scparadise.scadam.available_models

scparadise.scadam.available_models#

scparadise.scadam.available_models(species=None, tissue=None, suspension=None, query=None, min_accuracy=None, min_balanced_accuracy=None, n_levels=None, sort_by=None, ascending=False, compact=True, refresh=False)[source]#

Get a table of available pretrained scAdam models with optional filtering.

Parameters:

  • species (str (default: None)) – Filter models by species, using the Species column. (for example: ‘human’, ‘mouse’, ‘macaque’, ‘marmoset’).

  • tissue (str (default: None)) – Filter models by tissue or organ, using the Tissue/Organ column. (for example: ‘brain’, ‘adipose’, ‘blood’).

  • suspension (str (default: None)) – Filter models by assay input type from the Suspension column. (for example: ‘cells’ or ‘nuclei’).

  • query (str (default: None)) – Search substring across Model name, Description, Tissue/Organ, and Species.

  • min_accuracy (float (default: None)) – Keep only models with Accuracy greater than or equal to this value.

  • min_balanced_accuracy (float (default: None)) – Keep only models with Balanced Accuracy greater than or equal. to this value.

  • n_levels (int (default: None)) – Filter models by the number of annotation levels.

  • sort_by (str (default: None)) – Column name used for sorting.

  • ascending (bool (default: False)) – Sort order.

  • compact (bool (default: True)) – If True, return a compact subset of the most useful columns. If False, return the full table.

  • refresh (bool (default: False)) – If True, clear the cache and reload the table from the remote source.

Returns:

DataFrame with available pretrained scAdam models after applying all requested filters.

Return type:

pandas.DataFrame

Notes

The table is loaded from the official scParadise repository and contains metadata columns including Model name, Tissue/Organ, Species, Description, Suspension, Accuracy, Balanced Accuracy, and Number of Levels.

Examples

>>> available_models()
>>> available_models(species='human')
>>> available_models(tissue='adipose', suspension='cells')
>>> available_models(query='brain', min_accuracy=0.99)
>>> available_models(min_balanced_accuracy=0.95, n_levels=2)