Skip to content

Confirm label transfer approach for neuroblastoma #1152

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
sjspielman opened this issue Jun 2, 2025 · 2 comments
Open

Confirm label transfer approach for neuroblastoma #1152

sjspielman opened this issue Jun 2, 2025 · 2 comments
Assignees

Comments

@sjspielman
Copy link
Member

For #1151, we had been planning to use scArches to perform label transfer to annotate NB samples. The goal of this issue to confirm this approach:

  • Is there any reason to consider a different method? We know that scArches is reliably maintained, so there may not be, but it can't hurt to poke around a bit more in the latest-and-greatest literature.
  • What computational resources are we looking at, and do we need GPU? How do resource this affect environment setup, if at all?
    • For example, so we have a sense if LSfR could accommodate GPU needs (if relevant)or would we need more resources?
@sjspielman sjspielman self-assigned this Jun 2, 2025
@sjspielman
Copy link
Member Author

sjspielman commented Jun 5, 2025

Having spent some time reloading scArches and NBAtlas into my brain, the method indeed seems like the move. I'll look more into resource requirements coming up next, but first I want to note a couple initial items about NBAtlas that may be relevant to bear in mind:

  • NBAtlas integrated with scVI, but they do not make their scVI model available. They do make Seurat objects with raw and normalized counts, cell/batch metadata (including cell types!), and the scVI latent representation available, but that representation isn't the same as the model itself.
  • Either way, we'd want to use scANVI within the scArches framework, so we don't strictly need their scVI model. It might help us though to understand how it was created though:
    • Their code to build their scVI model is in this notebook
    • This appears to be their preprocessing data notebook to prepare each original dataset and export an anndata file for next steps, and that anndata file is slightly modified in this notebook. The output file from the second notebook matches the input file for the scVI notebook, so this seems right.

@sjspielman
Copy link
Member Author

Noting that this is the NBAtlas notebook where they ran scANVI/scArches to incorporate in-house data, starting from their scVI model. May be useful to keep a hold of: https://github.com/VIBTOBIlab/NBAtlas_manuscript/blob/90bba8a6ca023f34ba5dffb7875e9ebdaac2bdd5/scArches_SCANVI_NBAtlas_v2_manuscript.ipynb#L226

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

1 participant