key_added='spatial', copy=False)#

Compute spatial neighbor graph for an AnnData object.

This function constructs a spatial neighborhood graph based on spatial coordinates stored in adata.obsm[spatial_key]. The resulting adjacency and distance matrices are stored in adata.obsp and metadata describing the graph is stored in adata.uns.

The neighborhood graph can be computed using k-nearest neighbors, a fixed radius, or Delaunay triangulation depending on the parameters.

Parameters:

  • adata (anndata.AnnData) – Annotated data matrix containing spatial coordinates.

  • sample_key (str or None) – Column in adata.obs specifying sample or library identifiers. If provided, spatial neighbor graphs will be computed independently for each sample and then combined into a block-diagonal matrix.

  • coord_type ({"grid", "generic"} or None) –

    Type of spatial coordinate system.

    • "grid": coordinates follow a lattice/grid structure.

    • "generic": arbitrary spatial coordinates.

    If None, the coordinate type is inferred automatically.

  • n_neighs (int, default=10) – Number of nearest neighbors used to construct the spatial graph.

  • radius (float or None, optional) – Radius for defining spatial neighbors. Only applicable when coord_type="generic".

  • delaunay (bool, default=True) – Whether to construct neighbors using Delaunay triangulation.

  • spatial_key (str, default="spatial") – Key in adata.obsm where spatial coordinates are stored.

  • n_rings (int, default=1) – Number of neighbor rings used when coord_type="grid".

  • percentile (float or None, optional) – Percentile threshold used for distance filtering.

  • set_diag (bool, default=False) – Whether to set diagonal elements of the connectivity matrix.

  • key_added (str, default="spatial") – Prefix used for storing results in adata.obsp and adata.uns.

  • copy (bool, default=False) – If True, return adjacency and distance matrices instead of storing them in adata.

Returns:

  • tuple of scipy.sparse.spmatrix, optional – If copy=True, returns:

    • adjacency matrix

    • distance matrix

  • Otherwise, results are stored in

    • adata.obsp[f"{key_added}_connectivities"]

    • adata.obsp[f"{key_added}_distances"]

    • adata.uns[f"{key_added}_neighbors"]

Notes

The neighbor graph is computed separately for each sample when sample_key is provided. The resulting matrices are merged using block-diagonal concatenation to avoid cross-sample connections.

Examples

>>> spatial_neighbors(adata, sample_key="sample", coord_type="generic")
>>> adata.obsp["spatial_connectivities"]
>>> adata.obsp["spatial_distances"]
Contents