SCPCP000006_05_explore_COPYKAT

[maud-p]

Purpose/implementation Section

This PR is following the discussion from the PR#776.
It explores the results generated in PR#801.

Please link to the GitHub issue that this pull request addresses.

#790

What is the goal of this pull request?

We wanted to explore the copykat results generated in PR#801.

copykat results have been obtained with or without normal cells as reference, using either an euclidean or statistical (spearman) method for CNV heatmap clustering.
This impact the final decision made by copykat for each cell to be either aneuploid or diploid, and it is thus crucial to explore the results using the different methods.
For each of theselected samples, we explore the results in the notebooks 05_cnv_copykat_{distance_parameter}_exploration_{sample_id}.html.
These notebooks are inspired by the plots written for the Ewing Sarcoma analysis in 03-copykat.Rmd.

Briefly describe the general approach you took to achieve this goal.

I get inspiration from the Ewing Sarcoma analysis in 03-copykat.Rmd to quickly check the results of copykat.

If known, do you anticipate filing additional pull requests to complete this analysis module?

Yes, same kind of notebook for exploration of infercnv results.

What is your summary of the results?

Looking at copykat heatmaps, I can really see how the function annotated a cell as aneuploid or diploid when using spearman as a clustering distance. The results seems to be more meaningfull when using an euclidean distance (i.e. we see CNV in aneuploid predicted cells and not diploid, mostly).
For that reason, I think that using spearman clustering distance should be avoided.

Provide directions for reviewers

I am not confident about continuing with copykat.
I'll do the same kind of exploration of the infercnv results in another PR, maybe adding some additional umap of predicted_cell.types to have everything in the same notebook.

Author checklists

Analysis module and review

• This analysis module uses the analysis template and has the expected directory structure.
• The analysis module README.md has been updated to reflect code changes in this pull request.
• The analytical code is documented and contains comments.
• Any results and/or plots this code produces have been added to your S3 bucket for review.

Reproducibility checklist

• Code in this pull request has been added to the GitHub Action workflow that runs this module.
• The dependencies required to run the code in this pull request have been added to the analysis module Dockerfile.
• If applicable, the dependencies required to run the code in this pull request have been added to the analysis module conda environment.yml file.
• If applicable, R package dependencies required to run the code in this pull request have been added to the analysis module renv.lock file.

[sjspielman]

Well, I'm convinced - let's not use copyKAT 😄
I'm also really glad that you were able to adapt code from a different analysis module, too - it's a great way to work within the OpenScPCA project!

I left a few very small wording comments, but otherwise this seems good to go! Once you make the small changes I suggested, I will approve this and we can move onto looking at inferCNV.

[sjspielman]

Wow - does it really differ between runs? I'm surprised to hear this (but I believe you!!), because the copyKAT code itself actually hardcodes a random seed; I actually filed an issue about this last week: navinlabcode/copykat#121. So, I'd expect it to be very reproducible. This means that it must be calling something else that is somehow un-setting the seed.

Nothing we can do about it, it's just interesting to note...

[maud-p]

yes, I figured this out when running multiple times to save the heatmap not in jpeg but png format. I can send you examples tomorrow (I struggle with the vpn connexion now 😞, hopefully better tomorrow! )

[maud-p]

@sjspielman some example to illustrate.
For the sample -208, I show you the heatmap of CNV for copykat ran with an euclidean distance, no reference, two days apart (to then save the heatmap in png instead of jpeg:

This is not completly different but not identical.

[maud-p]

for sample -205

[maud-p]

for -184 it is much more striking ❓❗

[sjspielman]

@maud-p can you also regenerate the html notebooks since the Rmd has changed? Thanks!

[maud-p]

@maud-p can you also regenerate the html notebooks since the Rmd has changed? Thanks!

Sure, it is running! Sorry for missing this part!

[sjspielman]

Thanks, this is ready to go! The GitHub Action failure is not something to worry about - or at least, it wasn't caused by your code, so it doesn't prevent this from moving forward.

[sjspielman]

Hm actually, we do need to try and get that check to pass, which is a matter of "re-triggering" the github action. Can you make a commit with a small "meaningless" change, like something very small (even just a space) somewhere in the README file? That will hopefully get the action to work.

[sjspielman]

@jashapiro can you have a look at what's up with this pre-commit action (I don't have permissions)?This PR is otherwise approved.

[jashapiro]

@jashapiro can you have a look at what's up with this pre-commit action (I don't have permissions)?This PR is otherwise approved.

We seem to be timing out on cloning the repo. This could be a more significant problem than just this PR: There is a time limit of 30s for cloning and we seem to be taking ~25 seconds in other PRs, meaning we are probably quite close to the repo size limit for the service.

[sjspielman]

@maud-p We're going to have to look into this action failure on our end a bit more. Until then, you can feel free to file the next PR looking at infer CNV.

[maud-p]

Wow I was out of office few hours and problem already solved! Thank you so much @sjspielman @jashapiro 🚀