add extent tumor resection/lgg to tp53/telomerase survival analysis

[jharenza]

Purpose/implementation Section

What scientific question is your analysis addressing?

After discussion with @adamcresnick, I am adding to the survival analysis two additional terms: extent of tumor resection and LGG group. This is to make sure our results are not strictly due to tumors having been fully resected or belonging to LGG group.

What was your approach?

I add in extent of tumor resection and LGG group for univariate analysis. I also add these terms to the multivariate analyses.

What GitHub issue does your pull request address?

NA

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?

We will want to create a new forest plot for this result. Update: done in #1343

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?

• Extent of tumor resection: gross/total and LGG group: LGG are not surprisingly very strong positive predictors of overall survival.

• TP53 and telomerase scores alone are each significant with or without LGG samples or samples with tumors fully resected. Note: the p-values did get larger when removing these groups, suggesting it is good we are adding in these terms! TP53 score p went from e-13 to e-8 and telomerase score p went from e-17 to e-8.

• HGG group is a very strong negative prognostic indicator of survival (p=5e-35).

• LGG group is a strong positive predictor of survival (p=1e-4).

• Interaction models are still not significant/only marginally significant, thus additive are the way to go.

tp53_score+telomerase_score+extent_of_tumor_resection+lgg_group

• The strongest predictor here is extent of tumor resection with gross/near total resection having a survival benefit and partial resection nearing significance for a survival benefit. High TP53 scores, telomerase scores, and non-LGG are all significant risk factors for poorer overall survival.

tp53_score+telomerase_score+extent_of_tumor_resection+hgg_group

• HGG group:HGG and high telomerase scores are significant risk factors for poorer overall survival. Extent of tumor resection: gross/near total is a survival advantage. High TP53 scores trend toward a significant risk for poorer overall survival.

Additive models: tp53_score+telomerase_score+extent_of_tumor_resection for each cancer group

• Same results as before

Reproducibility Checklist

• [x] The dependencies required to run the code in this pull request have been added to the project Dockerfile.
• [x] This analysis has been added to continuous integration.

Documentation Checklist

• [ ] This analysis module has a README and it is up to date.
• [ ] This analysis is recorded in the table in analyses/README.md and the entry is up to date.
• [x] The analytical code is documented and contains comments.

[jharenza]

I've left some comments about some of the models here, since I'm not sure what exactly they are contributing. More comments and explanations would be helpful, or some of these models can be removed.

In addition, I'm not sure why there are separate models, tp53_score+telomerase_score+extent_of_tumor_resection+hgg_group and tp53_score+telomerase_score+extent_of_tumor_resection+lgg_group. Is there a reason not to just do: tp53_score+telomerase_score+extent_of_tumor_resection+hgg_group + lgg_group?

I actually didn't think about combining LGG and HGG groups, but I did just make the change. I was going back and forth about this because I was thinking that they tell a slightly different story re: tp53 (everything else being similar in separate vs combined analyses).

• The tp53 scores are not a significant predictor when you consider HGG_group, but they are when you consider LGG_group - I think this makes sense because LGGs already have such a good prognosis, so adding in a dysregulated tumor suppressor affects survival, whereas samples being grade 3 or 4 gives more weight to prognosis than tp53 scores, since there is likely much more complex biology at play (and potentially brain region/inability to resect, among others). That is how I interpreted, but is that a correct thinking?
• We also see that tp53 scores are a significant negative predictor in DMGs in #1332 (and at the end of this notebook), which is to say that in HGGs, this seems to be perhaps subtype-specific and maybe in non-HGGs this is more generalized, but maybe I am over-interpreting.

I am ok doing one model, cleaning this up, and updating to use one FP for the main figure, so I did that in the latest commit.

Also not sure why there is a diff for results/TableS3-RNA-results-table.xlsx ?

I kept seeing this, too, so must have accidentally added it somewhere, so removed here.