Cannot modify AnnData object

Help anndata
1

Hi,

I have a strange problem with the AnnData object. I can’t seem to modify any of the fields.

adata = sc.read_10x_h5(file)
adata.var_names_make_unique()

AnnData object with n_obs × n_vars = 6185 × 32285
var: ‘gene_ids’, ‘feature_types’, ‘genome’

when I run :

sc.pp.filter_cells(adata, min_counts=2000, inplace=False)

filtered out 1228 cells that have less than 2000 counts

it’s fine (don’t modify the object) but when I run :

sc.pp.filter_cells(adata, min_counts=2000, inplace=True) # → modify the object

I have :

/usr/local/lib/python3.11/dist-packages/anndata/_core/views.py:79: ImplicitModificationWarning: Trying to modify attribute .var of view, initializing view as actual.
container[idx] = value
/usr/local/lib/python3.11/dist-packages/anndata/_core/anndata.py:1908: UserWarning: Variable names are not unique. To make them unique, call .var_names_make_unique.
utils.warn_names_duplicates(“var”)
/usr/local/lib/python3.11/dist-packages/anndata/_core/views.py:79: ImplicitModificationWarning: Trying to modify attribute .var of view, initializing view as actual.
container[idx] = value
/usr/local/lib/python3.11/dist-packages/anndata/_core/anndata.py:1908: UserWarning: Variable names are not unique. To make them unique, call .var_names_make_unique.
utils.warn_names_duplicates(“var”)
/usr/local/lib/python3.11/dist-packages/anndata/_core/views.py:79: ImplicitModificationWarning: Trying to modify attribute .var of view, initializing view as actual.
container[idx] = value

ad infinitum, or more exactly until a maximum recursion depth exceeded error is raised.

In fact I can’t modify any of the .obs or .var fields, even this gives me the same error (adding a ‘test’ field):

adata.var[‘test’] = adata.var[‘gene_ids’]

Thanks for your help !

(packages version):

scanpy==1.9.8 anndata==0.10.5.post1 umap==0.5.4 numpy==1.26.1 scipy==1.11.3 pandas==2.1.2 scikit-learn==1.3.2 statsmodels==0.14.1 pynndescent==0.5.10

2

I believe the issue is the version of pandas. Right not anndata actually it’s not compatible with pandas==2.1.2, which should have meant pip would complain about it.

If updating your version of pandas doesn’t help, please open an issue on the anndata issue tracker: Issues · scverse/anndata · GitHub

3

Indeed, that’s what it was. Upgrading to pandas 2.2 solved the problem - thank for your help.

1 Like