Added S3 figure panels B, C, D

[sjspielman]

This PR address issue #1255 and creates three PDF panels B, C, and D to use in Figure S3.

Figure S3B is created by copying over the PDF from the relevant module. Therefore that module was also updated to export a PDF image, and its notebook was rerun to ensure v21 data.

Figures S3C-D were created directly in the script added here, where I used the simpsons color scale to color the jitter points for each # chromothripsis regions.

[sjspielman]

Note - I am now seeing that the intended FigS3B had some color-related png changes when I re-ran the notebook related to the palette. Investigating this now.

Edit: This notebook was using the deprecated display_group, so I am also updating the notebook to use cancer_group instead. As part of review @jaclyn-taroni, please advise if broad_histology is better?

[jaclyn-taroni]

Yea, I was just headed to chime in about the palette when I saw your edited comment @sjspielman. I think you'd want to use broad_histology_display, but perhaps drop the samples where broad_histology_display == "Other" because then we'd be lumping disparate diagnoses together and I'm not sure that's what we want to do here.

Edit: I would also consider if it's better to save some output from the module used to create the heatmap (e.g., some RDS file) and use the new & correct palette in your figures/ script vs. altering the module to use the new palette. I don't know how big of a lift that would be. I generally have avoided going back into modules where I can, even though it is what I would do if time constraints didn't exist.

[sjspielman]

@jaclyn-taroni So, actually, are we sure about cancer group? S2 apparently relates back to Figs 2-3, which focus on cancer groups?

✅ for your edit suggestion. "The less module the better", it is!

[jaclyn-taroni]

So, actually, are we sure about cancer group? S2 apparently relates back to Figs 2-3, which focus on cancer groups?

No, I'm not sure. I just thought broad_histology_display was more analogous to what was used previously. But if we're going "less module the better," then we probably will have a much easier time making that call/making tweaks. I'm fine with cancer_group being what we try first.

[sjspielman]

@jaclyn-taroni @jharenza This is ready for a re-review. The heatmap (Fig S3B) code is all in figures/scripts/ without touching the original module, and heatmap uses cancer groups, where "Other" is on the bottom.

The other chromothripsis S3C-D plots are still as they were.

[sjspielman]

I like the new cancer group figure for B, and I agree with @jaclyn-taroni on dropping the "Other" samples from this plot.

👍

but the original figure was set up to be colored as a gradient with the highest number of chromothripsis regions being the darkest.

I wasn't sure about this palette, actually. Here it describes we should only use the gradient palette for numeric data, not for groupings like this. @jaclyn-taroni , any thoughts?

[jaclyn-taroni]

Coloring the points in those plots seems a little like duplicating information to me? I suppose the color is somewhat helpful matching between the two panels, but there are not a ton of groups to begin with.

[jharenza]

Coloring the points in those plots seems a little like duplicating information to me? I suppose the color is somewhat helpful matching between the two panels, but there are not a ton of groups to begin with.

I am fine with B&W here, plus we don't have a color legend anyway