spatial-statistics

spatial-statistics

When to use

Measure spatial enrichment, co-occurrence, autocorrelation or graph structure. Use spatial-genes for SVG ranking and spatial-domains for regions.

Use from a step

from skills._sdk.notebook import load_skill, read_input, write_output
library = load_skill("spatial-statistics")
data = read_input("input.h5ad")
data = library.analyze(data, analysis_type="neighborhood_enrichment", random_state=123)
write_output(library.results_table(data), "tables/results.csv")

Run examples/example_step.py through the step runner.

API

analyze(adata, *, analysis_type: str='neighborhood_enrichment', random_state: int=123, **parameters)

Run one spatial analysis in place and return the same AnnData.

Gene analyses read X (log-normalized expression); cluster and network analyses read labels, coordinates and spatial_connectivities. Existing spatial graphs are reused unless force_graph_rebuild=True. Bivariate Moran's backend cannot be seeded: results vary between runs.

:param adata: Spatial AnnData; cluster-aware methods require categorical labels. :param analysis_type: CLI default neighborhood_enrichment; also ripley, co_occurrence, moran, geary, local_moran, getis_ord, bivariate_moran, network_properties or spatial_centrality. :param random_state: Permutation/simulation seed, matching CLI default 123. Ignored by bivariate_moran; its run_info records this limitation. :param parameters: Method keyword options from references/parameters.md; omitted values retain backend wrapper defaults. Unknown options raise. :returns: The same AnnData with JSON-encoded diagnostics and result tables. :raises ValueError: Unknown analysis or invalid input. :raises TypeError: An option does not belong to the chosen method. :raises ImportError: Missing backend; use install_skill_deps with squidpy, esda, libpysal or networkx as reported in the error.