Move immune-deconv figure code

[sjspielman]

Inspired by my PRs #1569 and #1570, I got a little excitable and also moved out the immune-deconv panel code for easier updating moving forward!

Along the way I made a couple figure updates including using shape=16 to make Illustrator happier with transparent points (thanks again for the tip @envest!) and re-sizing some panels that are compiled with Graphic. I also updated the CD274 expression axis label to match the same format that we have in figure S5B-C (all GENE log2(FPKM+1)). I recompiled accordingly!

Again, the generate-figures.sh script will need to be updated and that's coming up next!

[sjspielman]

@jharenza I've updated this as follows...

• Some general code re-organization to remove unused/unneeded variables
• For quantiseq-cell_types-molecular_subtypes.pdf, I've now included the legend embedded
• Turns out there is no EPN in the cd8/cd4 plot because there are no EPN ratios to show! They are all infinite or NaN, except for unclassified EPN samples which would not show up in the plot anyways. But I have ordered this plot in the same order as quantiseq-cell_types-molecular_subtypes.pdf. I don't have a legend in there since colors are the same, but can easily add one in!

[sjspielman]

For the cancer groups plot in Figure 5, I think we can have a slightly larger text in the facet strips?

I entirely agree!! But whenever I bump this up from the current value, it gets too big right away! It has been mystifying me. Maybe I'll try using rel() instead of setting an exact size.

Also, for the CD8/4 ratio plot, it looks like the ETMR (part of embryonal) is out of order of the broad histology.

Ok, will check this. Edit - Woops, ETMR was missing from these plots!

[sjspielman]

@jharenza more updates -

• I tried fighting more with the facet text size, and it just won't budge. Feel free to five it a go if you want! I might have to do it manually in Illustrator?
• I realized I was missing a few N >= 3 subtypes for those histologies, so I ended up removing my hardcoding to programmatically get the right subtypes and order. Are there any new subtypes that got added in from this code change we should remove now though?
• I also changed the boxplot color in the MB expression to a darker gray, since it was a bit harder to see the current gray with the shape=16 addition, which again helps avoid alpha weirdness during compiling.

[jharenza]

I tried fighting more with the facet text size, and it just won't budge. Feel free to five it a go if you want! I might have to do it manually in Illustrator?

AHhh, I tried a bunch of combinations of wrapping the text after fewer characters and then increasing the strip text size, but see what you mean. Some names are too long and if you wrap too soon, you get 3 rows for DNET and then you go in circles.

I realized I was missing a few N >= 3 subtypes for those histologies, so I ended up removing my hardcoding to programmatically get the right subtypes and order. Are there any new subtypes that got added in from this code change we should remove now though?

This looks good to me!

I also changed the boxplot color in the MB expression to a darker gray, since it was a bit harder to see the current gray with the shape=16 addition, which again helps avoid alpha weirdness during compiling.

👍

[sjspielman]

This code does not run through CI, so can merge before checks.