复杂热图

作者

[编辑] 柴银盈

[贡献] 金子伦, 郑虎

示例

复杂热图（Complex Heatmap）是一种高级数据可视化工具，主要用于展示多维数据的矩阵关系，并在基础热图（Heatmap）上增加了多层次注释、聚类分析、数据拆分等功能。它能够更清晰地呈现复杂数据的内在结构和关联性。

环境配置

系统要求：跨平台（Linux/MacOS/Windows）
编程语言：R
依赖包：ComplexHeatmap, dendextend, tidyr, circlize, gridExtra, pheatmap

# 安装包
if (!requireNamespace("ComplexHeatmap", quietly = TRUE)) {
  install.packages("ComplexHeatmap")
}
if (!requireNamespace("dendextend", quietly = TRUE)) {
  install.packages("dendextend")
}
if (!requireNamespace("tidyr", quietly = TRUE)) {
  install.packages("tidyr")
}
if (!requireNamespace("circlize", quietly = TRUE)) {
  install.packages("circlize")
}
if (!requireNamespace("gridExtra", quietly = TRUE)) {
  install.packages("gridExtra")
}
if (!requireNamespace("pheatmap", quietly = TRUE)) {
  install.packages("pheatmap")
}

# 加载包
library(ComplexHeatmap)
library(dendextend)
library(tidyr)
library(circlize)
library(gridExtra)
library(pheatmap)

数据准备

数据主要来自TCGA数据库的甲基化数据

#data_mat(连续型)
data <- readr::read_csv("https://bizard-1301043367.cos.ap-guangzhou.myqcloud.com/TCGA.BRCA.sampleMap_HumanMethylation27_ch.csv")

data <- as.data.frame(data)
rownames(data) <- data[,1]
data <- data[,-1]

data_TCGA <- na.omit(data[1:20,1:20])

data_TCGA <- as.matrix(data_TCGA)  

#离散型
discrete_mat = matrix(sample(1:5, 100, replace = TRUE), 10, 10)

可视化

1. 基础复杂热图

连续型变量

这里我们以TCGA数据库的甲基化数据为例

# 连续型变量
Heatmap(data_TCGA , name = "Methylation")

上图为甲基化水平的复杂热图，列为基因位点，行为样本。红色表示高甲基化水平，蓝色表示低甲基化水平。边轴为聚类分析结果。

离散型变量

这里我们以生成数据为例

# 离散型变量
Heatmap(discrete_mat,  name = "mat")

上图为生成的离散变量的复杂热图，边轴为聚类分析结果。

2. 个性化

2.1 颜色自定义

连续型变量

# 连续型变量颜色自定义
col_fun = colorRamp2(c(0, 0.5, 1),  #设置截断点
                     c("blue", "white", "pink")) 
Heatmap(data_TCGA, name = "Methylation", col = col_fun, 
              column_title = "自定义三段颜色")

# 连续型变量颜色自定义
Heatmap(data_TCGA, name = "Methylation", col = rev(rainbow(10)), 
        column_title = "创建颜色向量")

离散型变量

# 离散型变量颜色自定义
colors <- c(
  "1" = "#BBEDA7",  
  "2" = "#F3D8FA",  
  "3" = "#FFECA1",  
  "4" = "#D9E6F7",  
  "5" = "#EFBEC6"   
)
Heatmap(discrete_mat, name = "mat", col = colors,
        column_title = "离散型变量颜色自定义")

2.2 边框自定义

# 边框自定义
Heatmap(data_TCGA, name = "Methylation", 
        border = "black", #border值可为FALSE或者颜色
        rect_gp = gpar(col = "white", lwd = 2),#lty线型, lwd线宽
        column_title = "边框自定义")

3. 聚类

可以自定义聚类方法，树状图位置、大小等

# 聚类
Heatmap(data_TCGA, name = "Methylation", 
       # row_dend_side = "right", 
       # column_dend_side = "bottom", #可调整树状图位置
        column_dend_height = unit(1, "cm"), 
        row_dend_width = unit(1, "cm"),#可调整树状图距离
        column_title = "聚类自定义",
        clustering_distance_rows = "pearson", #聚类方法，可自设函数
      # show_parent_dend_line = FALSE  #可隐藏虚线
        )

树状图渲染

我们还利用dendextend包对树状图进行颜色渲染

# 树状图渲染
row_dend = as.dendrogram(hclust(dist(data_TCGA)))
row_dend = color_branches(row_dend, k = 2)  #生成分支
#数字表示颜色数量
Heatmap(data_TCGA, name = "Methylation", 
        cluster_rows = row_dend,
      # row_dend_gp = gpar(col = "red")  #也可自定义颜色
        )

上图可见树状图分支为不同颜色

4. 分割

连续型变量

采用k-means聚类分割

# 连续型变量分割
Heatmap(data_TCGA, 
        name = "Methylation", 
        row_km = 2,  #设置分割数
        column_km = 3)

利用向量分割

# 利用向量分割
Heatmap(data_TCGA, name = "Methylation", 
        row_split = rep(c("A", "B"), 9), 
        column_split = rep(c("C", "D"), 10))

利用矩阵分割

# 利用矩阵分割
Heatmap(data_TCGA, name = "Methylation", 
        row_split = data.frame(rep(c("A", "B"), 9), 
                               rep(c("C", "D"), each = 9)))

以上分割方法均可行列独立混使用，设置show_parent_dend_line = FALSE可隐藏虚线

离散型变量

按指定行列进行分割

# 按指定行列进行分割
Heatmap(discrete_mat, 
        name = "mat", 
        col = 1:4, 
        row_split = discrete_mat[,1])#根据矩阵数据中的第一列进行分割

参数分割

可根据热图分割对标题，聚类参数等进行分割

# 参数分割
Heatmap(data_TCGA, name = "Methylation", 
        row_km = 2, 
        row_title_gp = gpar(col = c("#9A090A", "#0D4B99"), font = 1:2),
        row_names_gp = gpar(col = c("#748901", "#E6AC03"), fontsize = c(10, 12)),
        column_km = 3, 
        column_title_gp = gpar(fill = c("#BFD641", "#FE9900", "#CC6CE7"), font = 1:3), #fill用于填充
        column_names_gp = gpar(col = c("#1D6103", "#E08804", "#5E0676"), fontsize = c(10, 12, 8)))

注释分割

可根据热图分割对边注释进行分割

# 注释分割
Heatmap(data_TCGA, name = "Methylation", row_km = 2, column_km = 3,
        top_annotation = HeatmapAnnotation(foo1 = 1:20, bar1 = anno_points(runif(20))),
        right_annotation = rowAnnotation(foo2 = 18:1, bar2 = anno_barplot(runif(18)))
)

上图中添加的点图、热图、条图注释均随主热图分割

5. 标签添加

添加数据

cell_fun 接受一个函数，该函数会被应用到热图的每个单元格上。函数的参数如下：

j: 列索引（列号）。
i: 行索引（行号）。
x, y: 单元格中心点的坐标（基于绘图设备的坐标系）。
width, height: 单元格的宽度和高度。
fill: 单元格对应的数值（即矩阵中的值）。

# 添加数据
small_mat = data_TCGA[1:9, 1:9]
col_fun = colorRamp2(c(0, 0.5, 1), c("#EBA7FE", "white", "pink"))
Heatmap(small_mat, name = "Methylation", col = col_fun,
        cell_fun = function(j, i, x, y, width, height, fill) {
          grid.text(sprintf("%.1f", small_mat[i, j]), x, y, gp = gpar(fontsize = 10))
        })

# 添加数据
small_mat = data_TCGA[1:9, 1:9]
Heatmap(small_mat, name = "Methylation",  col = col_fun,
        cell_fun = function(j, i, x, y, width, height, fill) {
          if(small_mat[i, j] > 0.5) #设定显示范围
            grid.text(sprintf("%.1f", small_mat[i, j]), x, y, gp = gpar(fontsize = 10)) #添加文本/数值
        })

上图仅显示甲基化水平大于0.5的数据

添加图形

# 添加图形
small_mat = data_TCGA[1:9, 1:9]
Heatmap(small_mat, name = "Methylation", col = col_fun,
        row_km = 2, column_km = 2,
        layer_fun = function(j, i, x, y, w, h, fill) {
          ind_mat = restore_matrix(j, i, x, y)
          ind = unique(c(ind_mat[1, 4], ind_mat[3, 2])) #定位,每个分割对应位置添加图形
          grid.points(x[ind], y[ind], pch = 16, size = unit(4, "mm")) #添加点
        }
)

应用场景

上图是基于 AVR 数据[1]中 59 个差异甲基化 ATG 位点的无监督层次聚类热图。红色表示 DNA 甲基化程度高，绿色表示甲基化程度低。

参考文献

[1] Radhakrishna U, Albayrak S, Alpay-Savasan Z, Zeb A, Turkoglu O, Sobolewski P, Bahado-Singh RO. Genome-Wide DNA Methylation Analysis and Epigenetic Variations Associated with Congenital Aortic Valve Stenosis (AVS). PLoS One. 2016 May 6;11(5):e0154010. doi: 10.1371/journal.pone.0154010. PMID: 27152866; PMCID: PMC4859473.

[2] Gu, Z. (2016) Complex heatmaps reveal patterns and correlations in multidimensional genomic data. Bioinformatics.Gu, Z. (2022) Complex Heatmap Visualization. iMeta.

[3] Tal Galili (2015). dendextend: an R package for visualizing, adjusting, and comparing trees of hierarchical clustering. Bioinformatics. DOI:10.1093/bioinformatics/btv428

[4] Wickham H, Vaughan D, Girlich M (2024). tidyr: Tidy Messy Data. R package version 1.3.1, CRAN: Package tidyr.

[5] Gu, Z. (2014) circlize implements and enhances circular visualization in R. Bioinformatics.

[6] Auguie B (2017). gridExtra: Miscellaneous Functions for “Grid” Graphics. R package version 2.3, CRAN: Package gridExtra.

[7] Kolde R (2019). pheatmap: Pretty Heatmaps. R package version 1.0.12, CRAN: Package pheatmap.