sfaira.data.StoreSingleFeatureSpace

class sfaira.data.StoreSingleFeatureSpace(adata_by_key: Dict[str, anndata._core.anndata.AnnData], indices: Dict[str, numpy.ndarray], obs_by_key: Union[None, Dict[str, dask.dataframe.core.DataFrame]] = None, data_source: str = 'X')

Data set group class tailored to data access requirements common in high-performance computing (HPC).

This class does not inherit from DatasetGroup because it entirely relies on the cached objects. This class is centred around .adata_by_key and .indices.

.adata_by_key is a dictionary (by id) of backed anndata instances that point to individual h5ads. This dictionary is intialised with all h5ads in the store. As the store is sub-setted, key-value pairs are deleted from this dictionary.

.indices have keys that correspond to keys in .adata_by_key and contain index vectors of observations in the anndata instances in .adata_by_key which are still kept. These index vectors are a form of lazy slicing that does not require data set loading or re-writing. As the store is sub-setted, key-value pairs are deleted from this dictionary if no observations from a given key match the sub-setting. If a subset of observations from a key matches the subsetting operation, the index set in the corresponding value is reduced. All data retrieval operations work on .indices: Generators run over these indices when retrieving observations for example.

Attributes

adata_by_key	Anndata instance for each selected data set in store, sub-setted by selected cells.
adata_memory_footprint	Memory foot-print of data set k in MB.
data_by_key	Data matrix for each selected data set in store, sub-setted by selected cells.
dataset_weights
genome_container
idx	Global indices.
indices	Indices of observations that are currently exposed in adata of this instance.
n_obs	Number of observations selected in store.
n_vars	Number of selected features per organism in store
obs_by_key
organism	Organism of store.
organisms_by_key	Data set-wise organism label as dictionary of data set keys.
shape
var
var_names	Feature names of selected genes by organism in store.
data_source

Methods

checkout([idx, batch_size, ...])	Yields an instance of a generator class over observations in the contained data sets.
get_subset_idx(attr_key, values, excluded_values)	Get indices of subset list of adata objects based on cell-wise properties.
load_config(fn)	Load a config file and recreates a data sub-setting.
subset(attr_key[, values, excluded_values, ...])	Subset list of adata objects based on cell-wise properties.
write_config(fn)	Writes a config file that describes the current data sub-setting.