AttributeError: 'numpy.ndarray' object has no attribute 'todense' with H5AD file as input for 'scaden simulate'

[WanderingHedgie]

Hi !
I want to use H5AD as input format to use scaden simulate as it is indicated in the help comand, but here is the error I get :

(/CONDAS/users/username/scaden) [username@gknwwd2 example]$ scaden simulate -n 100 -d ./example_data/ -f h5ad --pattern "*.h5ad"

     ____                _            
    / ___|  ___ __ _  __| | ___ _ __  
    \___ \ / __/ _` |/ _` |/ _ \ '_ \ 
     ___) | (_| (_| | (_| |  __/ | | |
    |____/ \___\__,_|\__,_|\___|_| |_|

    
INFO     Datasets: ['reference']                                                             bulk_simulator.py:84
INFO     Simulating data from reference                                                      bulk_simulator.py:89
INFO     Loading reference dataset ...                                                      bulk_simulator.py:141
Traceback (most recent call last):
  File "/CONDAS/users/username/scaden/bin/scaden", line 10, in <module>
    sys.exit(main())
  File "/CONDAS/users/username/scaden/lib/python3.8/site-packages/scaden/__main__.py", line 48, in main
    cli()
  File "/CONDAS/users/username/scaden/lib/python3.8/site-packages/click/core.py", line 1157, in __call__
    return self.main(*args, **kwargs)
  File "/CONDAS/users/username/scaden/lib/python3.8/site-packages/click/core.py", line 1078, in main
    rv = self.invoke(ctx)
  File "/CONDAS/users/username/scaden/lib/python3.8/site-packages/click/core.py", line 1688, in invoke
    return _process_result(sub_ctx.command.invoke(sub_ctx))
  File "/CONDAS/users/username/scaden/lib/python3.8/site-packages/click/core.py", line 1434, in invoke
    return ctx.invoke(self.callback, **ctx.params)
  File "/CONDAS/users/username/scaden/lib/python3.8/site-packages/click/core.py", line 783, in invoke
    return __callback(*args, **kwargs)
  File "/CONDAS/users/username/scaden/lib/python3.8/site-packages/scaden/__main__.py", line 207, in simulate
    simulation(
  File "/CONDAS/users/username/scaden/lib/python3.8/site-packages/scaden/simulate.py", line 22, in simulation
    bulk_simulator.simulate()
  File "/CONDAS/users/username/scaden/lib/python3.8/site-packages/scaden/simulation/bulk_simulator.py", line 90, in simulate
    self.simulate_dataset(dataset)
  File "/CONDAS/users/username/scaden/lib/python3.8/site-packages/scaden/simulation/bulk_simulator.py", line 104, in simulate_dataset
    data_x, data_y = self.load_dataset(dataset)
  File "/CONDAS/users/username/scaden/lib/python3.8/site-packages/scaden/simulation/bulk_simulator.py", line 201, in load_dataset
    x = pd.DataFrame(data_h5ad.X.todense()) 
AttributeError: 'numpy.ndarray' object has no attribute 'todense'

The reason of this is that I write my H5AD file at the end of preprocessing provided in Jupyter Notebooks available in paper_data.tar.gz on figshare (version 1).

The thing is, after the preprocessing, the data is stored in the anndata.X as numpy array with dense data (obtained throug preprocessing with regress_out() function) and cannot be handled by the simulate function. I suppose that the simulate function aimed to process sparse data, even in H5AD format, with a sparse matrix (see Scipy sparse formats).

Maybe it would be a good idea to add a warning in the documentation and/or add a verification in the code to handle this case.

Hope it will be useful for someone !

WanderingHedgie