omicverse.pp.highly_variable_features

omicverse.pp.highly_variable_features(data, batch=None, flavor='pegasus', n_top=2000, span=0.02, min_disp=0.5, max_disp=inf, min_mean=0.0125, max_mean=7, n_jobs=-1)[source]

Select highly variable features (HVF/HVG) for downstream modeling.

Parameters:

  • data (anndata.AnnData) – AnnData matrix with genes in columns.

  • batch (str, optional) – Column name in data.obs for batch-aware HVF selection.

  • flavor ({"pegasus", "Seurat"}, default="pegasus") – Algorithm used to rank variable genes.

  • n_top (int, default=2000) – Number of top genes to keep.

  • span (float, default=0.02) – Loess span used by Pegasus flavor.

  • min_disp (float, default=0.5) – Lower bound of normalized dispersion for Seurat flavor.

  • max_disp (float, default=np.inf) – Upper bound of normalized dispersion for Seurat flavor.

  • min_mean (float, default=0.0125) – Lower bound of mean expression for Seurat flavor.

  • max_mean (float, default=7) – Upper bound of mean expression for Seurat flavor.

  • n_jobs (int, default=-1) – Number of worker threads for computations where applicable.

Returns:

Updates data.var['highly_variable_features'] and ranking fields.

Return type:

None