univi.datasets.load

Contents

univi.datasets.load#

univi.datasets.load(name, keys=None, *, data_dir=None, counts_layer='counts', force=False, progress=True, **reader_kwargs)[source]#

Download (if needed) and load a registered dataset as AnnData objects.

Parameters:

  • name – Registry name, e.g. "pbmc_multiome_10k". See list_datasets().

  • keys – Subset of files to load (e.g. ["rna"]). Defaults to all.

  • data_dir – Cache directory (overrides UNIVI_DATA_DIR).

  • counts_layer – If this layer is missing and .X holds non-negative integers, copy .X into .layers[counts_layer] so the UniVI preprocessors can find raw counts. Set to None to leave objects untouched.

  • force – Re-download even if a verified cached copy exists.

  • reader_kwargs – Passed to the dataset-specific reader for archive datasets (e.g. scNMT-seq).

Returns:

{key: AnnData}. Keys listed under the entry’s paired field share identical, identically ordered obs_names.

Return type:

dict

Contents