mosaicmpi.Dataset

class mosaicmpi.Dataset(adata: AnnData, force_migrate: bool = False)

Creates a Dataset object from an anndata.AnnData object.

Parameters:

• adata (anndata.AnnData) – AnnData object with data

• force_migrate (bool, optional) – forces conversion of AnnData objects even when adata.X and adata.raw.X are not linearly scaled relative to each other, defaults to False

Raises:

• RuntimeError – Backed-mode Anndata objects cannot be migrated

• ValueError – Error is raised when force is False and adata is non-linearly scaled.

Returns:

Object with expression and metadata

Return type:

Dataset

Attributes

has_cnmf_results	Test for wehther Dataset contains cNMF results for the dataset
hvf	Highly variable features used for cNMF
hvf_stats	Outputs the highly variable genes dataframe in pandas-compatible format.
is_imputed	Outputs the imputation status of the dataset.
is_normalized	Outputs the normalization status of the dataset.
mosaicmpi_version	mosaicMPI version used to create the dataset
patient_id_col	Outputs the normalization status of the dataset.

Methods

add_cnmf_results(cnmf_output_dir, cnmf_name)	After factorization, add completed cNMF results in [cnmf_output_dir]/[cnmf_name] to the dataset object.
append_to_history(entry)	Add entry to Dataset history.
calculate_cnmf_prediction_error([k])	Calculate cNMF prediction error using the method in the original cNMF package.
cross_validate_imputation(imputer[, n_folds])	Perform k-fold cross validation of imputation on the dataset without modifying the data.
from_anndata(adata[, force_migrate])	Creates a Dataset object from an anndata.AnnData object.
from_df(data, is_normalized[, sparsify, ...])	Creates a Dataset object from a pandas DataFrame.
from_h5ad(h5ad_file[, force_migrate, backed])	Creates a Dataset object from an AnnData-compatible .h5ad file.
get_approximation([k, program_type])	Return the approximated data by multiplying the programs and usage matrices for a given rank (k).
get_category_overrepresentation(layer[, ...])	Calculate Pearson residual of chi-squared test, associating programs for each rank (k) to categories of samples/observations.
get_history()	Returns timestamped history of Dataset object.
get_metadata_correlation(layer[, method])	Calculate correlation of program usage to numerical metadata across samples/observations.
get_metadata_df([include_categorical, ...])	Get sample/observation metadata.
get_printable_metadata_type_summary()	Return a printable summary of metadata and the types.
get_programs([k, type])	Get feature scores for programs.
get_usages([k, discretize, normalize])	Generate dataframe of program usage.
impute_knn([n_neighbors, weights, ...])	Imputation for completing missing values using k-Nearest Neighbors.
impute_zeros([cross_validate, n_folds])	Imputation by filling missing values with zeros.
initialize_cnmf(cnmf_output_dir, cnmf_name)	Initialize a cNMF run for subsequent factorization.
map_gene_ids(source_species, dest_species, ...)	Map the feature IDs in place for a dataset.
remove_cnmf_results()
remove_unfactorizable_features()	Removes features with missing values or zero variance from the data matrix.
remove_unfactorizable_observations()	Removes observations with all zeros from the data matrix.
select_hvf([stratify_by, stratify_mode, ...])	Select highly variable features (HVFs) for cNMF factorization.
select_hvf_cnmf([stratify_by, ...])
select_hvf_stdeconvolve([stratify_by, ...])
to_df([normalized])	Get data matrix as a pd.DataFrame.
update_obs(obs)	Update the observation metadata with a new metadata matrix
validate_cnmf_prediction_errors([tolerance])	Validate the dataset and cNMF solutions for each rank by comparing the prediction error values stored in the object [self.adata.uns.kvals] to those calculated from the dataset's data matrices [based on self.adata.X and self.adata.varm['cnmf_gep_raw']].
validate_feature_stats([tolerance])	Validate the dataset and cNMF solutions for each rank by comparing the calculated feature statistics (mean, SD, variance) stored in the object [self.adata.var] to those calculated from the dataset's data matrices [based on self.adata.X].
write_h5ad(filename[, safe_mode])	Write dataset to .h5ad file.