Improve scanpy compatibility #774
Conversation
Sync changes from `development` into `main`
jashapiro
changed the title
bin/reformat_anndata.py
Outdated
| adata.var["highly_variable"] = adata.var.gene_ids.isin( | ||
| adata.uns["highly_variable_genes"] | ||
| ) |
There was a problem hiding this comment.
I think you might want to only do this if args.mask_var != "None"? And does this fail if adata.uns["highly_variable_genes"] does not exist? If so, we should add a check for that.
There was a problem hiding this comment.
Yes, you are correct, as I realized when looking at the first comment!
| move_counts_anndata.py --anndata_file ${rna_h5ad_file} | ||
| ${has_adt ? "move_counts_anndata.py --anndata_file ${feature_h5ad_file}" : ''} | ||
| reformat_anndata.py --anndata_file ${rna_h5ad_file} | ||
| ${has_adt ? "reformat_anndata.py --anndata_file ${feature_h5ad_file}" : ''} |
There was a problem hiding this comment.
Do you want to include the PCA information in these objects too like you do for the processed individual objects? I believe we do re-calculate PCs after merging.
There was a problem hiding this comment.
I wasn't planning to change anything for the merged files. We don't move the reduced dims to X_PCA and X_UMAP for those, do we?
There was a problem hiding this comment.
We do move them in those objects.
Line 70 in f215046
I can see both arguments, but I think for consistency between objects we would want to do this for merged objects?
There was a problem hiding this comment.
I agree that we should make the names match. I think I don't want to add the .uns["pca"] object for merged data though, since we use a different function for the PCA calculation for merged data that I am not sure would be equivalent to what ScanPy would do. So I think probably better to just leave that off?
There was a problem hiding this comment.
I think that makes sense. So for the merged object the only change should be the lower case naming.
Co-authored-by: Ally Hawkins <54039191+allyhawkins@users.noreply.github.com>
|
Test run here completed successfully, and at first examination the outputs are as expected. |
Closes #773
I'm filing this as a draft because I have not yet really tested it, but I wanted to get something up while it was fresh in my mind.
The goal here is to improve the compatibility with scanpy as described in #773, so I have done 4 main things:
X_pcaandX_umapto match scanpy.highly_variablecolumn to thevartableunsThe last step involves exporting the variance explained data and then importing that separately, with the assumption that the PCA was centered and highly variable genes were used. I could (and probably should) make those assumptions into arguments for the script just to be safe and maybe a bit future-proof.
There are probably also a few places where I am making other assumptions that I should check more explicitly. As I said, a draft...
I also renamed the script to be a bit more generally named, but annoyingly there were apparently enough changes that GitHub is not displaying it as a rename but as a deletion and recreation. Probably because it was run through a code formatter automatically.