PART 3: refactor multipanel plot step for dimension reduction

[jaclyn-taroni]

Purpose/implementation

This should be the last step in my refactor of the transcriptomic dimension reduction analysis module. Here I'm adding the capability to generate multipanel plots from the plot lists generated with scripts/get-plot-list.R.

Issue

#9

Directions for reviewers

Are this plots good enough for the moment?

Docker and continuous integration

Check all those that apply or remove this section if it is not applicable.

• The dependencies required to run the code in this pull request have been added to the project Dockerfile.
• This analysis has been added to continuous integration.

[jaclyn-taroni]

I should note that we still need to add script(s) for running everything and probably a README.

[cbethell]

LGTM 👍 It is definitely easier to read and the plotting is much more flexible.
I do have one question listed below, but beyond that, it looks ready to merge!

[cbethell]

In this file, why do you add the R scripts for the other transcriptomic analyses above, but the shell script for this PR?

[jaclyn-taroni]

This shell script can be run in CI for two main reasons: 1) the files it acts on are committed to the repository so when the code gets checked out they will be there 2) even if that was not the case it would run over any files in that directory.

[cbethell]

Ahh okay, gotcha. So in that case, would we want to keep the R scripts for the first 3 executable parts of this directory in the final collective shell script or would we use each of the individual shell scripts?

[jaclyn-taroni]

I would expect the shell script that gets implemented specifically for running this in CI would execute the following:

• A subset by strategy step
• The dimension reduction steps, but t-SNE should be skipped using the --skip_tsne flag for the polyA subset files because there probably aren't enough of them and perplexity needs to be lowered when t-SNE is not skipped.
• Generating plot lists
• The multipanel plot list

It's the second step that diverges from how you would 'normally' run the pipeline for analysis.