Mismatched sample labels about sccoda HOT 8 CLOSED

Hi! Thanks for reporting this bug!

From what I can tell from your description, the issue arises due to the indices of data_scanpy_1.X and data_scanpy_1.obs being misaligned. This could happen in dat.from_scanpy because anndata does not check automatically if indices align. I just pushed a fix for this in the branch update_0.1.6. I hope that this solves the problem!

from sccoda.

Thanks. I'm working in Terra with scCODA installed via pip. How do I go about pulling the fix?

from sccoda.

You can install an exact branch from github with pip: pip install git+https://github.com/theislab/scCODA.git@update_0.1.6.

The version number of your scCODA installation should then be 0.1.6. If not, you can try uninstalling scCODA (pip uninstall sccoda) before installing with the same command as above.

from sccoda.

I was able to download it and I can see version 0.1.6 in my packages folder, but I am still getting the same mismatched output. Could something else be causing the problem?

from sccoda.

Hmm, then there seems to be another issue that causes this problem. Could you please help me pinning down the cause of it by checking whether the rows in your data_scanpy_1.X (cell counts) and in data_scanpy_1.obs (covariates) are correctly matched (e.g. does row 10 in data_scanpy_1.X contain lots of cell type 5 or is it row 20)?

Also, you could plot viz.stacked_barplot(data_scanpy_1, feature_name="samples",cmap=umap_cmap) to plot the data with the index of data_scanpy_1.obs as the x-axis labels, if you are more familiar with these labels than with the numbers 1-20.

Thanks in advance!

from sccoda.

viz.stacked_barplot(data_scanpy_1, feature_name="samples",cmap=umap_cmap)

data_scanpy_1.obs:

Now it seems like the samples are lined up correctly. Sample WR_4396_AS4 is the correct sample containing cell type 5. But now it looks like I can't put the data in the order I assign in cov_df. If I assign the samples numbers 1-20 based on the treatment, it mixes them up as such:

viz.stacked_barplot(data_scanpy_1, feature_name="Cond",cmap=umap_cmap)

from sccoda.

I'm glad that the samples and metadata are aligned now. This is the bug that I tried to fix with the new version.

The plotting function uses per default the same order as the data, which is the order in data_scanpy_1.X (the cov_df was shuffled in order to match the order of samples there). To order the x-axis in the plot, you can set the desired order in the level_order argument, in your case: viz.stacked_barplot(data_scanpy_1, feature_name="Cond",cmap=umap_cmap, level_order=['1','2','3','4','5','6','7','8','9','10','11','12','13','14','15','16','17','18','19','20']).

I hope that this gives you the desired plot!

from sccoda.

Awesome. Thank you. Everything is looking good now.

from sccoda.