add function to read adat files into widely used AnnData object

[gitbenlewis]

add function to read adat files into widely used AnnData object
read_adat_2_AnnData()
Returns an adata object instead a somalogic adat file path from the filepath/name.
Parameters
----------
path_or_buf : Union[str, TextIO]
Path or buffer that the file will be read from
Examples
--------
>>> adata=read_adat_2_AnnData(example_data_file)
Returns
-------
adata object instead of : Adat object

[kyoung73]

Hi @MegaDesk! First, thank you for your contribution! To maintain a lightweight codebase, we would prefer to avoid introducing additional dependencies, such as AnnData, into the source code for now. In the future, we may consider adding AnnData as an optional dependency and add a function like this in. It is recommended to create a helper function downstream (in your code) to encapsulate the entire workflow in a single function, simplifying the process. While slightly more convoluted, the following approach demonstrates how this can be achieved using read_adat:

import somadata as sd
import anndata as ad

adat = sd.read_adat("tests/data/control_data.adat")

row_metadata = {name: adat.index.get_level_values(name).tolist() for name in adat.index.names}
column_metadata = {name: adat.columns.get_level_values(name).tolist() for name in adat.columns.names}
header_metadata = {key: str(value) for key, value in adat.header_metadata.items()}

adata = ad.AnnData(X = adat.values, obs = row_metadata, var = column_metadata, uns = header_metadata)

Alternatively, I updated the code (1.2.0) to streamline using parse_file and default all header metadata values to be string type. Hope this makes things easier:

from somadata import parse_file
import anndata as ad
import numpy as np

rfu_matrix, row_metadata, column_metadata, header_metadata = parse_file("tests/data/control_data.adat", compatibility_mode=True)

adata = ad.AnnData(X = np.array(rfu_matrix), obs = row_metadata, var = column_metadata, uns = header_metadata)

Please reach out if there's any questions. Happy to discuss further!

[gitbenlewis]

Hi @kyoung73 ,

Yes these are all workable solutions to access the somadata.io.adat.file.parse_file() function.

I look forward to the addition of AnnData as an optional dependency. It is a perfect for this type of data (feature matrix with row/column metadata).

The multi index panda dataframe based object is not very end-user friendly in comparison to operations on an Anndata object.

Also storing the the data matrix and column/row metadata in Anndata object would remove the need for most of the helper functions in the somadata code base.