cellink.tl.external.compute_celltype_programs

cellink.tl.external.compute_celltype_programs(adata, celltype_col, *, min_cells_per_type=10, use_raw=False, method='wilcoxon', prefix='celltype', out_dir=None, save=True)

Compute cell-type gene programs by differential expression (one vs rest).

Each gene receives a probabilistic score in [0, 1] derived from the Wilcoxon rank-sum test p-value, following the sc-linker paper:

X = -2 * log(p_adj) [chi2_2 distributed] score = (X - min(X)) / (max(X) - min(X))

Parameters:

• adata (AnnData) – AnnData with log-normalised expression in adata.X.

• celltype_col (str) – Column in adata.obs with cell-type labels.

• min_cells_per_type (int (default: 10)) – Cell types with fewer cells are skipped.

• use_raw (bool (default: False)) – Passed to sc.tl.rank_genes_groups.

• method (str (default: 'wilcoxon')) – DE method; "wilcoxon" is recommended.

• prefix (str (default: 'celltype')) – Prefix for the DE key stored in adata.uns.

• out_dir (str | Path | None (default: None)) – If provided and save=True, saves CSV files here.

• save (bool (default: True)) – Whether to write output CSV files.

Return type:

dict[str, DataFrame]

Returns:

dict ``{“pval”: DataFrame, “logfold”: DataFrame, “score”: DataFrame,

”genescores”: DataFrame}``

All DataFrames are (genes x cell_types).