omicverse.utils.roe

Contents

omicverse.utils.roe#

omicverse.utils.roe(adata, sample_key, cell_type_key, pval_threshold=0.05, expected_value_threshold=5, order=None)[source]#

Compute the Ro/e (observed/expected) cell-type enrichment matrix.

Parameters:

  • adata (AnnData) – Object with sample and cell-type annotations in .obs.

  • sample_key (str) – adata.obs column with the sample / tissue / condition labels (these become the Ro/e columns).

  • cell_type_key (str) – adata.obs column with the cell-type / cluster labels (these become the Ro/e rows).

  • pval_threshold (float, default=0.05) – Significance level for the global χ² test of cluster–sample independence. Recorded as a QC flag; it does not gate the returned matrix.

  • expected_value_threshold (float, default=5) – Minimum expected count below which the χ² approximation is considered unreliable. For a 2×2 table the global p-value then falls back to Fisher’s exact test; for larger tables a warning is emitted.

  • order (list of str or None, default=None) – Explicit column (sample) order. A comma-separated string is also accepted for backwards compatibility. None keeps the natural order. (The legacy sentinel 'F' is still accepted = None.)

Returns:

Ro/e matrix, cell types (rows) × samples (columns). Also stored, together with the observed / expected tables and the χ² result, in adata.uns['roe']; adata.uns['roe_results'] and adata.uns['expected_values'] are kept for backwards compatibility.

Return type:

pandas.DataFrame

Contents