When using schard::h5ad2seurat() in Windows 11 to read an AnnData object as a Seurat object, categorical variables with string values in the metadata are converted into numbers 0123456.

[cystone]

Would you please check out the issue.
Now I use sceasy to read h5ad in R temporarily.

h5ad_file = paste0('sce_qc.h5ad')
sceasy::convertFormat(h5ad_file, from="anndata", to="seurat",
outFile='tmp_sce_qc.rds'))

sce = schard::h5ad2seurat(h5ad_file)

sce0 = readRDS('sce_qc.rds')
[email protected] %>% head()

[iaaka]

Hi, thanks for reporting it. Could you please share input file (as small as possible), exact code you use and your [undesired] output?

[cystone]

I found that it woks well in scanpy workflow.
The necessary files for test were archived into a zip file.
tmp (2).zip
Thanks for your work.

[iaaka]

Hi @cystone, I've just pushed fixed version, could you please re-install and check whether it works for you?

[cystone]

I just reinstalled the package a few minutes ago, yet the issue persists.

devtools::install_github("cellgeni/schard", force = T)
sce = schard::h5ad2seurat('tmp_sce_qc.h5ad')
[email protected] %>% head()

doublet_score cellIndex datasetID sampleID cancerType tissueOrgan
AAACCTGAGATGTAAC-1 9.044023e-05 2 0 0 0 0
AAACCTGAGGACAGAA-1 1.687027e-02 7 0 0 0 0
AAACCTGAGGTGATTA-1 1.251088e-02 9 0 0 0 0
AAACCTGAGTCAAGCG-1 2.340808e-03 11 0 0 0 0
AAACCTGCAAGAGGCT-1 1.320272e-03 17 0 0 0 0
AAACCTGCACCGAAAG-1 1.726603e-04 20 0 0 0 0

I plan to delve into the source code of the package, as I believe it to be an exemplary tool for converting between Seurat and Scanpy.

[iaaka]

That is strange, I can confirm that schard was not working correctly with the file you send but it is fixed now and I cannot see how it can be platform-dependent. Just to make sure we are on same version, could you please share source code of schard::h5ad2data.frame for your installation?