Mixscape classification

[NguyenLuongHa]

Report

While further exploring the pipeline, I have encountered a couple of additional questions:

1. When separating KO and NP, I noticed that in typical practices, results are iteratively updated until convergence is achieved.
   However, your approach seems to focus on finding a consistent result between two consecutive iterations. Could you kindly clarify this difference?
   (Please see the attachment for further details.)

2. In the tutorial document, the code snippet:

    ms.mixscape(adata=mdata["rna"], control="NT", labels="gene_target", layer="X_pert")  

Should it include the "split_by" parameter?

Version information

No response

[NguyenLuongHa]

@Lilly-May
In pull request #683, you're using random_state for the GaussianMixture object. This suggests that the while loop might be intended to be iterative, updating its behavior with each iteration.

Additionally, within the while loop, the guide_cells variable should likely use the new_class_name instead of the original labels. This would make the loop operate on the newly assigned class names

[Lilly-May]

Hi @NguyenLuongHa, thanks for opening this issue. I'll look into this ASAP!