Clustered ggplot2 heatmap

clustered_heatmap.R

library(tidyverse)

get_order <- function(df,  distmethod = "pearson", hclustmethod = "average", output_ordername = "order"){
  #Get the row ordering from a clustering
  hr <- hclust(as.dist(1-cor(t(df), method=distmethod)), method=hclustmethod)
  order <- data.frame(hr$labels[hr$order])
  order$ordering <- rownames(order)
  names(order) <- c("ID", output_ordername)
  return(order)
}

mtcars[is.na(mtcars)] <- 0 # Not needed for mtcars

# Get correlation between all variables
mtcars_cor <- cor(mtcars, mtcars, method  = "pearson") 

# Cluster correlations, get clustered row/column order
corr_order <- get_order(mtcars_cor) %>% 
  as_tibble() %>%
  mutate(ID = as.character(ID))

mtcars_cor %>% 
  as_tibble(rownames = "ID1") %>% 
  gather(ID2, r, -ID1) %>%
  mutate(ID1 = fct_relevel(ID1, corr_order$ID)) %>% # Gets rows in clustered order
  mutate(ID2 = fct_relevel(ID2, corr_order$ID)) %>% # Gets columns in clustered order
  ggplot(aes(x= ID1, y = ID2, fill = r)) + 
  geom_tile() + 
  scale_fill_gradient(limit = c(0,1), low = "white", high = "black", name = "Pearson's r") +
  theme(axis.text.x = element_text(angle = 45, vjust = 1, hjust = 1)) + 
  theme(axis.title.x = element_blank(), axis.title.y = element_blank())