Revision: Run and assess relevant survival models at threshold

[sjspielman]

To assess the affects of tumor purity on our results, we should re-run certain survival models, considering only samples that have tumor_fraction>= median tumor fraction (#1622).

Specifically, we should run the following models and compare results to original results with all tumors:

• Model shown in 4F: survival ~ is_lgg + resection + tp53 + extend
  • https://github.com/AlexsLemonade/OpenPBTA-manuscript/blob/b0fdf6f24fe9c021f22e8b7da348708f68877ee5/content/03.results.md?plain=1#L251-L254
• Model shown in 5D: survival~CD274 expr + resection + quantiseq_fraction
  • https://github.com/AlexsLemonade/OpenPBTA-manuscript/blob/b0fdf6f24fe9c021f22e8b7da348708f68877ee5/content/03.results.md?plain=1#L286
• (This is not a survival model, but related to 5D so it can go here as well): Results in 5E (CD274 expression ~ MB molecular subtype)
  • https://github.com/AlexsLemonade/OpenPBTA-manuscript/blob/b0fdf6f24fe9c021f22e8b7da348708f68877ee5/content/03.results.md?plain=1#L287-L289

[sjspielman]

I'll note that other modules that were reassessed at a tumor purity threshold were broadly consistent with original results that feed into this module, suggesting (but not meaning for sure!) that results might not be strongly influenced.