CellChat AttributeError: 'SparseCSRView' object has no attribute 'A'

[AmyOlex]

Describe the bug
I am trying to run CellChat and I am getting a AttributeError: 'SparseCSRView' object has no attribute 'A'.

To Reproduce
Here is the code to call the method. The "adata" is a Seurat object with the Data slot (log normalized) converted to an h5ad file using R and imported as an anndata formatted object in Python using the read_h5ad method. I was able to get CellPhoneDB method to run fine.

cellchat(adata,
use_raw=False,
groupby='cell_type',
# NOTE by default the resource uses HUMAN gene symbols
resource_name='consensus',
expr_prop=0.25,
verbose=True, key_added='cellchat_res')

Screenshots
No screen shot, but here is the traceback:

AttributeError Traceback (most recent call last)
Cell In[23], line 2
1 # run cellchat
----> 2 cellchat(adata,
3 use_raw=False,
4 groupby='cell_type',
5 # NOTE by default the resource uses HUMAN gene symbols
6 resource_name='consensus',
7 expr_prop=0.25,
8 verbose=True, key_added='cellchat_res')

File ~/anaconda3/envs/env_liana_p311/lib/python3.11/site-packages/liana/method/sc/_Method.py:244, in Method.call(self, adata, groupby, resource_name, expr_prop, min_cells, groupby_pairs, base, supp_columns, return_all_lrs, key_added, use_raw, layer, de_method, n_perms, seed, n_jobs, resource, interactions, mdata_kwargs, inplace, verbose)
241 if supp_columns is None:
242 supp_columns = []
--> 244 liana_res = liana_pipe(adata=adata,
245 groupby=groupby,
246 resource_name=resource_name,
247 resource=resource,
248 interactions=interactions,
249 expr_prop=expr_prop,
250 min_cells=min_cells,
251 supp_columns=supp_columns,
252 return_all_lrs=return_all_lrs,
253 groupby_pairs=groupby_pairs,
254 base=base,
255 de_method=de_method,
256 verbose=verbose,
257 _score=self._method,
258 n_perms=n_perms,
259 seed=seed,
260 n_jobs=n_jobs,
261 use_raw=use_raw,
262 layer=layer,
263 mdata_kwargs=mdata_kwargs
264 )
265 if inplace:
266 adata.uns[key_added] = liana_res

File ~/anaconda3/envs/env_liana_p311/lib/python3.11/site-packages/liana/method/sc/_liana_pipe.py:171, in liana_pipe(adata, groupby, resource_name, resource, interactions, groupby_pairs, expr_prop, min_cells, base, de_method, n_perms, seed, verbose, use_raw, n_jobs, layer, supp_columns, return_all_lrs, _score, _methods, _consensus_opts, _aggregate_method, mdata_kwargs)
167 print(f"Generating ligand-receptor stats for {adata.shape[0]} samples "
168 f"and {adata.shape[1]} features")
170 # Get lr results
--> 171 lr_res = _get_lr(adata=adata,
172 resource=resource,
173 groupby_pairs=groupby_pairs,
174 mat_mean=mat_mean,
175 mat_max=mat_max,
176 relevant_cols=_key_cols + _add_cols + _complex_cols,
177 de_method=de_method,
178 base=base,
179 verbose=verbose
180 )
182 # Ligand and receptor score based on unfiltered cluster mean and cluster std. Handles protein complexes
183 if (M.ligand_cdf in _add_cols) or (M.receptor_cdf in _add_cols):

File ~/anaconda3/envs/env_liana_p311/lib/python3.11/site-packages/liana/method/sc/_liana_pipe.py:308, in _get_lr(adata, resource, groupby_pairs, relevant_cols, mat_mean, mat_max, de_method, base, verbose)
306 dedict[label]['logfc'] = _calc_log2fc(adata, label)
307 if isinstance(mat_max, np.float32): # cellchat flag
--> 308 dedict[label]['trimean'] = _trimean(temp.X / mat_max)
310 pairs = (pd.DataFrame(np.array(np.meshgrid(labels, labels))
311 .reshape(2, np.size(labels) * np.size(labels)).T)
312 .rename(columns={0: P.source, 1: P.target}))
314 if groupby_pairs is not None:

File ~/anaconda3/envs/env_liana_p311/lib/python3.11/site-packages/liana/method/sc/_liana_pipe.py:463, in _trimean(a, axis)
462 def _trimean(a, axis=0):
--> 463 quantiles = np.quantile(a.A, q=[0.25, 0.75], axis=axis)
464 median = np.median(a.A, axis=axis)
465 return (quantiles[0] + 2 * median + quantiles[1]) / 4

AttributeError: 'SparseCSRView' object has no attribute 'A'

Thank you for any assistance!

[dbdimitrov]

Hi @AmyOlex,

This appears to be related to a scipy version. I will address in the following update but in the meantime, downgrading scipy to e.g. 1.10.1 will resolve it.

Daniel

[Zach-Sten]

downgrading scipy to 1.13.1 also fixes this if for some reason others need a more recent version of scipy.

[AmyOlex]

Thank you for the suggestions. I tried downgrading to v1.13.1, but that caused dependency conflicts with Numba. Installing v1.10.1 resolved all the issues though and CellChat is working now. Thanks!

[dbdimitrov]

I should double check this and make sure that it works with more recent scipy version

[Rose-Hedderman]

This worked for me:
conda create --name CCDB python=3.11 numpy=2.0
pip install liana cell2cell decoupler omnipath seaborn==0.11
pip install matplotlib==3.8.0
conda install scipy=1.13.1 numpy=1.26.4 numba=0.60.0