HoloNet.tools.filter_ce_tensor#

HoloNet.tools.filter_ce_tensor(ce_tensor, adata, lr_df, elements_expr_df_dict, w_best, n_pairs=200, thres=0.1, distinguish=True, copy=True)#

Filter the edge in calculated CE tensor, removing the edges with low specificities.

For each LR pair, select faked ligand and receptor genes, which have similar expression levels with the ligand and receptor gene in the dataset. Then calculate the background CE tensor using faked LR genes,

Using permutation tests, require filtered edges with communication event strength larger than a proportion of background strengthes.

Parameters
ce_tensor Tensor

Calculated CE tensor (LR_pair_num * cell_num * cell_num) by “compute_ce_tensor” function

adata AnnData

Annotated data matrix.

lr_df DataFrame

A preprocessed LR-gene dataframe.

elements_expr_df_dict dict

metadata from ‘elements_expr_df_calculate’ function.

w_best float

A distance parameter in edge weighting function controlling the covering region of ligands. ‘default_w_visium’ function provides a recommended value of w_best.

n_pairs int (default: 200)

The number of faked ligand and receptor genes.

thres float (default: 0.1)

We require filtered edges with communicatin event strength larger than a proportion of background strengthes. The parameter is the proportion.

distinguish bool (default: True)

If True, set the different w_best for secreted ligands and plasma-membrane-binding ligands.

copy bool (default: True)

If False, change the input ce_tensor and save memory consumption.

Return type

Tensor

Returns

: A CE tensor which removed the edges with low specificities (LR_pair_num * cell_num * cell_num).