recovering all genes for subsets

[notimenocall]

Hi,
Thanks for the great package!
I would like to perform scvelo on a subset. To do that, I clustered my data with highly variable genes and subset the cluster of interest. At this point, the vars only contain the hvgs, and so I usually do the following to recover all the genes for reclustering:
T_cells = adata[adata.obs.louvain.str.contains('T cells')] T_cells=sc.AnnData(T_cells.raw.X,obs=T_cells.obs,var=T_cells.raw.var)
However, this approach does not recover the 'layers' (spliced/unspliced). Is there an easy way to do this that I missed?
Thanks again!

[VolkerBergen]

Why not just T_cells = adata[adata.obs.louvain.str.contains('T cells')].copy() without the second command. Or is there a reason for creating a new anndata?

[notimenocall]

Thanks for your quick response!
Because the adata only contains hvg and I was trying to recover all the genes with the 2nd line from the T_cells.raw, so I can re-select hvg again just from the T cells cluster. However, this approach does not recover the 'layers' (spliced/unspliced).

[VolkerBergen]

.raw has no layers, because it is only meant to store .X for DE testing, so the layers are not restorable anymore for all genes at this point. You would have to re-read the data and then subset to the T_cells. The pre-computed (on hvg) cluster annotations, embeddings etc. can then be load into the anndata object.

[notimenocall]

Okay. It makes sense to me. I just want to make sure I didn't miss any magical function to could accomplish this. Thank you very much again!