ggplot2绘制相关矩阵热图：变量排序及绘图问题求助

Got it, let's put together a complete, reproducible solution for your correlation heatmap with clustering-ordered variables using ggplot2. Here's a step-by-step breakdown you can adapt to your data:

Step 1: Load Required Packages #

First, install and load the tools we'll need for correlation calculation, data reshaping, and plotting:

# Install packages if you haven't already
install.packages(c("ggplot2", "reshape2", "Hmisc", "dplyr"))

# Load libraries
library(ggplot2)
library(reshape2)
library(Hmisc)
library(dplyr)

Step 2: Prepare Your Data #

We'll use the built-in mtcars dataset as a working example—replace this with your actual data frame.

# Replace mtcars with your own data frame
data <- mtcars

Step 3: Compute the Correlation Matrix #

Use rcorr() to calculate Pearson correlations (add type = "spearman" inside the function if you need rank-based correlations instead):

# Generate correlation matrix (extract only the correlation values with $r)
cormatx <- rcorr(as.matrix(data))$r

Step 4: Reorder Variables via Hierarchical Clustering #

This step ensures variables are sorted by their correlation similarity:

# Convert correlation values to a distance metric (0 = perfect correlation, 1 = no correlation)
dist_matrix <- as.dist((1 - cormatx)/2)

# Perform hierarchical clustering
cluster_hcl <- hclust(dist_matrix)

# Reorder the correlation matrix using the cluster order
cormatx_ordered <- cormatx[cluster_hcl$order, cluster_hcl$order]

Step 5: Reshape to Long Format for ggplot #

We need to convert the square matrix into a long data frame. Crucially, we'll set factor levels to preserve the cluster order (so ggplot doesn't reorder variables alphabetically):

# Melt the ordered matrix into long format
melted_cormatx <- melt(cormatx_ordered, na.rm = TRUE)

# Lock in the cluster order for x and y axes
melted_cormatx <- melted_cormatx %>%
  mutate(
    Var1 = factor(Var1, levels = rownames(cormatx_ordered)),
    Var2 = factor(Var2, levels = colnames(cormatx_ordered))
  )

Step 6: Build the Heatmap with ggplot2 #

This final code creates a clean, informative heatmap with correlation values, a color gradient, and readable axes:

ggplot(melted_cormatx, aes(x = Var1, y = Var2, fill = value)) +
  # Create heatmap tiles with white borders
  geom_tile(color = "white") +
  # Add rounded correlation values to each tile
  geom_text(aes(label = round(value, 2)), color = "black", size = 3) +
  # Color gradient: blue for negative correlations, red for positive, white at 0
  scale_fill_gradient2(
    low = "#0072B2",
    mid = "white",
    high = "#D55E00",
    midpoint = 0,
    limit = c(-1, 1),
    space = "Lab",
    name = "Pearson\nCorrelation"
  ) +
  # Clean up the theme for readability
  theme_minimal() +
  theme(
    axis.text.x = element_text(angle = 45, vjust = 1, size = 10, hjust = 1),
    axis.title.x = element_blank(),
    axis.title.y = element_blank(),
    panel.grid.major = element_blank(),
    panel.border = element_blank(),
    panel.background = element_blank(),
    axis.ticks = element_blank(),
    legend.justification = c(1, 0),
    legend.position = c(0.6, 0.7),
    legend.direction = "horizontal"
  ) +
  # Ensure tiles are square-shaped
  coord_fixed()

Quick Customization Tips: #

• Missing Values: rcorr() automatically handles NA values with pairwise deletion—no extra steps needed.
• Color Scheme: Swap out the low/high hex codes to match your preferred color palette.
• Text Size: Adjust the size argument in geom_text() to make correlation values larger or smaller.

内容的提问来源于stack exchange，提问作者Sophie