omicverse.metabol.corr_network

Contents

omicverse.metabol.corr_network#

omicverse.metabol.corr_network(adata, *, group_col=None, group=None, features=None, method='spearman', abs_r_threshold=0.5, padj_threshold=0.05, layer=None)[source]#

Pairwise metabolite correlation network within a single condition.

Parameters:

  • group_col (Optional[str] (default: None)) – If both given, restrict to samples where adata.obs[group_col] == group. If group_col is given but group is None a ValueError is raised — partial filtering makes no sense. If both are None, all samples are used.

  • group (Optional[str] (default: None)) – If both given, restrict to samples where adata.obs[group_col] == group. If group_col is given but group is None a ValueError is raised — partial filtering makes no sense. If both are None, all samples are used.

  • features (Optional[Sequence[str]] (default: None)) – Optional subset of metabolite names (reduces O(p^2)).

  • method (str (default: 'spearman')) – "spearman" (default, heavy-tailed-friendly) or "pearson".

  • abs_r_threshold (float (default: 0.5)) – Keep pairs with |r| >= abs_r_threshold in the output edge list.

  • padj_threshold (float (default: 0.05)) – Keep pairs with BH-FDR padj < padj_threshold.

  • layer (Optional[str] (default: None)) – AnnData layer (default None → adata.X).

Returns:

Filtered edge list, columns feature_a, feature_b, r, pvalue, padj. Sorted by |r| descending. Intended as input to networkx.Graph (nx.from_pandas_edgelist(df, source='feature_a', target='feature_b', edge_attr='r')).

Return type:

pd.DataFrame

Contents