scdrs.method.downstream_group_analysis#

scdrs.method.downstream_group_analysis(adata: AnnData, df_full_score: DataFrame, group_cols: List[str], fdr_thresholds: List[float] = [0.05, 0.1, 0.2]) → Dict[str, DataFrame][source]#

scDRS group-level analysis.

For each annotation in group_cols and each group of cells in the annotation, compute:

  1. Proportion of FDR < 0.1 cells.

  2. Group-level trait association.

  3. Group-level heterogeneity.

connectivities is expected in adata.obsp for the group-level heterogeneity analysis. Recommended parameters: sc.pp.neighbors(adata, n_neighbors=15, n_pcs=20).

Parameters:
adataanndata.AnnData

Single-cell data of shape (n_cell, n_gene). Assumed to be size-factor-normalized and log1p-transformed.

df_full_scorepd.DataFrame

scDRS .full_score file for a given trait.

group_colslist of str

List of column names in adata.obs used to define cell groups.

Returns:
dict_df_resDict[str, pd.DataFrame]

Group-level statistics (n_group, n_stats) keyed by the group names.