如何在ggplot离散y轴类别间添加自定义p值括号

问题 #

我拥有如下数据集：

plot2=structure(list(partition = c("2", "2b", "2c"), partition_name = structure(1:3, levels = c("Cardiac_significant_enh partition PRS Original OR", 
"Cardiac_significant_enh partition PRS OR × enh_ES", "Cardiac_significant_enh partition PRS OR × enh_TS_tpm"
), class = c("ordered", "factor")), Num_SNP = c(3157, 3157, 3157
), CoD_per_SNP = c(62.1186923796969, 61.1820673263678, 60.9483894486912
), R2 = c(0.0196108711842703, 0.0193151786549343, 0.0192414065489518
), LCL = c(0.010974448167152, 0.0135183879280951, 0.0112302824404708
), UCL = c(0.0282472942013886, 0.0251119693817734, 0.0272525306574328
), R2type = c("Lee", "Lee", "Lee"), xlabel = structure(c(3L, 
1L, 2L), levels = c("Cardiac significant enh\npartition PRS OR × enh ES (SNP\nN=3157)", 
"Cardiac significant enh\npartition PRS OR × enh TS tpm\n(SNP N=3157)", 
"Cardiac significant enh\npartition PRS Original OR (SNP\nN=3157)"
), class = "factor")), row.names = 4:6, class = "data.frame")

并通过以下代码绘制了ggplot图表：

ggplot(data = plot2,
         aes(
           y=reorder(plot2$xlabel, desc(partition)), 
           x=plot2$R2*100, xmin = plot2$LCL*100, xmax=plot2$UCL*100, 
           colour=plot2$R2type,group=plot2$R2type
           # label=round(R2*100,2)
         )) +  
  geom_pointrange(
    position=pos, size=1, lwd=1) +
  scale_color_manual(values=MetBrewer::met.brewer("Johnson", 2),
                     expression(paste(R^2)))+
  ggrepel::geom_label_repel(
    aes(
      y=reorder(xlabel, desc(partition)), 
      x=R2*100,
      colour=R2type,group=R2type,
      label=round(R2*100,2)
    ),
    position = pos, size = rel(4),  show.legend = F,min.segment.length = 0,
    point.padding = NA, box.padding = 0.5)+ #nudge_y = -0.2, 
  scale_x_continuous(limits = c(0, NA), expand = expansion(mult = c(0, .15))) + 
  scale_y_discrete(expand = expansion(add = .7)) + #.6 is default - add a little spacing to the sides
  xlab("") +  ylab(paste0("HCM diagnosis ~"))+theme_bw() +
  theme(
    legend.position = "none",  
    text=element_text(lineheight = 0.8, angle = 0, size = 21, color = "gray8")
  )

当前图表的分类变量位于y轴，我希望使用ggpubr包的geom_bracket()函数在离散y轴的不同类别间添加自定义p值括号，但该函数默认仅支持对离散x轴的类别进行比较。我需要实现能够选择特定两两比较（例如第一行与第三行类别）并添加自定义p值括号的效果。

解决方案 #

geom_bracket()本质是通过坐标定位绘制括号，针对y轴离散类别的场景，只需手动映射离散y值对应的数值位置，再指定括号的水平位置即可实现需求。

步骤1：确认离散y轴的数值映射 #

ggplot中离散y轴的类别默认按从下到上的顺序对应数值1、2、3...。结合你的代码reorder(xlabel, desc(partition))，最终y轴从上到下的类别及对应数值为：

• "Cardiac significant enh\npartition PRS Original OR (SNP\nN=3157)" → 数值3
• "Cardiac significant enh\npartition PRS OR × enh ES (SNP\nN=3157)" → 数值2
• "Cardiac significant enh\npartition PRS OR × enh TS tpm\n(SNP N=3157)" → 数值1

若要比较第一行（最上方，数值3）和第三行（最下方，数值1），直接使用这两个数值作为括号的y起始/结束位置即可。

步骤2：修改代码添加自定义括号 #

在原有ggplot代码基础上，添加geom_bracket()层，指定括号的水平位置、y轴对应数值和自定义p值标签：

完整修改后的代码如下：

library(ggplot2)
library(ggrepel)
library(ggpubr)
library(MetBrewer)

# 补充原代码未定义的position参数
pos <- position_dodge(width = 0.5)

ggplot(data = plot2,
       aes(
         y = reorder(xlabel, desc(partition)), 
         x = R2*100, xmin = LCL*100, xmax = UCL*100, 
         colour = R2type, group = R2type
       )) +  
  geom_pointrange(position = pos, size = 1, lwd = 1) +
  scale_color_manual(values = MetBrewer::met.brewer("Johnson", 2),
                     name = expression(paste(R^2)))+
  ggrepel::geom_label_repel(
    aes(
      x = R2*100,
      label = round(R2*100,2)
    ),
    position = pos, size = rel(4),  show.legend = F, min.segment.length = 0,
    point.padding = NA, box.padding = 0.5)+
  scale_x_continuous(limits = c(0, 3.5), expand = expansion(mult = c(0, .15))) + 
  scale_y_discrete(expand = expansion(add = .7)) +
  xlab("") +  ylab(paste0("HCM diagnosis ~"))+theme_bw() +
  # 添加自定义p值括号
  geom_bracket(
    x = 3.2,  # 括号水平位置，避开数据点范围
    y.position = c(3, 1),  # 要比较的两个y类别对应数值
    label = "p = 0.85",  # 自定义p值
    tip.length = 0.02,  # 括号尖端长度
    size = 1,  # 括号线条粗细
    color = "gray8"
  ) +
  theme(
    legend.position = "none",  
    text = element_text(lineheight = 0.8, angle = 0, size = 21, color = "gray8")
  )

关键说明 #

• 若要比较其他类别，只需调整y.position的数值（比如比较前两个类别用c(3,2)）。
• x参数需设置在x轴数据范围之外（原数据UCL*100最大约2.82，设为3.2可避免重叠），可根据需求灵活调整。
• 如需添加多组括号，重复调用geom_bracket()，每组设置对应的y.position和label即可。

内容的提问来源于stack exchange，提问作者Emos