lots of integers in the python REPL during running spectral embedding

[beyondpie]

Hi Kai,

Today I used SnapATAC2 2.6 for spectral embedding, when I use the commands

a = snap.tl.spectral(brain_h3k27ac, features = 'selected', inplace = True)
# or 
snap.tl.spectral(brain_h3k27ac, features = 'selected', inplace = True)

in an interactive way, I noticed that lots of integers separated by comma are printed in the python REPL during running. Do you know what are they? Have you noticed this before? Is that possible to ignore the temporary log?

Thanks! Songpeng

[beyondpie]

This might because the scanpy AnnData format used directly for spectral embedding. I will check this later.

[beyondpie]

The full story is this:

1. We have a scanpy AnnData h5ad file. We use SnapATAC2 to read it and backed as 'r+'.
2. It works at first for feature selection (inplace = False).
3. Then we perform spectral embedding (inplace = True), the error is when saving spectral embedding results, the dimension does not match, i.e., number of cells does not match number of features.

I think the error is the how to data is saved (order of rows or columns) in h5ad between scanpy AnnData and SnapATAC2 AnnData.

Do you have any comments?

Thanks! Songpeng

[kaizhang]

Do you have a test data for me?

[beyondpie]

Later we figured it out. The original matrix is saved as CSC format, and after make it to CSR, it works.

[kaizhang]

I see. If you are using the nightly version, the error message should be much better.