pulling survival module from OpenPBTA

analyses/survival-analysis/README.md

@@ -0,0 +1,76 @@
+## Perform survival analysis for samples with 
+
+**Module authors:** Candace Savonen (ALSF CCDL), Run Jin (D3b), Stephanie J. Spielman (ALSF CCDL), Jo Lynne Rokita (D3b)
+
+In this module, we perform survival analyses by Kaplan-Meier, log-rank, and/or cox regression for samples in OpenPBTA using different covariates. 
+Detailed information about covariates can be found below.
+
+
+### Usage
+```sh
+bash run-survival.sh
+```
+
+### Module contents
+
+#### Function
+`util/survival_models.R` performs kaplan-meier, log rank, or cox regression for univariate and/or multivariate analyses.
+- The `survival_analysis` function performs survival modeling and returns a list with three objects:
+    1. The original model fit object
+    2. The summary table
+    3. The original dataframe
+- The `fit_save_model` function wraps the function `survival_analysis()` to perform and export a survival model
+
+
+#### Scripts
+
+`survival-analysis_template.Rmd` shows the usage of function `survival_analysis` to perform Kaplan-Meier, log-rank and cox regression survival analyses. 
+Particularly, it demonstrates runninng Kaplan-Meier and log-rank survival analysis on a categorical variable, using `germline_sex_estimate` as an example.
+It also demonstrates how to run cox regression survival analysis on a continuous variable, using `tmb` (tumor mutation burden) as an example.
+
+`survival-analysis_subtypes.Rmd` runs the following analyses:
+1. Univariate models
+- Kaplan-Meier survival analysis on HGG samples by `molecular_subtype`
+- Cox regression univariate analysis on HGG samples by `molecular_subtype`
+
+`survival-analysis_tp53_telomerase.Rmd` runs the following analysis:
+1. Univariate models 
+- extent of tumor resection (cox)
+- LGG broad histology (cox)
+- HGG broad histology (cox)
+- Cancer group (cox)
+- Cancer group (log rank)
+
+2. Multivariate models
+- Interaction model: `TP53 classifier score * telomerase score * extent of tumor resection * LGG group` across entire PBTA cohort
+- Additive model: `TP53 classifier score + telomerase score + extent of tumor resection + LGG group + HGG_group` across entire PBTA cohort
+- Additive models: `TP53 classifier score + telomerase score + extent of tumor resection` for each `cancer_group`
+
+3. Fit results are saved as `.RDS` files for each analysis
+
+`survival-analysis_immune.Rmd` runs the following analyses:
+1. Univariate models:
+- PD-L1 expression (stranded samples only)
+
+2. Multivariate models:
+- CD274 (PD-L1) expression, controlling for RNA library
+- `quanTIseq immune cell fractions + extent of tumor resection + LGG broad histology`, across entire PBTA cohort
+- `quanTIseq immune cell fractions + CD274 expression + extent of tumor resection + LGG broad histology` across entire PBTA cohort
+- `quanTIseq immune cell fractions + CD274 expression + extent of tumor resection` across suitable `cancer_groups` (DECEASED N>=3, thus no LGG)
+
+
+#### Output
+
+Within the `results` folder are folders containing survival model fit results for each analysis described above.
+```
+results/immune/*.RDS (`survival-analysis_immune.Rmd`)
+results/subtypes/*.RDS (`survival-analysis_subtypes.Rmd`)
+results/template/*.tsv (`survival-analysis_template.Rmd`)
+results/tp53_telomerase/*.RDS (`survival-analysis_tp53_telomerase.Rmd`)
+```
+
+The `plots` folder contains Kaplan-meier curve for the log rank analysis listed in parentheses below.
+```
+plots/KM_hgg_subtypes.pdf (`survival-analysis_subtypes.Rmd`)
+plots/survival_curve_gender.pdf (`survival-analysis_template.Rmd`)
+```

analyses/survival-analysis/results/template/cox_regression_tmb.tsv

@@ -0,0 +1,2 @@
+term	estimate	std.error	statistic	p.value	conf.low	conf.high
+tmb	0.005170351703977231	0.0020840890711969116	2.480868872369188	0.013106257732241913	0.0010856121838576593	0.009255091224096674

analyses/survival-analysis/results/template/logrank_gender.tsv

@@ -0,0 +1,4 @@
+germline_sex_estimate	N	obs	exp
+Female	987	270	276.4181799066573
+Male	1216	333	326.3838899517439
+Unknown	1	0	0.19793014159880817

analyses/survival-analysis/run_survival.sh

@@ -1,12 +1,20 @@
 #!/bin/bash
 
-# K S Gaonkar
+# Jo Lynne Rokita (D3b), Run Jin (D3b), and Stephanie Spielman (CCDL)
 
-# Run fusion_filtering
+# Run survival analysis 
 
 set -e
 set -o pipefail
 
+# Set the working directory to the directory of this file
+cd "$(dirname "${BASH_SOURCE[0]}")"
+
 # Plot survial curves of subtypes in HGG/DMG samples
-Rscript -e "rmarkdown::render('analyses/survival-analysis/survival-analysis_HGG_DMG.Rmd')"
+Rscript -e "rmarkdown::render('survival-analysis_subtypes.Rmd', clean = TRUE)"
+
+# Build survival models to assess tp53, telomerase, cancer group, and HGG group effects
+Rscript -e "rmarkdown::render('survival-analysis_tp53_telomerase.Rmd', clean = TRUE)"
 
+# Build survival models to assess molecular subtype, PDL1 expression, and immune cell fraction effects
+Rscript -e "rmarkdown::render('survival-analysis_immune.Rmd', clean = TRUE)"