nichesphere.tl.PIC_BGdoubletsOEratios_parallel

nichesphere.tl.PIC_BGdoubletsOEratios_parallel(seed, adata, annot_col, target_pairs, downsample_cells=False, cell_list_downsample=None, downsampling_factor=6, pseudocount=1e-10)

Generate one set of O/E ratios for cell type pair co-localization probabilities from randomly paired single cells. Running this function n times (e.g. via get_PIC_BG_OEratios_DF) builds a per-pair empirical background distribution that can be used for significance testing in PIC-seq data or single-sample datasets.

Parameters:

• seed (int) – Random seed for reproducibility.

• adata (AnnData) – Singlets (scRNA-seq) anndata object.

• annot_col (str) – Name of the cell type annotation column in adata.obs.

• target_pairs (list or pd.Index) – Cell type pairs to evaluate (e.g. colocPerSample.columns).

• downsample_cells (bool, default False) – Whether to downsample over-represented cell types before pairing. Useful when the gated singlets data does not reflect true in-tissue proportions.

• cell_list_downsample (list or None, default None) – Cell types to downsample (used only when downsample_cells=True).

• downsampling_factor (int, default 6) – Reduction factor passed to downsample_class() (used only when downsample_cells=True).

• pseudocount (float, default 1e-10) – Small value added to observed and expected counts before division to avoid division-by-zero or log(0) in downstream steps.

Returns:

oe_ratios – O/E ratios for each pair in target_pairs, indexed by pair name.

Return type:

pd.Series