data_split.Rmd

---
title: "Untitled"
author: "Reuben"
date: "2024-12-03"
output: html_document
---

```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = TRUE)
```

# Clear environment
```{r}
rm(list = ls())
```

# Load patient data
```{r}
patient_data <- read.delim("ws3_grampian_patient_data.txt")
```

# Plot distribution of treatment responses
```{r}
library(ggplot2)
metadata <- as.data.frame(patient_data[,c(1,20)])
metadata <- metadata[-c(1,2),]

# Plot a barplot of the treatment response types
ggplot(metadata, aes(x = as.factor(Response.to.Treatment), fill = as.factor(Response.to.Treatment))) +
  geom_bar() +
  geom_text(stat = 'count', aes(label = ..count..), vjust = -0.5) +
  scale_fill_hue(c = 100) +
  theme(legend.position = "none")
```
# Split the data into two subsets - one for unsupervised MOFA models, one for supervised model
```{r}
library(caret)

# set a random seed for the data split
set.seed(123)

#split the data in half. Half the data for the unsupervised MOFA models and half for the supervised model.
# Create a balanced partition, with an equal number of each treatment type in each split
trainIndex <- createDataPartition(metadata$Response.to.Treatment, p = 0.5, list = FALSE, times = 1)

# Subset the data
unsupervised_data <- metadata[trainIndex, ]
supervised_data <- metadata[-trainIndex, ]

# Check the distribution of the response variable in each split
print("Unsupervised")
print(table(unsupervised_data$Response.to.Treatment))
print("Supervised")
print(table(supervised_data$Response.to.Treatment))

# Plot distributions of both subsets
ggplot(unsupervised_data, aes(x = as.factor(Response.to.Treatment), fill = as.factor(Response.to.Treatment))) +
  geom_bar() +
  geom_text(stat = 'count', aes(label = ..count..), vjust = -0.5) +
  labs(x = "Response to Treatment")+
  scale_fill_hue(c = 100) +
  theme(legend.position = "none")

ggplot(supervised_data, aes(x = as.factor(Response.to.Treatment), fill = as.factor(Response.to.Treatment))) +
  geom_bar() +
  geom_text(stat = 'count', aes(label = ..count..), vjust = -0.5) +
  labs(x = "Response to Treatment")+
  scale_fill_hue(c = 100) +
  theme(legend.position = "none")
```

# Save data subsets
```{r}
saveRDS(unsupervised_data, "unsupervised_ID")
saveRDS(supervised_data, "supervised_ID")
```