sc_reconstruction.metrics.compute_all_metrics#

sc_reconstruction.metrics.compute_all_metrics(adata_true, adata_pred, *, s_genes=None, g2m_genes=None, geneset_dict=None, progeny_model=None, pathway_dict=None, cytokine_dict=None, deg_refs=None, min_cells=20, perturb=None, knn_k=20, use_rep=None)[source]#

Run every metric for which the required inputs are available.

Returns one flat {metric_name: score} dict. Metrics whose required inputs are missing are returned as NaN.

Parameters:

  • perturb (tuple[anndata.AnnData, anndata.AnnData] | None) – (adata_pert_true, adata_ctrl) if KNN purity should be computed against the predicted perturbation in adata_pred.

  • min_cells (int) – Minimum cells-per-gene cutoff forwarded to the biological metrics. Lower this on small test sets (e.g. 100-cell tutorial slices) where the default of 20 would filter most cell-cycle / signature genes.

  • adata_true (anndata.AnnData)

  • adata_pred (anndata.AnnData)

  • s_genes (Sequence[str] | None)

  • g2m_genes (Sequence[str] | None)

  • geneset_dict (Mapping[str, Sequence[str]] | None)

  • progeny_model (DataFrame | None)

  • pathway_dict (Mapping[str, Sequence[str]] | None)

  • cytokine_dict (Mapping[str, Sequence[str]] | None)

  • deg_refs (tuple[anndata.AnnData, anndata.AnnData] | None)

  • knn_k (int)

  • use_rep (str | None)

Return type:

dict[str, float]

On this page