no extra anndata

[flying-sheep]

This still needs some checking, I did this mostly blindly (where I didn’t you see that I modified the text)

I also didn’t verify that setting non-HVGs to nan and then doing regression is

1. necessary, or
2. yields the expected results

once we know, maybe we can e.g. add a mask to regress_out.

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[ilan-gold]

necessary

What do you mean by this? It seems necessary in so far as the results match and means we don't need a separate object.

[review-notebook-app]

View / edit / reply to this conversation on ReviewNB

ilan-gold commented on 2025-06-17T08:02:41Z
----------------------------------------------------------------

Would rename the layer something like "scaled_hvg" just to make it crystal clear what's going on

[flying-sheep]

It seems necessary in so far as the results match and means we don't need a separate object.

I meant that I didn't actually check if the regression yields different results without the kludgy "masking" I'm doing here.

I'll do that!

[ilan-gold]

I'll do that!

Ok, can't hurt, But visually the results look the same so I was going on that. Good on you for being thorough haha :)

[flying-sheep]

OK, setting the nans is not actually necessary.

and with rank_genes_groups' parameter mask_var, this should work perfectly!

[review-notebook-app]

View / edit / reply to this conversation on ReviewNB

ilan-gold commented on 2025-06-17T11:22:36Z
----------------------------------------------------------------

Why use the mask here and below, but not above? I would think they should all be the same