V23 manuscript run

[sjspielman]

I am opening this PR as a Draft as it is, well, a lot...and perhaps it needs to be split up.

This PR represents a local Big Run - RUN_LOCAL=1 bash scripts/run-manuscript-analyses.sh, and there are rather a lot of diffs. I have tried to catalog how I handled everything in this spreadsheet: https://docs.google.com/spreadsheets/d/1xRGlPYzfMcGnLBIMGTJuHJ5LyCyOSuIOKcp_GyGKMnU/edit?usp=sharing
Within this spreadsheet, I indicate for each file that has a diff whether I indeed committed it or just checked it out (so it won't appear in this PR as a diff!) because it was already reasonably up-to-date. I made the decision to generally commit binary files even if there were no apparent diffs due to file metadata which will likely have changed to better reflect that the file was generated with V23.
I've highlighted a couple rows in the spreadsheet that I think are worth definitely checking.

There is also one easy thing to review here! I updated to V23 in figures/generate-figures.sh. The rest is less easy but I hope the spreadsheet will help, and/or a new strategy for merging all this in.

From this endeavor, I identified that at least these figure panels will need to be updated for resubmission, and associated figures recompiled -

• Figure 3
  • interaction panels A-B
  • chromothripsis panels C-D)
• Fig 2 - all oncoprints
• Fig S5A-D oncoprints
  • Note that S5 compilatio needs to be updated to say "Ependymoma" as well anyways (Revision: Replace "Ependymal tumor(s)" --> "Ependymoma" in figures #1652)
• Survival plots
  • HGG KM curve (4H)
  • likely some of the survival forest plots (potentially 4F-H and 5D)

[jaclyn-taroni]

Here's my best guess as to what's going on with RN7SL2--IGF2:

• In the base histologies file (what is used to generate this fusion file prior to release), the following broad histologies are represented when you review the 5 BS IDs with a RN7SL2--IGF2 fusion: Benign tumor, Ependymal tumor, Diffuse astrocytic and oligodendroglial tumor, Neuronal and mixed neuronal-glial tumor, Low-grade astrocytic tumor
• In the pbta-histologies.tsv file in the release, the broad histologies are: Ependymal tumor, Diffuse astrocytic and oligodendroglial tumor, Neuronal and mixed neuronal-glial tumor, Low-grade astrocytic tumor (one biospecimen [BS_DGTPWJ59] went from Benign tumor to Neuronal and mixed neuronal-glial tumor)
• In analyses/fusion_filtering/05-QC_putative_onco_fusion_distribution.Rmd, fusions are filtered out if they are in more than 4 histologies:

  OpenPBTA-analysis/analyses/fusion_filtering/05-QC_putative_onco_fusion_distribution.Rmd

  Lines 184 to 201 in 61dc81b

  	### get fusions that are found in more than 4 histologies
  	```{r get fusions that are found in more than 4 histologies }
  	# count number of fusions in putative oncogene annotated fused gene are in more than N (countHistology) histologies
  	FusionInNhist<-fusion_calls %>% dplyr::select(FusionName,broad_histology) %>% unique() %>% group_by(FusionName) %>% tally(name="count")
  	FusionInNhist<-FusionInNhist[FusionInNhist$count>countHistology,]
  	FusionInNhist
  	# plot broad_histologies that have Fusions that are potential false positives and found in multiple histologies
  	multiHistFusion<-fusion_calls %>% dplyr::filter(FusionName %in% FusionInNhist$FusionName) %>%
  	# we want to remove annots column since there is difference in arriba annotation (uniquely has duplication/translocation/deletion values) and StarFusion annotation which will be counted twice since they are not unique
  	dplyr::select(-annots) %>%
  	unique()
  	ggplot(multiHistFusion,aes(x=multiHistFusion$FusionName,fill=multiHistFusion$broad_histology))+geom_bar()+theme(axis.text.x = element_text(angle=90))+coord_flip()+ylab("count")+xlab("FusionName(Total)")
  	```

In summary: This fusion was found in more than 4 histologies in pbta-histologies-base.tsv and exactly 4 histologies in pbta-histologies.tsv due to the change in broad_histology value for one biospecimen.

[sjspielman]

Noting for my own reference that this is the line that filters out >4 histologies:

OpenPBTA-analysis/analyses/fusion_filtering/05-QC_putative_onco_fusion_distribution.Rmd

Line 208 in 61dc81b

putative_driver_fusions<-fusion_calls %>% dplyr::filter(!FusionName %in% FusionInNhist$FusionName) %>%

[jaclyn-taroni]

I went through the spreadsheet and took a look at the highlighted rows. I believe the changes in chromothripsis and interaction-plots are due to the changes in the independent specimen files. That is also likely true in the recurrently fused fusions files, as the stranded independent RNA-seq specimens file changed between v22 and v23. I have identified what I think is the cause of the addition of the RN7SL2--IGF2 lines in analyses/fusion_filtering/results/pbta-fusion-putative-oncogenic.tsv here: #1672 (comment)

All that to say – I don't believe anything here is concerning at this point.

[sjspielman]

Opening this up for "formal review" since things look mostly accounted for. Spelunk away!

[jaclyn-taroni]

I went through this fairly carefully yesterday (with particular emphasis on the highlighted modules), so I am going to go ahead and approve.

[sjspielman]

There are no code changes here that are part of CI, so this can be merged before checks.