age_slice_inspecting.Rmd

---
title: "AgeSliceHeirarchalClustering"
author: "Martyn Fyles"
date: "07/09/2021"
output: html_document
---


```{r}
source('Visualisation/my_functions.R')
```

# Libraries

```{r}
library(reshape2)
library(ggdendro)
library(ggplot2)
library(cowplot)
library(dplyr)
library(RColorBrewer)
library(scico)
library(ggpubr)
```

# Load in lookups

```{r}
SymptomsNameCategoryLookup= read.csv(
  'Data/Lookups/SymptomNameCategoryLookup.csv'
  )
categories = read.csv(
  'Data/Lookups/CategoriesLookup.csv'
  )

text_size = 0.01

remove_axis_arrange_legend = theme(
  axis.title.x=element_blank(),
        #axis.text.x=element_blank(),
        #axis.ticks.x=element_blank(),
        axis.line = element_blank(),
        axis.title.y=element_blank(),
        axis.text.y=element_blank(),
        axis.ticks.y=element_blank(),
        legend.position = "bottom",
        legend.box = "vertical",
        legend.justification = c(0.5,0))
```

## Dendrogram grobs

```{r}
category_legend = guides(colour = guide_legend(override.aes = list(size=5), nrow = 3))

y_scale = scale_y_continuous(
    name="Jaccard distance",
    breaks = c(0, 0.2, 0.4, 0.6, 0.8, 1.0),
    limits=c(-0.5, 1))

remove_axis_arrange_legend = theme(axis.title.x=element_blank(),
        axis.text.x=element_blank(),
        axis.ticks.x=element_blank(),
        axis.line = element_blank(),
        axis.title.y=element_blank(),
        axis.text.y=element_blank(),
        axis.ticks.y=element_blank(),
        legend.position = "bottom",
        legend.box = "vertical",
        legend.justification = ("right")
  )

size_scaling =  scale_size_continuous(
    breaks = c(20, 40, 60, 80, 100),
    labels = c(20, 40, 60, 80, 100),
    limits = c(0, 100),
    range = c(2,6)
   )

shifted_y_scale = scale_y_continuous(
    name="Jaccard distance",
    breaks = c(0, 0.2, 0.4, 0.6, 0.8, 1.0),
    limits=c(-0.5, 1))
```

Heatmap grobs

```{r}
heatmap_theme = theme(axis.text.y = element_blank(),
        axis.title.y = element_blank(),
        axis.ticks.y = element_blank(),
        legend.position = 'top',
        axis.text.x = element_text(
          angle = 90,
          hjust = 1,
          vjust = 0.5,
          size = 12))


z_axis_scale = scico::scale_fill_scico(
    palette = "roma",
    limits = c(0, 1))
```

# Load distance matrices

```{r}
get_distance_matrix_p2 = function(number, title){
  p2_slice = read.csv(paste0("Data/Slices/AgeSlices/P2/Jaccard_slice_", number, ".csv"))
  
  p2_slice_long = melt(p2_slice) %>%
  inner_join(SymptomsNameCategoryLookup, by = c('X' = 'symptom_name_raw')) %>%
  filter(dataset == 'CTAS') %>%
  inner_join(SymptomsNameCategoryLookup, by = c('variable' = 'symptom_name_raw')) %>%
  filter(dataset.x == 'CTAS') %>%
  filter(dataset.y == 'CTAS') %>%
  select(
    symptom_name_formatted.x,
    symptom_name_formatted.y,
    value
  )
  
  # pillar 2
  p2_slice_symptom_names = p2_slice[,1]
  p2_slice = p2_slice[-1]
  rownames(p2_slice) <- p2_slice_symptom_names
  p2_slice_dist = as.dist(p2_slice)
  
  p2_slice_symptom_names_df = data.frame(p2_slice_symptom_names)
  colnames(p2_slice_symptom_names_df) <- 'symptom_name_raw'
  
  p2_slice_symptom_names_df = p2_slice_symptom_names_df %>%
  inner_join(SymptomsNameCategoryLookup) %>%
  filter(dataset == 'CTAS')
  
  p2_slice_dendro_dd_row = as.dendrogram(
  hclust(
    p2_slice_dist,
    method = 'complete'
    )
  )
  
  p2_slice_dendro_data = dendro_data(p2_slice_dendro_dd_row)
  
  p2_slice_dendro_data$labels = p2_slice_dendro_data$labels %>% 
  left_join(
    p2_slice_symptom_names_df,
    by = c('label' = 'symptom_name_raw')
    )

  
  p2_slice_dendrogram <- ggplot(segment(p2_slice_dendro_data)) +
  geom_segment(
    aes(x=x, y=y, xend=xend, yend=yend)) + 
  geom_text(
    data=label(p2_slice_dendro_data),
    aes(
      label=symptom_name_formatted,
      x=x,
      y=-0.05,
      angle='0',
      hjust = 1,
      label_size = text_size)) +
  geom_point(
    data = p2_slice_dendro_data$labels,
    aes(
      x=x,
      y=0,
      size=50,
      color=category)) +
  labs(
    size = 'Cases with symptom (%)',
    color = 'Category') + 
  theme_cowplot(font_size = 12) + 
  remove_axis_arrange_legend + 
  shifted_y_scale +
  colour_scale_all + 
  size_scaling + 
  coord_flip() +
  category_legend
    
  # get the ordering from the dendrogram
  p2_slice_order = order.dendrogram(p2_slice_dendro_dd_row)
  
  # Order the levels according to their position in the dendrogram
  p2_slice_long$symptom_name_formatted.x = factor(
    x = p2_slice_long$symptom_name_formatted.x,
    levels = p2_slice_long$symptom_name_formatted.x[p2_slice_order],
    ordered = TRUE
  )
  
  p2_slice_long$symptom_name_formatted.y = factor(
    x = p2_slice_long$symptom_name_formatted.y,
    levels = p2_slice_long$symptom_name_formatted.x[rev(p2_slice_order)],
    ordered = TRUE
  )
  
  p2_slice_heatmap <- ggplot(
    data = p2_slice_long,
    aes(
      x = symptom_name_formatted.y,
      y = symptom_name_formatted.x
      )
    ) +
    geom_tile(aes(fill = value)) +
    theme_minimal() +
    z_axis_scale + 
    heatmap_theme +
    xlab("Symptom") +
    labs(
       title = title,
       fill = "Jaccard distance") 
  
  p2_slice_DendroHeatmap = ggdraw() +
    draw_plot(p2_slice_heatmap, x = 0, y = 0, width = .45, height = 1) +
    draw_plot(p2_slice_dendrogram, x = 0.36, y = 0.055, width = 0.56, height = 0.85)
  
  return(p2_slice_DendroHeatmap)
}
```

```{r}
slice_0_plot = get_distance_matrix_p2('0', 'Pillar 2 - age <10')
slice_0_plot
```

```{r}
slice_1_plot = get_distance_matrix_p2('1', 'Pillar 2 - age 10-20')
slice_1_plot
```


```{r}
slice_4_plot = get_distance_matrix_p2('4', 'Pillar 2 - age 40-50')
slice_4_plot
```


```{r}
slice_9_plot = get_distance_matrix_p2('9', 'Pillar 2 - age 90-100')
slice_9_plot
```

```{r}
ggsave(filename='slice_0_plot.png', plot = slice_0_plot, width = 15,
       height = 8)
ggsave(filename='slice_1_plot.png', plot = slice_1_plot, width = 15,
       height = 8)
ggsave(filename='slice_4_plot.png', plot = slice_4_plot, width = 15,
       height = 8)
ggsave(filename='slice_9_plot.png', plot = slice_9_plot, width = 15,
       height = 8)
```

```{r}
get_distance_matrix_sgss = function(number, title){
  sgss_slice = read.csv(paste0("Data/Slices/AgeSlices/SGSS/Jaccard_slice_", number, ".csv"))
  
  sgss_slice_long = melt(sgss_slice) %>%
  inner_join(SymptomsNameCategoryLookup, by = c('X' = 'symptom_name_raw')) %>%
  filter(dataset == 'CTAS') %>%
  inner_join(SymptomsNameCategoryLookup, by = c('variable' = 'symptom_name_raw')) %>%
  filter(dataset.x == 'CTAS') %>%
  filter(dataset.y == 'CTAS') %>%
  select(
    symptom_name_formatted.x,
    symptom_name_formatted.y,
    value
  )
  
  # pillar 2
  sgss_slice_symptom_names = sgss_slice[,1]
  sgss_slice = sgss_slice[-1]
  rownames(sgss_slice) <- sgss_slice_symptom_names
  sgss_slice_dist = as.dist(sgss_slice)
  
  sgss_slice_symptom_names_df = data.frame(sgss_slice_symptom_names)
  colnames(sgss_slice_symptom_names_df) <- 'symptom_name_raw'
  
  sgss_slice_symptom_names_df = sgss_slice_symptom_names_df %>%
  inner_join(SymptomsNameCategoryLookup) %>%
  filter(dataset == 'CTAS')
  
  sgss_slice_dendro_dd_row = as.dendrogram(
  hclust(
    sgss_slice_dist,
    method = 'complete'
    )
  )
  
  sgss_slice_dendro_data = dendro_data(sgss_slice_dendro_dd_row)
  
  sgss_slice_dendro_data$labels = sgss_slice_dendro_data$labels %>% 
  left_join(
    sgss_slice_symptom_names_df,
    by = c('label' = 'symptom_name_raw')
    )

  
  sgss_slice_dendrogram <- ggplot(segment(sgss_slice_dendro_data)) +
  geom_segment(
    aes(x=x, y=y, xend=xend, yend=yend)) + 
  geom_text(
    data=label(sgss_slice_dendro_data),
    aes(
      label=symptom_name_formatted,
      x=x,
      y=-0.05,
      angle='0',
      hjust = 1,
      label_size = text_size)) +
  geom_point(
    data = sgss_slice_dendro_data$labels,
    aes(
      x=x,
      y=0,
      size=50,
      color=category)) +
  labs(
    size = 'Cases with symptom (%)',
    color = 'Category') + 
  theme_cowplot(font_size = 12) + 
  remove_axis_arrange_legend + 
  shifted_y_scale +
  colour_scale_all + 
  size_scaling + 
  coord_flip() +
  category_legend
    
  # get the ordering from the dendrogram
  sgss_slice_order = order.dendrogram(sgss_slice_dendro_dd_row)
  
  # Order the levels according to their position in the dendrogram
  sgss_slice_long$symptom_name_formatted.x = factor(
    x = sgss_slice_long$symptom_name_formatted.x,
    levels = sgss_slice_long$symptom_name_formatted.x[sgss_slice_order],
    ordered = TRUE
  )
  
  sgss_slice_long$symptom_name_formatted.y = factor(
    x = sgss_slice_long$symptom_name_formatted.y,
    levels = sgss_slice_long$symptom_name_formatted.x[rev(sgss_slice_order)],
    ordered = TRUE
  )
  
  sgss_slice_heatmap <- ggplot(
    data = sgss_slice_long,
    aes(
      x = symptom_name_formatted.y,
      y = symptom_name_formatted.x
      )
    ) +
    geom_tile(aes(fill = value)) +
    theme_minimal() +
    z_axis_scale + 
    heatmap_theme +
    xlab("Symptom") +
    labs(
       title = title,
       fill = "Jaccard distance") 
  
  sgss_slice_DendroHeatmap = ggdraw() +
    draw_plot(sgss_slice_heatmap, x = 0, y = 0, width = .45, height = 1) +
    draw_plot(sgss_slice_dendrogram, x = 0.36, y = 0.055, width = 0.56, height = 0.85)
  
  return(sgss_slice_DendroHeatmap)
}
```