omicverse.micro.Ordinate.nmds

Contents

omicverse.micro.Ordinate.nmds#

Ordinate.nmds(n=2, random_state=0, write_to_obsm=True)[source]#

Non-metric multi-dimensional scaling on the distance matrix.

Wraps sklearn.manifold.MDS(dissimilarity=’precomputed’). NMDS preserves rank order rather than absolute distances — typically less distorted on Bray-Curtis / Jaccard than linear PCoA.

Parameters:

  • n (int, default 2) – Output dimensions.

  • random_state (int, default 0) – Seeds the four-restart NMDS init for reproducibility.

  • write_to_obsm (bool, default True) – Persist coords into adata.obsm[f'{dist_key}_nmds'] and the final stress value into adata.uns['micro'].

Return type:

pd.DataFrame indexed by sample with columns NMDS1..NMDSn.

Contents