decima.tools package

Submodules

decima.tools.evaluate module

decima.tools.evaluate.compare_marker_zscores(ad, key='cell_type')[source]
decima.tools.evaluate.compute_marker_metrics(marker_df, key='cell_type', tp_cutoff=1)[source]
decima.tools.evaluate.marker_zscores(ad, key='cell_type', layer=None)[source]
decima.tools.evaluate.match_criteria(df, filter_df)[source]

decima.tools.inference module

decima.tools.inference.predict_gene_expression(genes=None, model='ensemble', metadata_anndata=None, device=None, batch_size=8, num_workers=4, max_seq_shift=0, genome='hg38', save_replicates=False, float_precision='32')[source]

Predict gene expression for a list of genes

Parameters:

  • genes (list, optional) – List of genes to predict. Defaults to None.

  • model (str, optional) – Model to use for prediction. Defaults to ‘ensemble’.

  • metadata_anndata (str, optional) – Path to the metadata anndata file. Defaults to None.

  • device (str, optional) – Device to use for prediction. Defaults to None.

  • batch_size (int, optional) – Batch size for prediction. Defaults to 8.

  • num_workers (int, optional) – Number of workers for prediction. Defaults to 4.

  • max_seq_shift (int, optional) – Maximum sequence shift for prediction. Defaults to 0.

  • genome (str, optional) – Genome build for prediction. Defaults to ‘hg38’.

  • save_replicates (bool, optional) – Save the replicates for prediction. Defaults to False.

Raises:

ValueError – If the model is not ‘ensemble’ and save_replicates is True.

Returns:

AnnData object with the predicted gene expression.

Return type:

anndata.AnnData

decima.tools.interpret module

decima.tools.interpret.find_attr_peaks(attr, tss_pos=None, n=5, min_dist=6)[source]
decima.tools.interpret.scan_attributions(seq, attr, motifs, peaks, names=None, pthresh=0.001, rc=True, window=18)[source]

Module contents