omicverse.utils.LDA_topic

omicverse.utils.LDA_topic(adata, feature_type='expression', highly_variable_key='highly_variable_features', layers='counts', batch_key=None, learning_rate=0.001, ondisk=False)[source]

Latent Dirichlet Allocation (LDA) topic modeling for single-cell data using MIRA.

Parameters:

  • adata (anndata.AnnData) – Single-cell data object.

  • feature_type (str) – Feature modality passed to MIRA topic model.

  • highly_variable_key (str) – Var column indicating highly variable features.

  • layers (str) – Layer key containing count matrix.

  • batch_key (str or None) – Optional obs covariate for batch-aware modeling.

  • learning_rate (float) – Learning rate for model optimization.

  • ondisk (bool) – Whether to use on-disk dataset mode for large datasets.

Returns:

  • LDA_topic – Topic-model wrapper object.

  • Examples – >>> import omicverse as ov >>> # Basic LDA topic modeling >>> LDA_obj = ov.utils.LDA_topic(adata, feature_type=’expression’, … layers=’counts’) >>> # Determine optimal number of topics >>> LDA_obj.plot_topic_contributions(6) >>> # Fit model and predict topics >>> LDA_obj.predicted(13) >>> # Advanced classification with Random Forest >>> LDA_obj.get_results_rfc(adata, use_rep=’X_pca’, LDA_threshold=0.4)