Stereo-Seq Analysis Pipeline

[TolgaDuz]

Hello together,

Due to the significant increase in the popularity of the stereo-seq platform, I highly recommend implementing wrapper functions for this type of analysis.

I am currently having some difficulty in understanding how to read and analyze it. Do you have any suggestions or ideas? The back-up solution would be using stereopy, which couldn't really convince me.

Cheers!

[alikhuseynov]

is there public example dataset output that can be used to make a reader function?

[DrKaiChen]

I am also looking forward to Seurat developing a function to load the stereo-seq from BGI similar to Visum and CosMx.

[dcollins15]

Thanks for the suggestion!

I can certainly see why a loader would be handy 👍

In the meantime, assuming that you're able to read your counts and coordinates in from disk, you should be able to do something like:

assay_name = "Stereo"
fov_name = "fov"

object <- CreateSeuratObject(counts, assay = assay_name)

fov <- CreateFOV(
  coordinates, 
  type = "centroids",
  assay = assay_name,
  key = Key(fov_name, quiet = TRUE)
)

fov <- fov[Cells(object)]

object[[fov_name]] <- fov

[DrKaiChen]

Hi David Collins,

Thank you for providing the information. I am not sure whether we should use the sequencing-based analysis method (e.g., 10 × Visum) or the image-based analysis method (e.g., 10 × Xenium or Nanostring CosMx SMI) using the Seurat. I have already obtained the SAW outputs (https://github.com/STOmics/SAW). Could you please provide some guidance on which method would be approprite on the Stereo-seq (bin50 or cellbin)? If you need any information from me, please let me know. Really appreciate it!

Thank you,
Kai

[alikhuseynov]

Hi David Collins,

Thank you for providing the information. I am not sure whether we should use the sequencing-based analysis method (e.g., 10 × Visum) or the image-based analysis method (e.g., 10 × Xenium or Nanostring CosMx SMI) using the Seurat. I have already obtained the SAW outputs (https://github.com/STOmics/SAW). Could you please provide some guidance on which method would be approprite on the Stereo-seq (bin50 or cellbin)? If you need any information from me, please let me know. Really appreciate it!

Thank you, Kai

For object construction, if using CreateFOV then it is either VisiumV2 class or image-based method. For the analysis itself, I would say since Stereo-seq is sequencing-based method the analysis should be as for VisiumHD.

[DrKaiChen]

Hi David Collins,
Thank you for providing the information. I am not sure whether we should use the sequencing-based analysis method (e.g., 10 × Visum) or the image-based analysis method (e.g., 10 × Xenium or Nanostring CosMx SMI) using the Seurat. I have already obtained the SAW outputs (https://github.com/STOmics/SAW). Could you please provide some guidance on which method would be approprite on the Stereo-seq (bin50 or cellbin)? If you need any information from me, please let me know. Really appreciate it!
Thank you, Kai

For object construction, if using CreateFOV then it is either VisiumV2 class or image-based method. For the analysis itself, I would say since Stereo-seq is sequencing-based method the analysis should be as for VisiumHD.

Yes, I agree with you. Stereo-seq should be similar to the Visum HD.

[DrKaiChen]

Any example of creating a Seurat object using Stereo-seq data would be very helpful.

[jeanne09jiang]

I also had a similar problem. So I put together a pipeline for analyzing Stereo-seq data using Seurat, which works similar to 10x Visium HD. Hope this helps.
https://github.com/jeanne09jiang/Stereo-seq-2-seurat