dotools_py.tl.run_scvi

dotools_py.tl.run_scvi(adata, batch_key, gene_key='highly_variable', layer_counts='counts', categorical_covariates=None, continuous_covariates=None, n_hidden=128, n_latent=30, n_layers=3, dispersion='gene-batch', gene_likelihood='zinb', get_model=False, **kwargs)

Run scVI.

Run scVI to integrate sc/snRNA more information on scvi-tools.

Parameters:

adata AnnData

Annotated data matrix.

batch_key str

Column in adata.obs with batch information.

gene_key Union[str, Literal['all']] (default: 'highly_variable')

Boolean column in adata.var used to select the genes that will be used for the inference.

layer_counts str (default: 'counts')

Layer in adata.layers with raw counts.

categorical_covariates list (default: None)

Column in adata.obs with categorical covariates to correct for during scVI inference.

continuous_covariates list (default: None)

Column in adara.obs with continuous covariates to correct for during scVI inference.

n_hidden int (default: 128)

Number of hidden layers.

n_latent int (default: 30)

Dimensions of the latent space.

n_layers int (default: 3)

Number of layers

dispersion Literal['gene', 'gene-batch', 'gene-label', 'gene-cell'] (default: 'gene-batch')

Gene dispersion mode for scVI.

gene_likelihood Literal['zinb', 'nb', 'poisson', 'normal'] (default: 'zinb')

Gene likelihood.

get_model bool (default: False)

Return the trained model.

kwargs

Additional arguments for scvi.model.SCVI.

Return type:

SCVI | None

Returns:

Returns None or the trained scVI model if get_model is set to True. The latent space is saved in the AnnData under X_scVI.