liana.mt.lrMistyData

liana.mt.lrMistyData(adata, resource_name='consensus', resource=None, nz_threshold=0.1, use_raw=False, layer=None, spatial_key='spatial', kernel='misty_rbf', bandwidth=100, set_diag=False, cutoff=0.1, zoi=0, verbose=False)

Generate a MistyData object from an AnnData object in ligand-receptor format.

Parameters:

adata AnnData

AnnData object

resource_name str (default: 'consensus')

The name of the resource to use. See show_resources for available resources.

resource DataFrame | None (default: None)

A resource in the form of a pandas DataFrame. If None, the resource is selected using select_resource.

nz_threshold float (default: 0.1)

The threshold for the number of non-zero entries in each view.

use_raw bool (default: False)

Whether to use the raw data of the AnnData object.

layer str | None (default: None)

The layer of the AnnData object to use.

spatial_key str (default: 'spatial')

The key in adata.obsm where the spatial coordinates are stored.

kernel Literal['gaussian', 'exponential', 'linear', 'misty_rbf'] (default: 'misty_rbf')

A radial basis function kernel to use for the generation of the connectivity matrix for the extra view. Default is ‘misty_rbf’, a kernel derivative of a Gaussian kernel.

bandwidth float (default: 100)

The bandwidth of the kernel.

set_diag bool (default: False)

Whether to set the diagonal of the connectivity matrix to 1.

cutoff float (default: 0.1)

The minimum value cutoff for the connectivity matrix.

zoi float (default: 0)

Zone of indifference of the kernel, i.e. the kernel is set to 0 for distances smaller than zoi.

verbose bool (default: False)

Whether to print progress.

Return type:

MistyData

Returns:

A MistyData object with receptors in the intra view & ligands in the extra view.

Examples

Splits the data by the roles in a ligand-receptor resource: receptors become the targets ('intra'), and the ligands of neighbouring cells the predictors ('extra'):

>>> import liana as li
>>> adata = li.ds.generate_toy_spatial()
>>> misty = li.mt.lrMistyData(adata, bandwidth=200)

Fitting this asks, per receptor, which neighbouring ligands predict it – call the object with bypass_intra=True so that the receptors are not predicted from each other. See liana.mt.MistyData.__call__().