search

AlexsLemonade / OpenPBTA-analysis

The analysis repository for the Open Pediatric Brain Tumor Atlas Project
Other
99 stars 66 forks source link

OncoPrint figures #1200

Closed jaclyn-taroni closed 2 years ago

jaclyn-taroni commented 2 years ago

⚠️ Stacked on #1193

Addressing #1177 and #1178, but made slightly different choices than what is detailed in those issues. Importantly, I've only removed cancer groups with less than 10 samples for the other CNS plot. In the current version of #1193, cancer groups with n < 10 in the other 3 broad histologies are randomly assigned a hex code from the Greys color brewer palette. There are over 20 cancer groups with n < 10 that'd be included in the Other CNS plot (fewer than that when you consider which groups actually show up in the plots using the other CNS genes of interest list), so it's not feasible to randomly assign greys to them in the same way. Instead, the Other CNS cancer groups with n < 10 are in a supplemental PDF where they are assigned colors based on the Paired color brewer palette.

Similar to #1194, I've created an Rscript (figures/scripts/fig2-oncoprint-landscape.R) that creates individual panels & legends as PDFs (figures/pdfs/fig2/panels) and then I've assembled a draft figure in Illustrator. You can see that draft as figures/pngs/figure2.png. I've removed the old script that used multipanelfigure to assemble the figure from the PNGs in the module (figures/scripts/oncoprint-landscape.R).

I needed to rerun the oncoprint-landscape module because we had been manually updating the OncoPrint palette, rather than updating the script that has all the palettes, so some colors got dropped in #1161 when the tumor descriptor palette was created and those colors were missing when things were run as part of #1189. I've also deprecated the former figure 2 script to hopefully avoid confusion.

jaclyn-taroni commented 2 years ago

This is a large number of changed files, but everything in analyses/oncoprint-landscape/plots is just a rerun now that I've fixed the oncoprint palette and I'm removing a good number of figures that are deprecated.

jaclyn-taroni commented 2 years ago

@jharenza I've updated the color palette and rerun the figures. Can you take a look before I invest any more effort, i.e., attempt to put together the multipanel version?

jharenza commented 2 years ago

@jharenza I've updated the color palette and rerun the figures. Can you take a look before I invest any more effort, i.e., attempt to put together the multipanel version?

The colors I mentioned look fine, but what is "Complex_event"? I hadn't seen that before and it is black, along with nonsense mutations, so that would need a new color if we use it. I would think it would fit into other categories, unless it is an SV of some sort?

jaclyn-taroni commented 2 years ago

but what is "Complex_event"? I hadn't seen that before and it is black, along with nonsense mutations, so that would need a new color if we use it. I would think it would fit into other categories, unless it is an SV of some sort?

A couple notes:

jaclyn-taroni commented 2 years ago

@jharenza ready for another look

jharenza commented 2 years ago

Complex event isn't documented as far as I can tell outside of: https://github.com/PoisonAlien/maftools/blob/db66c013049e04999eca8e6aaa440cfe1eb08703/R/oncoplot.R#L644 which I take to mean that we expect to see it when a gene is impacted by both by a mutation and a copy number alteration of some kind

Interesting, I had not noticed that before and I wonder if that was earlier code before the CNVs were added as smaller rectangles in which we are able to see a mutation/fusion + a CNV with two different colors. That begs the question why is it showing as complex rather than the two colors for mut/cnv, unless perhaps a multi-hit SNV + CNV? I can't look into that right now, but probably can later tonight. If you get to it before me, let me know.

jaclyn-taroni commented 2 years ago

From a cursory look at the code, I think it might arise when there are multiple CNVs. If we set writeMatrix = TRUE when we call oncoplot(), that might help confirm/reject that idea.

jharenza commented 2 years ago

From a cursory look at the code, I think it might arise when there are multiple CNVs. If we set writeMatrix = TRUE when we call oncoplot(), that might help confirm/reject that idea.

Just noting that we should explain this in the figure legend.

I can't distinguish between the green for complex event and missense, but otherwise, the flank/amp colors are ok for me.

jaclyn-taroni commented 2 years ago

Going to request a review from @cbethell! The main file to take a look at is: figures/scripts/fig2-oncoprint-landscape.R. We can add the compiled figures later.