ROC_curves.Rmd

---
title: "Roc Curves"
output: html_notebook
---

 

```{r}
# load tidyverse and tidymodels packages
library(tidyverse)
library(broom)
library(yardstick)

# load cowplot to change plot theme
library(cowplot)
library(ROCR)
library(pROC)
```

```{r}
#Change character variables to columns
predictions_xgboost[sapply(predictions_xgboost, is.character)] <- lapply(predictions_xgboost[sapply(predictions_xgboost,is.character)], as.factor)



predictions_xgboost %>%
  group_by(model) %>% 
  roc_curve(status, .pred_aggressive_all) %>%
    ggplot(
    aes(
      x = 1 - specificity, 
      y = sensitivity, 
      color = model)
  ) + # plot with ROC curves for each model
  geom_line(size = 1) +
  geom_abline(slope = 1, intercept = 0, size = 0.4) +
  scale_color_manual(values = c("#48466D", "#3D84A8","red","green","orange","pink","black")) +
  coord_fixed() +
  theme_cowplot()
  
 
```



#plot ROC curves with pROC

```{r}

roc1<- predictions_xgboost %>% 
  filter(model=="all") %>% 
  roc(status, .pred_aggressive_all)

roc2<- predictions_xgboost %>% 
  filter(model=="clinical") %>% 
  roc(status, .pred_aggressive_all)

roc3 <- predictions_xgboost %>% 
  filter(model=="clin_ngs") %>% 
  roc(status, .pred_aggressive_all)

roc4 <- predictions_xgboost %>% 
  filter(model=="clin_cyto") %>% 
  roc(status, .pred_aggressive_all)

roc5 <- predictions_xgboost %>% 
  filter(model=="ngs") %>% 
  roc(status, .pred_aggressive_all)

roc6 <- predictions_xgboost %>% 
  filter(model=="cyto") %>% 
  roc(status, .pred_aggressive_all)

roc7<- predictions_xgboost %>% 
  filter(model=="parsimony") %>% 
  roc(status, .pred_aggressive_all)

g2 <- ggroc(list("All Features"=roc1, "Clinical"=roc2, "Clinical + NGS"=roc3, "Clinical + Cytogenetic"=roc4, "NGS Only"=roc5, "Cytogenetic Only"=roc6, "Parsimonious Model"=roc7))+ theme_minimal() + geom_abline(slope=1, intercept=1, linetype="dashed", alpha=0.7, color="grey") + coord_equal() + labs(x="Specificity", y="Sensitivity", color="Model")


g2






```
#Plot ROC of GLM and Parimonious Model

```{r}
roc1<- predictions_xgboost_GLM %>% 
  filter(model=="all features") %>% 
  roc(status, .pred_aggressive)

roc2<- predictions_xgboost_GLM %>% 
  filter(model=="parsimony") %>% 
  roc(status, .pred_aggressive)

roc3 <- predictions_xgboost_GLM %>% 
  filter(model=="GLM") %>% 
  roc(status, .pred_aggressive)



g3 <- ggroc(list("All Features"=roc1, "Parsimonious Model"=roc2, GLM=roc3)) + theme_minimal() + geom_abline(slope=1, intercept=1, linetype="dashed", alpha=0.7, color="grey") + coord_equal() + labs(x="Specificity", y="Sensitivity", color="Model")

g3
```