ValueError: You need to run `pp.neighbors` first to compute a neighborhood graph.

[KaiLi-2324]

Hi, @dawe . Thanks for developing SCHIST. I got the following error "ValueError: You need to run pp.neighbors first to compute a neighborhood graph." when I was running the function scs.inference.fit_model on my spatial ATAC-seq data with this tutorial "Analysis of spatially resolved ATAC-seq data". The screenshot is shown below. Do you know the reason to this error? Thanks very much!

Best

[dawe]

Hello, you need to compute a graph before applying the model, e.g. by sc.pp.neighbors(adata) where adata is substituted with your data object (sdata.table?).

I see that there's an error on the instruction page and the command is missing.

[KaiLi-2324]

Hi @dawe . Thanks for the prompt reply. I ran sc.pp.neighbors on sdata.table, and I got the following error "ValueError: A must be of floating or complex floating data type."

[dawe]

Hello @KaiLi-2324, sorry for that. I have fixed the docs page but readthedocs is failing in publishing the correct commands. If you run sc.pp.neighbors() on AnnData it defaults to adata.obsm['X_pca'] for the computation of kNN graph. In your case you may want to specify the proper embedding (X_spectral), so:

basis='spectral'
sc.pp.neighbors(spdata.table, key_added=f'{basis}_neighbors', use_rep=f'X_{basis}')
scs.inference.fit_model(spdata.table, neighbors_key=f'{basis}_neighbors')

The error in your command is raised by the fact PCA needs floating point numbers and your adata.X count matrix is integer type.

[KaiLi-2324]

Hi @dawe . Thanks for the reply. I got the following error "AttributeError: 'SpatialData' object has no attribute 'pl'" when I was running spdata.pl.render_images().pl.render_shapes(color='nsbm_level_1', fill_alpha=.7).pl.show(title='Level 1', colorbar=True). The error is shown below

[dawe]

Could this be an issue with spatialdata instead? Please refer to their manual (and issue tracker) to make this work. It comes to my mind I may have had some issues in plotting with spatialdata after updating to recent version, for this reason I've moved to squidpy. If you have it installed you can import it (like import squidpy as sq) and create an AnnData object from the spdata table you are interested in. Then you can add the coordinate for plotting with something like

adata.obsm['spatial'] = sdata.tables['table'].obsm['spatial'] # just copy the coordinates
spatial_key = "spatial" # this is needed by squidpy
adata.uns[spatial_key] = {library_id: {}} # this as well
sq.pl.spatial_scatter(adata, color='nsbm_level_1', 
                     img=False, 
                     size=.5)

Of course, you have to read squidpy manual to make it work with the desired layout.

[dawe]

I realized the documentation had errors and missing commands (due to wrong writing of markdown). Fixed in 443ab46