HoloNet.predicting.get_gene_expr

HoloNet.predicting.get_gene_expr(adata, lr_df=None, complex_db=None, gene_list_on_interest=None, min_mean=0.05, min_disp=0.5, max_sparse=0.5, remove_lr_gene=True)

Filter out the genes with too low expression levels or too low dispersions. Filter out the genes expressed in too fewer cells (or spots). Filter out the mitochondria genes and ligand (or receptor) genes.

Parameters

adata AnnData

Annotated data matrix.

lr_df DataFrame | NoneOptional[DataFrame] (default: None)

A pandas dataframe, must contain three columns: ‘Ligand_gene_symbol’, ‘Receptor_gene_symbol’ and ‘LR_pair’.

gene_list_on_interest List[str] | NoneOptional[List[str]] (default: None)

User provided gene list for the predefined genes on interest.

min_mean float (default: 0.05)

The minimum value of the mean expression level of filtered genes. (provided in adata.var.means)

min_disp float (default: 0.5)

The minimum value of the dispersions of filtered gene expressions. (provided in adata.var.dispersions_norm)

max_sparse float (default: 0.5)

The percentage of cells required to express the filtered genes.

remove_lr_gene bool (default: True)

If True, filter out the ligand (or receptor) genes, to avoid data Leakage in predicting target gene expression using ligand–receptor pairs.

Return type

Tuple[Tensor, List[str]]

Returns

: target_all_gene_expr : Tensor

Generated target gene expression matrix (cell by gene, scaled to make the maximum expression level of each target gene become 1)

used_gene_listList[str]

The filtered target gene list for the following workflow.