sc_reconstruction.metrics.metric_pathway#

sc_reconstruction.metrics.metric_pathway(adata_true, adata_pred, *, progeny_model=None, pathway_dict=None, correlation_measure='spearman', overlap_threshold=5, min_cells=30)[source]#

Pathway activity correlation (PROGENy ULM scores).

PROGENy pathway weights come from [Schubert et al., 2018]; the univariate-linear-model scoring is provided by decoupler [Badia-i-Mompel et al., 2022]. Wraps PathwayCalculator.pathway_score_similarity().

Parameters:

progeny_model (DataFrame | None) – PROGENy weight DataFrame (from decoupler.get_progeny(), [Badia-i-Mompel et al., 2022, Schubert et al., 2018]). If None, it is fetched lazily.
pathway_dict (Mapping[str, Sequence[str]] | None) – Optional explicit pathway → gene list mapping. Falls back to PROGENy coverage.
adata_true (anndata.AnnData)
adata_pred (anndata.AnnData)
correlation_measure (str)
overlap_threshold (int)
min_cells (int)

Return type:

float