Closed Jieran-S closed 9 months ago
I think it makes sense to update Banksy to preprocess as they recommend. I have changed the script accordingly.
Thinking about it, I also found a way to use the existing preprocessing workflow in bass - the script has been updated accordingly
For now we can leave it as implemented in this PR. In the future we might want to try out library size normalization as a preprocessing step in general anyway.
Since our transform data option is default log1p transformation. I added an additional scaling step for GraphST and SCAN-IT. Using parameters in their source codes.
GraphST: https://github.com/JinmiaoChenLab/GraphST/blob/main/GraphST/preprocess.py SCAN-IT: https://github.com/zcang/SCAN-IT/blob/main/examples/Slide-seq/scanit.ipynb
For reviewers:
These are the only 2 methods I found using scaled data. Is there any other method I miss?
BANKSY requires a lib-norm (not log-norm) preprocessing. What should we do about that?