R语言Dendrogram树状图绘图美化及导出问题求助

原问题概述 #

尝试在同一张画布绘制4种不同聚类算法的树状图，各划分为4个簇，原绘图效果如下：

原使用代码如下：

data <- df[, c(2, 1305:2185)]

data_HC <-
  data %>% remove_rownames %>% column_to_rownames(var = "Name")

h1 <- data_HC %>% dist %>% hclust(method='average') %>% as.dendrogram
h2 <- data_HC %>% dist %>% hclust(method='complete') %>% as.dendrogram
h3 <- data_HC %>% dist %>% hclust(method='ward.D') %>% as.dendrogram
h4 <- data_HC %>% dist %>% hclust(method='single') %>% as.dendrogram

compare_clusters <- function(data_1, data_2, data_3, data_4){
  par(mfrow=c(2,2))
  cols = c('red', 'green', 'blue', 'pink')
  plot(data_1, main='Average Linkage')
  cut_avg_h1 <- cutree(data_1, k = 4)
  rect.dendrogram(data_1 , k = 4, border = cols)
  plot(data_2, main='Complete Linkage')
  cut_avg_h2 <- cutree(data_2, k = 4)
  rect.dendrogram(data_2 , k = 4, border = cols)
  plot(data_3, main="Ward's Linkage")
  cut_avg_h3 <- cutree(data_3, k = 4)
  rect.dendrogram(data_3 , k = 4, border = cols)
  plot(data_4, main='Single Linkage')
  cut_avg_h4 <- cutree(data_4, k = 4)
  rect.dendrogram(data_4 , k = 4, border = cols)
}

plot <- compare_clusters(h1, h2, h3, h4)

需要解决4个需求：

• 裁剪Ward法树状图，仅显示高度50及以下的部分，突出4个簇
• 缩小所有样本标签字号，避免重叠
• 缩小树枝和标签之间的空隙
• 实现绘图结果的本地导出

修改后可运行代码 #

# 数据预处理部分保持不变
data <- df[, c(2, 1305:2185)]
data_HC <- data %>% remove_rownames %>% column_to_rownames(var = "Name")

h1 <- data_HC %>% dist %>% hclust(method='average') %>% as.dendrogram
h2 <- data_HC %>% dist %>% hclust(method='complete') %>% as.dendrogram
h3 <- data_HC %>% dist %>% hclust(method='ward.D') %>% as.dendrogram
h4 <- data_HC %>% dist %>% hclust(method='single') %>% as.dendrogram

compare_clusters <- function(data_1, data_2, data_3, data_4){
  par(mfrow=c(2,2), mar = c(2,2,2,2)) # 缩小页边距进一步节省空间
  cols = c('red', 'green', 'blue', 'pink')
  # 所有plot新增cex（标签字号）、offset（树枝与标签间距）参数
  plot(data_1, main='Average Linkage', cex = 0.5, offset = 0.1)
  rect.dendrogram(data_1 , k = 4, border = cols)
  
  plot(data_2, main='Complete Linkage', cex = 0.5, offset = 0.1)
  rect.dendrogram(data_2 , k = 4, border = cols)
  
  # Ward法新增ylim参数裁剪高度，仅显示0-50区间
  plot(data_3, main="Ward's Linkage", cex = 0.5, offset = 0.1, ylim = c(0,50))
  rect.dendrogram(data_3 , k = 4, border = cols)
  
  plot(data_4, main='Single Linkage', cex = 0.5, offset = 0.1)
  rect.dendrogram(data_4 , k = 4, border = cols)
}

# 先开绘图设备，设置合适的宽高避免标签截断，单位为像素，分辨率可按需调整
png("dendrogram_compare.png", width = 1200, height = 1000, res = 150)
compare_clusters(h1, h2, h3, h4)
dev.off() # 关闭设备完成保存

参数说明 #

• cex = 0.5：调整样本标签的缩放比例，数值越小字号越小，可根据样本数量自行调整
• offset = 0.1：调整标签和树枝末端的距离，默认值为1，数值越小间距越小
• ylim = c(0,50)：限制Y轴（树高度）的显示范围，仅展示高度0到50的部分，实现顶部分支裁剪
• 保存时用png()函数可直接导出为png格式，需要导出矢量图可替换为pdf("dendrogram_compare.pdf", width = 12, height = 10)

内容的提问来源于stack exchange，提问作者AP_98