search

AlexsLemonade / OpenPBTA-analysis

The analysis repository for the Open Pediatric Brain Tumor Atlas Project
Other
100 stars 67 forks source link

V22 imm deconv rerun and figure updates #1458

Closed jharenza closed 2 years ago

jharenza commented 2 years ago

Purpose/implementation Section

What scientific question is your analysis addressing?

This PR will update the immune deconvolution figures for the manuscript

What was your approach?

I looked at the cancer groups with N >15, as was done before (14 groups), but then went down to 12 as were in the original figure so not to have too many panels. I removed "Other HGG" and "Other LGG" in this instance because we can mention in text that they followed patterns similar to pilocytic and DMG, respectively, and thus were not plotted in this figure.

What GitHub issue does your pull request address?

1368

Directions for reviewers. Tell potential reviewers what kind of feedback you are soliciting.

Which areas should receive a particularly close look?

NA

Is there anything that you want to discuss further?

Currently a WIP blocked by updates needed in #1401, per this comment.

Interestingly, we did not have cranio subtypes in the CD8+/CD4+ ratio plot before and this subtype has the highest ratio! In our recent proteomics paper, the cranios were described as below, so I think this is something worth highlighting as well in the manuscriptl.

The Epithelial cluster, containing, as expected, only CP tumors, which originate from odontogenic epithelium, was characterized by upregulation of EMT, immune-related pathways, as well as CTLA4 and PD-1 molecules ([Figures 2](https://www.cell.com/cell/fulltext/S0092-8674(20)31451-3?_returnURL=https%3A%2F%2Flinkinghub.elsevier.com%2Fretrieve%2Fpii%2FS0092867420314513%3Fshowall%3Dtrue#gr2)A and 2B; [Table S2](https://www.cell.com/cms/10.1016/j.cell.2020.10.044/attachment/d644f606-3f5a-4897-9427-34829dbb95bf/mmc2)). Therefore, CP could potentially benefit from immune checkpoint therapy, as reported previously ([Coy et al., 2018](https://www.cell.com/cell/fulltext/S0092-8674(20)31451-3?_returnURL=https%3A%2F%2Flinkinghub.elsevier.com%2Fretrieve%2Fpii%2FS0092867420314513%3Fshowall%3Dtrue#bib37)).

Note: I did not yet copy figures over to the figures directory because I want to wait until immune-deconv is rerun and merged. However, this can be reviewed now.

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

Yes

Results

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

What is your summary of the results?

Reproducibility Checklist

Documentation Checklist

sjspielman commented 2 years ago

Noting this https://github.com/AlexsLemonade/OpenPBTA-analysis/issues/1474#issuecomment-1170405180 to ensure we get the right cancer groups in these plots.

jharenza commented 2 years ago

Update @sjspielman - based on our offline convo, I reran the whole module here.

I also changed the two supp figures containing molecular subtypes to black and white because: 1) the LGG subtypes are annotated as "LGG, X", but when lumping the subtypes together, they contain multiple LGG colors. You can see pdl1_distributions_plot as an example. I left it as is since we do not use it. The other option I was thinking was to just use the "Other LGG color" but technically it is not only "other", it is "all". I guess we can use the broad histology color for this as well, but will be a one-off update here. Just some thoughts to discuss. 2) we don't have legends being printed

sjspielman commented 2 years ago

Since this PR also re-runs the module itself (no results changed with v22!), this PR also closes #1502.

jharenza commented 2 years ago

@sjspielman you are such a plot whiz!

I think I got everything - ready for re-review!

jharenza commented 2 years ago

@sjspielman made the updates and re-copied figures over, see what you think!

jharenza commented 2 years ago

This looks good! I just have one stylistic discussion point. We've colored points by cancer_group_display, which leads to multiple colors in a single boxplot for the plots showing subtypes. That outcome is definitely not wrong, but potentially a little visually busy? I wonder whether we might want to use broad_histology colors instead of cancer_group_display? Honestly I can't decide!

Haha, that is why I went with B&W, but I don't think it looks terrible - it is only two cancer groups in the LGG subtyping which have different colors. I suppose the advantage of seeing both colors is that we do not see clusters of pilocytic at one end of the boxplot and other lgg at the other end. But, I have no strong opinion. Since this is in the supplement, I think it is also fine to be this way? Let me check out using broad histology.

jharenza commented 2 years ago

@sjspielman updated the two subtype plots to use broad hist hex - let me know what you think!

jharenza commented 2 years ago

This is so weird, I know I reviewed this again yesterday, but I don't see my review here at all! Sadly it looks like GitHub swallowed my comment into the ether :(

👎

Regarding quantiseq-cell_types-molecular_subtypes.pdf, I think one more quick change would help that has to do with the order of the legend vs the x-axis. Both of them are ordered alphabetically, and as a consequence there's a mismatch. The x-axis subtypes run... embryonal, stellar, hgg, ependymal, embryonal again, lgg, embryonal again. It would be nice if the x-axis could run in order of histologies, in the same order that matches the legend. I don't mind what the order is as long as it all matches. Let me know if I can help with this code at all! It's likely just a bunch of arranging and dealing with factors.

Updated!