classify module

classify.get_model_path(ref_model, ref_adata)[source]

Fetch the model and reference data paths directly from user input.

classify.parse_arguments()[source]

Parse command-line arguments.

classify.validate_files(adata_file, RNApath, metapath, umappath, ADTpath, protein)[source]

Check if input files exist.

classify.load_proteins_from_file(file_path)[source]

Load protein names from a file.

classify.validate_data_integrity(RNA_counts, ADT_counts)[source]
classify.load_data(adata_file, ref_adata, RNApath, metapath, umappath, ADTpath, protein)[source]

Load AnnData objects and external data.

classify.preprocess_data(adata, protein, protein_adata, ref, meta, batch, proteins_to_check, protein_suffix, proteintech, output_dir, patient, mouse)[source]
classify.integrate_protein_data(adata, protein_adata, meta, proteins_to_check, protein_suffix, proteintech)[source]

Integrate protein expression data into the AnnData object.

Supports two protein naming formats: - Standard (default): CD16-TotalSeqC → CD16ADT (use –protein_suffix) - ProteinTech: prot:CD16.65090.1 → CD16ADT (use –proteintech flag)

classify.initialize_protein_data(adata, ref)[source]

Initialize protein expression data with zeros.

classify.prepare_adata_for_totalvi(adata, batch, ref, output_dir, patient, mouse)[source]

Prepare AnnData for TOTALVI.

classify.align_protein_data(adata, ref, output_dir, patient)[source]

Align protein expression data to the reference.

classify.train_totalvi_model(adata, ref_model, ref, adversarial_classifier)[source]

Train the TOTALVI model on the query data.

classify.classify_latent_space(vae_q, adata, classifier_type)[source]

Classify using BBC or BRF based on the selected model.

classify.save_results(adata, predictions, probs, output_dir, patient, vae_q, classifier_type, mouse)[source]

Save all relevant results based on the classifier used.

classify.classify_cells(adata, classifier_type, output_dir, patient)[source]

Classifies cells based on {classifier_type} probabilities stored in adata.obs. Assigns ‘ML_transition’, ‘unclassified’, or the label with the highest probability.

classify.main()[source]

Main function to execute the process.