omicverse.micro.combine_studies

omicverse.micro.combine_studies#

omicverse.micro.combine_studies(studies, study_names=None, rank='genus', study_key='study', min_prevalence=0.0)[source]#

Stitch a list of per-study AnnDatas into a single cross-cohort table.

Parameters:

  • studies (Sequence[AnnData]) – List of per-study AnnData objects (samples × ASVs or pre-collapsed genera). Each must already carry taxonomy columns in var if a rank other than None is requested.

  • study_names (Optional[Sequence[str]] (default: None)) – Optional list aligned with studies to label each cohort. Default: ['study_0', 'study_1', …].

  • rank (Optional[str] (default: 'genus')) – Collapse each study to this taxonomic rank before concatenating (so feature labels align across studies). Pass None to skip collapsing — only sensible when all studies already share the same ASV ids.

  • study_key (str (default: 'study')) – Column name to write the per-sample study label into.

  • min_prevalence (float (default: 0.0)) – Optional per-study prevalence filter applied before union. A taxon has to appear in >= this fraction of samples in at least one study to survive.

Returns:

Shape (Σn_samples, n_union_features). The obs carries the original per-study metadata (inner join on columns shared by all studies) plus obs[study_key]. var carries the union of feature names (no taxonomy column — the rank collapse already flattened that). X is a sparse CSR of int64 counts; features absent from a given study are zero.

Return type:

AnnData

Contents