---
title: "A.4 -- Survival and TCGA"
author:
  Martin Morgan <Martin.Morgan@RoswellPark.org><br/>
date: "11 - 12 September 2017"
output:
  BiocStyle::html_document:
    toc: true
    toc_depth: 2
vignette: >
  % \VignetteIndexEntry{A.4 -- Survival and TCGA}
  % \VignetteEngine{knitr::rmarkdown}
---

```{r style, echo = FALSE, results = 'asis'}
knitr::opts_chunk$set(
    eval=as.logical(Sys.getenv("KNITR_EVAL", "TRUE")),
    cache=as.logical(Sys.getenv("KNITR_CACHE", "TRUE"))
)
suppressPackageStartupMessages({
    library(tidyverse)
    library(broom)
    library(cgdsr)
    library(survival)
    library(survminer)
})
```

# Getting the data

## Discovery

```{r}
library(tidyverse)
library(cgdsr)
url <- "http://www.cbioportal.org/public-portal/"
cbiop <- CGDS(url)

studies <- getCancerStudies(cbiop) %>% as.tibble()
## View(studies)
study <- "skcm_tcga"

cases <- getCaseLists(cbiop, study) %>% as.tibble()
## View(cases)
allcases <- "skcm_tcga_all"
mRNAcases <- "skcm_tcga_rna_seq_v2_mrna"

profiles <- getGeneticProfiles(cbiop, study) %>% as.tibble()
## View(profiles)
profile = "skcm_tcga_rna_seq_v2_mrna_median_Zscores"

genes <- c('CD63', 'CD9', 'CD81')
```

## Download

```{r}
mRNA <- getProfileData(cbiop, genes, profile, mRNAcases) %>% 
    rownames_to_column("id") %>% as.tibble()

clinical <- getClinicalData(cbiop, allcases) %>%
    rownames_to_column("id") %>% as.tibble()
```

## Cleaning

```{r}
data <- full_join(clinical, mRNA)

clean <- data %>% 
    filter(!is.na(CD63), !is.na(CD9), !is.na(CD81)) %>%
    select(
        id, SAMPLE_TYPE, 
        OS_STATUS, OS_MONTHS, 
        DFS_STATUS, DFS_MONTHS,
        CD63, CD9, CD81
    ) %>%
    mutate(
        SAMPLE_TYPE = factor(SAMPLE_TYPE),
        OS_STATUS = factor(OS_STATUS),
        DFS_STATUS = factor(DFS_STATUS)
    )

clean %>% summary()
```

# Analysis

## Testing continuous variables

```{r}
library(survival)
library(broom)   # tidy()
surv <- clean %>% with(Surv(OS_MONTHS, OS_STATUS == "DECEASED"))
fits <- bind_rows(
    clean %>% coxph(surv ~ CD63, data = .) %>% tidy(),
    clean %>% coxph(surv ~ CD9, data = .)  %>% tidy(),
    clean %>% coxph(surv ~ CD81, data = .) %>% tidy()
) %>% as.tibble()
fits
```

## Visualization as Kaplan-Meier curves

```{r}
library(survminer)

cutpoints <- clean %>% 
    mutate(OS_STATUS = OS_STATUS == "DECEASED") %>% 
    surv_cutpoint("OS_MONTHS", "OS_STATUS", genes)
cats <- cutpoints %>% surv_categorize() %>% as.tibble()

surv <- cats %>% with(Surv(OS_MONTHS, OS_STATUS))
fit <- survfit(surv ~ CD81, data = cats)
ggsurvplot(fit, conf.int=TRUE)
```