cycombinepy.evaluate.scib_metrics

cycombinepy.evaluate.scib_metrics(adata, batch_key, label_key=None, embedding_key='X_pca', layer=None)

Run a minimal scib-metrics benchmark on an AnnData.

Computes a PCA on adata.X (or adata.layers[layer] if supplied) and evaluates batch-mixing metrics from scib_metrics. This is a convenient drop-in for comparing before/after correction; call it twice and diff the resulting dicts.

Returns a dict of scalar scores. Metrics that require a biological label are skipped if label_key is None.

Return type:

dict

Parameters:

• adata (AnnData)

• batch_key (str)

• label_key (str | None)

• embedding_key (str)

• layer (str | None)