scvi.data.fields.ObsmField#

class scvi.data.fields.ObsmField(registry_key, obsm_key, colnames_uns_key=None, is_count_data=False, correct_data_format=True)[source]#

An AnnDataField for an .obsm field in the AnnData data structure.

In addition to creating a reference to the .obsm field, stores the column keys for the obsm field in a more accessible .uns attribute.

Parameters

registry_key : str: Key to register field under in data registry.
obsm_key : str: Key to access the field in the AnnData .obsm mapping.
colnames_uns_key : str | NoneOptional[str] (default: None): Key to access column names corresponding to each column of the .obsm field in the AnnData .uns mapping. If None, checks if the field is stored as a dataframe. If so, uses the dataframe’s colnames. Otherwise, generates sequential column names (e.g. 1, 2, 3, etc.).
is_count_data : bool (default: False): If True, checks if the data are counts during validation.
correct_data_format : bool (default: True): If True, checks and corrects that the AnnData field is C_CONTIGUOUS and csr if it is dense numpy or sparse respectively.

Attributes table#

`COLUMN_NAMES_KEY`
`attr_key`	The key of the data field within the relevant AnnData attribute.
`attr_name`	The name of the AnnData attribute where the data is stored.
`is_empty`	Returns True if the field is empty as a function of its kwargs.
`registry_key`	The key that is referenced by models via a data loader.

Methods table#

`get_data_registry`()	Returns a nested dictionary which describes the mapping to the AnnData data field.
`get_field_data`(adata)	Returns the requested data as determined by the field for a given AnnData object.
`get_summary_stats`(state_registry)	Returns a dictionary comprising of summary statistics relevant to the field.
`register_field`(adata)	Sets up the AnnData object and creates a mapping for scvi-tools models to use.
`transfer_field`(state_registry, adata_target, ...)	Takes an existing scvi-tools setup dictionary and transfers the same setup to the target AnnData.
`validate_field`(adata)	Validates whether an AnnData object is compatible with this field definition.
`view_state_registry`(state_registry)	Returns a `rich.table.Table` summarizing a state registry produced by this field.

Attributes#

COLUMN_NAMES_KEY#

ObsmField.COLUMN_NAMES_KEY = 'column_names'#

attr_key#

ObsmField.attr_key#

Return type: str

attr_name#

ObsmField.attr_name#

Return type: str

is_empty#

ObsmField.is_empty#

Return type: bool

registry_key#

ObsmField.registry_key#

Return type: str

Methods#

get_data_registry#

ObsmField.get_data_registry()#

Returns a nested dictionary which describes the mapping to the AnnData data field.

The dictionary is of the form {“attr_name”: attr_name, “attr_key”: attr_key}. This mapping is then combined with the mappings of other fields to make up the data registry.

Return type: dict

get_field_data#

ObsmField.get_field_data(adata)#

Returns the requested data as determined by the field for a given AnnData object.

Return type: ndarray | DataFrameUnion[ndarray, DataFrame]

get_summary_stats#

ObsmField.get_summary_stats(state_registry)[source]#

Returns a dictionary comprising of summary statistics relevant to the field.

Parameters

state_registry : dict: Dictionary returned by register_field(). Summary stats should always be a function of information stored in this dictionary.

Return type

dict

Returns

summary_stats_dict The dictionary is of the form {summary_stat_name: summary_stat_value}. This mapping is then combined with the mappings of other fields to make up the summary stats mapping.

register_field#

ObsmField.register_field(adata)[source]#

Sets up the AnnData object and creates a mapping for scvi-tools models to use.

Return type: dict
Returns: dict A dictionary containing any additional state required for scvi-tools models not stored directly on the AnnData object.

transfer_field#

ObsmField.transfer_field(state_registry, adata_target, **kwargs)[source]#

Takes an existing scvi-tools setup dictionary and transfers the same setup to the target AnnData.

Used when one is running a pretrained model on a new AnnData object, which requires the mapping from the original data to be applied to the new AnnData object.

Parameters

state_registry : dict: state_registry dictionary created after registering an AnnData using an AnnDataManager object.
adata_target : AnnData: AnnData object that is being registered.
**kwargs: Keyword arguments to modify transfer behavior.

Return type

dict

Returns

dict A dictionary containing any additional state required for scvi-tools models not stored directly on the AnnData object.

validate_field#

ObsmField.validate_field(adata)[source]#

Validates whether an AnnData object is compatible with this field definition.

Return type: None

view_state_registry#

ObsmField.view_state_registry(state_registry)[source]#

Returns a rich.table.Table summarizing a state registry produced by this field.

Parameters

state_registry : dict: Dictionary returned by register_field(). Printed summary should always be a function of information stored in this dictionary.

Return type

Table | NoneOptional[Table]

Returns

state_registry_summary Optional rich.table.Table summarizing the state_registry.