Visualize quanTIseq fractions across molecular subtypes

[sjspielman]

This PR is continued from #1232.

Adds two visualizations that explore the cell type fractions estimated by quanTIseq across molecular subtypes from three different cancer groups: High-grade glioma astrocytoma, Ependymoma, Medulloblastoma. Unclassified subtypes and subtypes with <3 samples are not considered.

Two visualization options are included - the data is the same but the faceting is flipped. Note that one of the versions contains fractions of uncharacterized cell types, which doesn't fit (due to faceting scheme) in the other version. Points are currently colored by cancer group.

Is the analysis in a mature enough form that the resulting figure(s) and/or table(s) are ready for review?

Yes

Questions for reviewers

• Which plot, if either?
• Feedback on color scheme?
• Note that this code needs to be integrated into the 04 figures script.

Results

What types of results are included (e.g., table, figure)?

Two figures: version 1 and version 2

What is your summary of the results?

Reproducibility Checklist

• [x] 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.

Documentation Checklist

• [ ] This analysis module has a README and it is up to date.
• [ ] This analysis is recorded in the table in analyses/README.md and the entry is up to date.
• [ ] The analytical code is documented and contains comments.

[sjspielman]

@jharenza I've marked this as a draft PR since I imagine we'll have some back-and-forth about the figure conceptualization here. But, it's ready for you to have a look (just not "formal" review!).

[sjspielman]

Update - added one more plot as an option to consider. This barplot shows the fractions of characterized cell types, similar to the barplot I had made for molecular signature fractions.

[sjspielman]

I've updated this code based on some feedback from @jharenza and I feel this is ready for review now! Note that the code is now in a notebook inside the immune-deconv module for improved organization, so in the end (probably a separate PR) we can copy final figures over into the main figures/ directory.

• I solidified a version of the distributions of fractions across molecular subtypes for the four cancer groups of interest.
• I explored PDL1 expression, but it's important to note that there is mostly no variation in CD8+ expression levels in most groupings and CD8+ fractions are mostly around 0, except some of the MB samples.
• I added a figure showing PDL1 expression across the subtypes, faceted by cancer groups. Text is a little large here though, may need to shrink the whole figure down (?)

I suggest we proceed with survival analysis (again, note that given lack of variation in CD8+ scores, this is not likely going to be a useful predictor, but we'll find out!) in a subsequent PR and in its own notebook potentially in the survival-analysis/ module.

[sjspielman]

Noting that I also updated the analysis README to describe this notebook and its output.