Download the reference data for SingleR for stereopy

[synatkeamsk]

Dear Team,

It is me again and sorry for asking many question. Hope you do not mind. I have problem downloading the reference data for singleR in stereopy.
Manual download did not work. I downloaded via my HPC using
I downloaded this data using the below command in Bash Shell uisng this link. http://upload.dcs.cloud:8090/share/bb6fab82-2c16-46b2-a95e-6931338f31bf
wget "http://upload.dcs.cloud:8090/share/bb6fab82-2c16-46b2-a95e-6931338f31bf/download?path=%2FDemo_SingleR%2FReference%2FMouseRNAseqData.h5ad" -O MouseRNAseqData.h5ad. The file I got look like corrupted. Could you kindly help me guide to get the reference data to perform singleR in stereopy my mouse tumor. Now I could not do it, kindly help point me to the right reference file suitable to annotate my mouse tumors.

Kindly let me know whethe this is the correct dataset.

Thanks,

Synat

[wanruiwen-genomics-cn]

Dear @synatkeamsk,

Thank you for reaching out, and no need to apologize for asking questions!

Short answer: The download link you used is the correct one from the Stereopy tutorial, but the wget command needs a slight adjustment. Additionally, for mouse tumor annotation, you may want to consider alternative reference datasets beyond MouseRNAseqData.

---

Assessment

• Type: usage/question
• Severity: low
• Affected module: stereo/algorithm/single_r/single_r.py

What Is Happening

Stereopy's SingleR implementation does not include built-in reference data download functionality. Users must provide their own reference data as an .h5ad file. The Stereopy team has uploaded three reference files at the shared link you found:

1. BlueprintEncodeData.Rdata — human reference
2. MouseRNAseqData.Rdata — mouse reference (from R celldex package)
3. Mouse_brain_ref.anndata075.h5ad — mouse brain reference (already in h5ad format)

The file appearing corrupted is likely caused by the download URL encoding. The shared link serves an HTML download page, so wget may download the HTML page instead of the actual file.

Recommended Workflow

Step 1: Download the .Rdata file (not .h5ad) from the share link

The share link (http://upload.dcs.cloud:8090/share/bb6fab82-2c16-46b2-a95e-6931338f31bf) provides .Rdata files that need to be converted. Try downloading via a browser directly, or use the correct download path. Make sure the downloaded file size is reasonable (the .Rdata file should be several MB, not a few KB).

You can verify the file is not corrupted by checking:

file MouseRNAseqData.Rdata
# Should show: "data" or similar, NOT "HTML document"

If wget downloaded an HTML page instead, try:

# Use curl with redirect following
curl -L -o MouseRNAseqData.Rdata "http://upload.dcs.cloud:8090/share/bb6fab82-2c16-46b2-a95e-6931338f31bf/download?path=%2FDemo_SingleR%2FReference%2FMouseRNAseqData.Rdata"

Step 2: Convert .Rdata to .h5ad format

Stereopy provides an R script for this conversion. You need R with the anndata and argparser packages installed:

# Install R packages (if not already installed)
Rscript -e "install.packages(c('argparser')); BiocManager::install('SummarizedExperiment'); reticulate::py_install('anndata', pip=TRUE)"

# Convert
Rscript rda2h5ad.R --infile MouseRNAseqData.Rdata --outfile MouseRNAseqData.h5ad

The rda2h5ad.R script is available at: https://github.com/STOmics/Stereopy/blob/dev/docs/source/_static/rda2h5ad.R

Step 3: Use the reference in SingleR

import stereo as st

# Load your data and reference
data = st.io.read_h5ad("your_mouse_tumor_data.h5ad")
ref = st.io.read_h5ad("MouseRNAseqData.h5ad")

# Preprocess both
data.tl.normalize_total()
data.tl.log1p()
ref.tl.normalize_total()
ref.tl.log1p()

# Run SingleR
data.tl.single_r(
    ref_exp_data=ref,
    ref_use_col="label.main",  # This is the column name for MouseRNAseqData
    res_key="annotation"
)

Alternative Approaches

For mouse tumor annotation specifically:

• Option A: Use MouseRNAseqData — This is a general mouse reference from the R celldex package. It contains broad cell types (e.g., immune cells, fibroblasts, endothelial cells) and can be a good starting point. Use ref_use_col="label.main" for broad labels or ref_use_col="label.fine" for more specific subtypes.

• Option B: Use a tumor-specific reference — For better annotation of tumor microenvironment cells, consider downloading a more specialized scRNA-seq reference dataset (e.g., from Tabula Muris or a published tumor scRNA-seq atlas). Convert it to .h5ad format with cell type labels in an obs column, and use that as your reference.

• Option C: Get references directly from R celldex — If you have R installed, you can get any SingleR reference directly:

  # In R
  library(celldex)
  ref <- MouseRNAseqData()
  # Then convert using rda2h5ad.R or save with anndata

Useful References

Resource	Purpose
Stereopy SingleR Tutorial	Complete walkthrough with examples
R celldex references	All available SingleR references
rda2h5ad.R script	Format conversion tool

Notes

• The ref_use_col parameter depends on which reference you use. For MouseRNAseqData, use "label.main" or "label.fine". For the mouse brain reference, use "ClusterName".
• Make sure reference and query data share gene name conventions (both gene symbols or both gene IDs).
• If the shared download link (upload.dcs.cloud) is not accessible from your network, downloading references directly via the R celldex package and converting with rda2h5ad.R is a reliable alternative.

Please let us know if you have further questions!

Best regards,
Stereopy Maintainer