omicverse.single.cosg

omicverse.single.cosg(adata, groupby='CellTypes', groups='all', mu=1, remove_lowly_expressed=False, expressed_pct=0.1, n_genes_user=50, key_added=None, calculate_logfoldchanges=True, use_raw=True, layer=None, reference='rest', copy=False)[source]

Identify cluster-specific marker genes with COSG.

Parameters:

  • adata (anndata.AnnData) – Annotated data matrix for marker detection.

  • groupby (str, default='CellTypes') – Obs column containing cluster/group labels.

  • groups ({'all'} or Iterable[str], default='all') – Group subset to analyze.

  • mu (float, default=1) – Specificity penalty against expression in non-target groups.

  • remove_lowly_expressed (bool, default=False) – Whether to remove genes with low detection rates in target groups.

  • expressed_pct (float, default=0.1) – Minimum fraction of expressing cells when low-expression filtering is enabled.

  • n_genes_user (int, default=50) – Number of top marker genes retained per group.

  • key_added (str or None, default=None) – Key in adata.uns used to store COSG results.

  • calculate_logfoldchanges (bool, default=True) – Whether to compute log-fold changes.

  • use_raw (bool, default=True) – Whether to use adata.raw expression values when available.

  • layer (str or None, default=None) – Layer key used as expression matrix.

  • reference (str, default='rest') – Reference strategy for differential comparison.

  • copy (bool, default=False) – If True, run on a copy and return it.

Returns:

Returns modified copy when copy=True; otherwise updates input object in place.

Return type:

anndata.AnnData or None