fusor.find_initial_pivots AttributeError: 'AnnData' object has no attribute 'mean'

[mcaponegro]

Hi,

Thank you for the great package and documentation. I am really looking forward to applying this to my own data and testing different data modalities. I have completed the install and CODEX vignette successfully, however, I am running into the error below on my own data. Everything completes successfully up until this point.

I will say that I am converting a Seurat .Rds into AnnData and running the python code. This has worked well for me in the past. I am wondering if you think the issue lies within the arrays being converted or if it is something else I am missing.

sc <- import('scanpy', convert = FALSE)
py$adata <- sc$AnnData(
    X   = t(as.matrix(GetAssayData(seurat_object, assay = assay, slot='counts'))),
    obs = seurat_object[[]],
    var = GetAssay(seurat_object)[[]]
  )

...process as in maxfuse vingette...

The error:

>>> fusor.find_initial_pivots(
...     wt1=0.3, wt2=0.3,
...     svd_components1=25, svd_components2=20
... )
Now at batch 0<->0...

Traceback (most recent call last):
  File "<string>", line 1, in <module>
  File "/ldata_fast/python_envs/maxfuse/lib/python3.8/site-packages/maxfuse/model.py", line 619, in find_initial_pivots
    arr1 = utils.get_centroids(
  File "/ldata_fast/python_envs/maxfuse/lib/python3.8/site-packages/maxfuse/utils.py", line 271, in get_centroids
    centroids[curr_label, :] = arr[indices, :].mean(axis=1)
AttributeError: 'AnnData' object has no attribute 'mean'

Additionally, the step prior to investigate the elbow plot this error, if it helps.

>>> fusor.plot_singular_values(
    target='shared_arr1',
    n_components=None,
)

Traceback (most recent call last):
  File "<string>", line 1, in <module>
  File "/ldata_fast/python_envs/maxfuse/lib/python3.8/site-packages/maxfuse/model.py", line 346, in plot_singular_values
    _, s, _ = utils.robust_svd(arr=target, n_components=n_components, randomized=randomized_svd, n_runs=svd_runs)
  File "/ldata_fast/python_envs/maxfuse/lib/python3.8/site-packages/maxfuse/utils.py", line 308, in robust_svd
    u, s, vh = svds(arr*1.0, k=n_components) # svds can not handle integer values
TypeError: unsupported operand type(s) for *: 'AnnData' and 'float'

[mcaponegro]

update. my first error was not suppling an array to the fusor.plot_singular_values() function. when I convert to array using

rna_shared = rna_shared.X.copy()
protien_shared = protien_shared.X.copy()

I now get this error:

>>> fusor.find_initial_pivots(
...     wt1=0.3, wt2=0.3,
...     svd_components1=25, svd_components2=20
... )
Now at batch 0<->0...
Traceback (most recent call last):
  File "<string>", line 1, in <module>
  File "/ldata_fast/python_envs/maxfuse/lib/python3.8/site-packages/maxfuse/model.py", line 639, in find_initial_pivots
    match_utils.get_initial_matching(
  File "/ldata_fast/python_envs/maxfuse/lib/python3.8/site-packages/maxfuse/match_utils.py", line 179, in get_initial_matching
    arr1 = utils.svd_denoise(
  File "/ldata_fast/python_envs/maxfuse/lib/python3.8/site-packages/maxfuse/utils.py", line 334, in svd_denoise
    u, s, vh = robust_svd(arr, n_components=n_components, randomized=randomized, n_runs=n_runs)
  File "/ldata_fast/python_envs/maxfuse/lib/python3.8/site-packages/maxfuse/utils.py", line 308, in robust_svd
    u, s, vh = svds(arr*1.0, k=n_components) # svds can not handle integer values
  File "/ldata_fast/python_envs/maxfuse/lib/python3.8/site-packages/scipy/sparse/linalg/_eigen/_svds.py", line 447, in svds
    args = _iv(A, k, ncv, tol, which, v0, maxiter, return_singular_vectors,
  File "/ldata_fast/python_envs/maxfuse/lib/python3.8/site-packages/scipy/sparse/linalg/_eigen/_svds.py", line 49, in _iv
    raise ValueError(message)
ValueError: `k` must be an integer satisfying `0 < k < min(A.shape)`.

[mcaponegro]

further update: this was due to my shared array dimensions being less than the input for svd components. This solved my error with this function. I will close. Also, the above conversion from seurat seems to work here, for those interested.

[zongmingma]

Thanks a lot for your input and interest in our work! We will be happy to answer any further questions.