chrysalis.aa

chrysalis.aa(adata: AnnData, n_archetypes: int = 8, pca_key: Optional[str] = None, n_pcs: Optional[int] = None, max_iter: int = 200)

Run archetypal analysis on the low-dimensional embedding.

Calculates archetypes, alphas, loadings, and RSS (Residual Sum of Squares). Requires input calculated with chrysalis.pca.

Parameters:

  • adata – The AnnData data matrix of shape n_obs × n_vars. Rows correspond to cells and columns to genes.

  • n_archetypes – Number of archetypes (tissue compartments) to be identified.

  • pca_key – Define alternative PCA input key from .obm, otherwise chr_X_pca is used.

  • n_pcs – Number of PCs (Principal Components) to be used.

  • max_iter – Maximum number of iterations.

Returns:

Updates .uns with the following fields:

  • .uns[‘chr_aa’][‘archetypes’] – Archetypes.

  • .uns[‘chr_aa’][‘alphas’] – Alphas.

  • .uns[‘chr_aa’][‘loadings’] – Gene loadings.

  • .uns[‘chr_aa’][‘RSS’] – RSS reconstruvtion error.

Example usage:

>>> import chrysalis as ch
>>> import scanpy as sc
>>> adata = sc.datasets.visium_sge(sample_id='V1_Human_Lymph_Node')
>>> sc.pp.calculate_qc_metrics(adata, inplace=True)
>>> sc.pp.filter_cells(adata, min_counts=6000)
>>> sc.pp.filter_genes(adata, min_cells=10)
>>> ch.detect_svgs(adata)
>>> sc.pp.normalize_total(adata, inplace=True)
>>> sc.pp.log1p(adata)
>>> ch.pca(adata)
>>> ch.aa(adata, n_pcs=20, n_archetypes=8)