rapids_singlecell.gr.ligrec

Contents

rapids_singlecell.gr.ligrec#

rapids_singlecell.gr.ligrec(adata, cluster_key, *, clusters=None, interactions=None, complex_policy='min', threshold=0.01, corr_method=None, corr_axis='clusters', alpha=0.05, use_raw=True, copy=False, key_added=None, gene_symbols=None, n_perms=1000, interactions_params={}, transmitter_params={'categories': 'ligand'}, receiver_params={'categories': 'receptor'})[source]#

Perform the permutation test as described in [Efremova et al., 2020].

Parameters:
adata AnnData

Annotated data object.

cluster_key str

Key in obs where clustering is stored.

clusters list | None (default: None)

Clusters from obs ['{cluster_key}']. Can be specified either as a sequence of tuple or just a sequence of cluster names, in which case all combinations considered.

interactions DataFrame | Mapping | Sequence | None (default: None)

Interaction to test. The type can be one of:

  • pandas.DataFrame - must contain at least 2 columns named source and target.

  • dict - dictionary with at least 2 keys named source and target.

  • typing.Sequence - Either a sequence of str, in which case all combinations are produced, or a sequence of tuple of 2 str or a tuple of 2 sequences.

If None, the interactions are extracted from omnipath. Protein complexes can be specified by delimiting the components with _, such as alpha_beta_gamma.

complex_policy Literal['min', 'all'] (default: 'min')

Policy on how to handle complexes. Valid options are:

  • min - select gene with the minimum average expression. This is the same as in [Efremova et al., 2020].

  • all - select all possible combinations between source and target complexes.

threshold float (default: 0.01)

Do not perform permutation test if any of the interacting components is being expressed in less than threshold percent of cells within a given cluster.

corr_method str | None (default: None)

Correction method for multiple testing. See statsmodels.stats.multitest.multipletests() for valid options.

corr_axis Literal['interactions', 'clusters'] (default: 'clusters')

Axis over which to perform the FDR correction. Only used when corr_method != None. Valid options are:

  • interactions - correct interactions by performing FDR correction across the clusters.

  • clusters - correct clusters by performing FDR correction across the interactions.

alpha float (default: 0.05)

Significance level for FDR correction. Only used when corr_method != None.

use_raw bool (default: True)

Whether to access raw

copy bool (default: False)

If True, return the result, otherwise save it to the adata object.

key_added str | None (default: None)

Key in uns where the result is stored if copy = False. If None, ‘{cluster_key}_ligrec’ will be used.

gene_symbols str | None (default: None)

Key in var to use instead of var_names.

n_perms int (default: 1000)

Number of permutations for the permutation test.

interactions_params dict (default: {})

Keyword arguments for omnipath.interactions.import_intercell_network() defining the interactions. These datasets from [Türei et al., 2016] are used by default: omnipath, pathwayextra, kinaseextra and ligrecextra.

transmitter_params dict (default: {'categories': 'ligand'})

Keyword arguments for omnipath.interactions.import_intercell_network() defining the transmitter side of intercellular connections.

receiver_params dict (default: {'categories': 'receptor'})

Keyword arguments for omnipath.interactions.import_intercell_network() defining the receiver side of intercellular connections.

Return type:

DataFrame | None

Returns:

If copy = True, returns a dict with following keys:
Otherwise, modifies the adata object with the following key:
  • uns ['{key_added}'] -

    the above mentioned dict.

NaN p-values mark combinations for which the mean expression of one of the interacting components was 0 or it didn’t pass the threshold percentage of cells being expressed within a given cluster.

Contents