plot_gene_loadings error

[kanghelen]

Hello!

After running iNMF, I am trying to plot the gene loadings on UMAP/tSNE using plot_gene_loadings function. However, there is an error

    markers = get_factor_markers(                                                                                                       
              ^^^^^^^^^^^^^^^^^^^                                                                                                       
  File "/.conda/envs/inmf_env/lib/python3.12/site-packages/pyliger/tools/_marker.py", line 112, in get_factor_mark\
ers                                                                                                                                     
    cell_label = np.concatenate(                                                                                                        
                 ^^^^^^^^^^^^^^^                                                                                                        
TypeError: only integer scalar arrays can be converted to a scalar index   

The code on that line shows

cell_label = np.concatenate(
    np.repeat(dataset1, np.sum(labels[dataset1] == factor)),
    np.repeat(dataset1, np.sum(labels[dataset2] == factor)),
)

But when I tried to change it to the code below, there is no error. (Also note the change to "dataset2" on the third line.)

cell_label = np.concatenate(
    [np.repeat(dataset1, np.sum(labels[dataset1] == factor)),
    np.repeat(dataset2, np.sum(labels[dataset2] == factor))]
)

Could I ask if this is a reasonable fix?

[theAeon]

seems this is a dup of #5 but i'm going to close that one as its old

[theAeon]

give it a shot off the master branch and let me know!

[kanghelen]

Hi @theAeon! Thanks! It seems like I don't have the permission to push a branch.

[theAeon]

I think you may want to pull the branch, not push 😅

[kanghelen]

I don't think I can submit any changes to the repo (e.g. trying to push changes in a branch and create a pull request).

Sorry for any confusion. First time contributing to a public repo!

[theAeon]

oh! I see-you have to fork it for that. I was confused because I (think) I already fixed it on the master branch. Can you give it a test?

[kanghelen]

Got it! I'll try forking next time. Thanks!

I think you fixed the TypeError: only integer scalar arrays can be converted to a scalar index
There might be a logical error:

pyliger/src/pyliger/tools/_marker.py

Line 114 in b869a3b

np.repeat(dataset1, np.sum(labels[dataset2] == factor))),

The second part should be dataset2 so that wilcoxon test can compare between two datasets.

This will cause a subsequent error:

u, z = _mannwhitneyu(ranks[idx], ranks[~idx], T)
ZeroDivisionError: division by zero

[theAeon]

I'm getting a divide by zero error on my end even if I change that to dataset2. Currently waiting on the person who wrote it to take a look to figure out what that was supposed to be.

[theAeon]

I looked at the R source code and found that you're right about it being dataset2, not dataset1.

https://github.com/welch-lab/liger/blob/58e59572385688b5975f4d70c0537776a33edd74/R/factorMarker.R#L136-L137

Still not sure what the deal with the divide-by-zero error is.