opendvp.tl.impute_gaussian

impute_gaussian(adata, mean_shift=-1.8, std_dev_shift=0.3, perSample=False, layer_key='unimputed', uns_key='impute_gaussian_qc_metrics')

Impute missing values in an AnnData object using a Gaussian distribution.

This function imputes missing values in the data matrix using a Gaussian distribution, with the mean shifted and the standard deviation scaled. Imputation can be performed per protein (column) or per sample (row).

The original, un-imputed data matrix is stored in adata.layers. A DataFrame with quality control metrics for the imputation is stored in adata.uns. The QC metrics include the number of imputed values, the mean and standard deviation used for imputation, and a numpy array of the imputed values themselves for each feature.

Parameters:

• adata (anndata.AnnData) – AnnData object with missing values to impute.

• mean_shift (float, default -1.8) – Number of standard deviations to shift the mean of the Gaussian distribution.

• std_dev_shift (float, default 0.3) – Factor to scale the standard deviation of the Gaussian distribution.

• perSample (bool, default False) – If True, impute per sample (row); if False, impute per protein (column).

• layer_key (str, default 'unimputed') – Key under which to store the original, un-imputed data matrix in adata.layers.

• uns_key (str, default 'impute_gaussian_qc_metrics') – Key under which to store the imputation QC metrics DataFrame in adata.uns.

Returns:

AnnData object with imputed values in .X, the original matrix in .layers[layer_key], and QC metrics in .uns[uns_key].

Return type:

anndata.AnnData