Hello, I am planning to use decipher to batch correct my single cell embryo data across developmental stage, and i was wondering if the method requires annotation since i only have an adata.obs column for sample stage.
No, you dont. You can use it for plotting. @Justin_Hong any other thoughts?
can i use for batch correction across developmental stages? and how does it compare to other scVI tools in that regard?
Right now, the implementation in scvi-tools, doesnt support setup_anndata with batch_key, so that’s the main difference.
Ori is correct in that decipher is not intended to explicitly use a batch key to correct for batch correct. Instead, it relies on the data bottleneck to prioritize biological variation over batch variation. For a biological covariate like developmental stage I would recommend using a different model like scVI to correct these effects.
1 Like
Thank you for the response, but the assumption i got from decipher that it does take into account transcription change that would occur in different developmental stage. or am i missing something?
for scVI, i have different embryo some from the same developmental stage, each embryo part (e.g. Limb, Head) was sequenced as a different sample. would the following be correct as a setup:
SysVI.setup_anndata(
adata=adata,
batch_key=“embryo”,
categorical_covariate_keys=[“sample”],
)
for sysvi, it seems correct, although if it will not work you can also try to use a combination of all embryo-tissue couples as batch_key as they both have substantial batch effects.