Making AnnDataSet changes reflect in individual sample files

I have 16 samples I independently performed QC for. I have .h5ad files for each of these.

For these samples, I made an AnnDataSet, "all_samples.h5ad"and read into the variable data. I performed UMAP embedding, Harmony integration, generated gene matrix and MAGIC imputed for my genes of interest.

Based on these genes I want to apply cell type labels to each cell. But trying something like this doesn't work:

data.obs["cell_type"] = data.obs["leiden_1.5"].map(leiden_to_cell_type)

where "leiden_to_cell_type" maps leiden cluster to cell type annotation.

How do I make this work for the AnnDataSet? And how can I make new columns added to the AnnDataSet reflect in the individual .h5ad files that the AnnDataSet is referencing?

Thanks.

Start by reproducing the data.obs assignment on the AnnDataSet loaded from all_samples.h5ad, then inspect how its referenced individual .h5ad files handle observation-column updates. Determine whether propagation is supported or needs a documented workflow; done means cell-type columns can be added as requested and the result is clear for each sample file.