ggplot2：如何将图例按颜色拆分为自定义图例

Question

ggplot2：如何将图例按颜色拆分为自定义图例

3

我正在努力让绘图标签以特定的方式呈现。我正在使用ggplot2和tidyverse。

这是我的代码:

我希望图例有两个标题（=名称），一个是细胞类型HCT，另一个是细胞类型RKO。然后对于HCT和RKO，我想要具有相应颜色、线型和形状的试剂的图例。因此，基本上，我想将颜色图例分成两个单独的图例。我只是无法理解如何编写代码。这是我想要的绘图图例（请想象橙色正方形被填充）：

我需要改变我的geom_line和geom_point代码才能实现我想要的图例风格吗？还是有其他方法可以做到？我尝试搜索了一下，但没有找到任何相关内容（也许我只是没有使用正确的术语）。我已经尝试按照这里所做的步骤：如何在ggplot中合并颜色、线条样式和形状图例和将颜色和形状的图例组合成一个图例，但无法使其工作。（换句话说，我尝试改变scale_shape_manual等来适应我的需求，但没有成功。我还尝试使用interaction()）。

注意：我决定不使用facet_wrap，因为我想在同一张图上显示两种细胞类型。真实数据的图看起来有些不同，而且不会那么混乱。我已经成功地用ggpubr绘制了一个"facet_wrap"图。

注意2：我也没有使用stat_summary()，因为我需要取相同试剂浓度、试剂和细胞类型的平均值。根据我的数据，我找不到使stat_summary工作的方法。

这是我目前拥有的代码：

mean_mutated <- mutated %>% group_by(Reagent, Reagent.Conc, Cell.type) %>%
  summarise(Avg.Viable.Cells = mean(Mean.Viable.Cells.1, na.rm = TRUE))
mutated_0 = mutated %>% group_by(Reagent, Reagent.Conc, Cell.type) %>% filter(Reagent=="0") %>% 
  summarise(Avg.Viable.Cells = mean(Mean.Viable.Cells.1, na.rm = TRUE))
mutated_1 = mutated %>% group_by(Reagent, Reagent.Conc, Cell.type) %>% filter(Reagent=="1") %>% 
  summarise(Avg.Viable.Cells = mean(Mean.Viable.Cells.1, na.rm = TRUE))
mutated_2 = mutated %>% group_by(Reagent, Reagent.Conc, Cell.type) %>% filter(Reagent=="2") %>% 
  summarise(Avg.Viable.Cells = mean(Mean.Viable.Cells.1, na.rm = TRUE))

#linetype by reagent
ggplot() +  
  #the scatter plot per cell type -> that way I can color them the way I want to, I believe
  #the mean/average line plot 
  geom_point(mean_mutated, mapping= aes(x = as.factor(Reagent.Conc), y = Avg.Viable.Cells, shape=as.factor(Reagent), color=Cell.type)) +
  geom_line(mutated_1, mapping= aes(x = as.factor(Reagent.Conc),y = Avg.Viable.Cells, group=Cell.type, color=Cell.type, linetype = "1"))+
  geom_line(mutated_2, mapping= aes(x = as.factor(Reagent.Conc),y = Avg.Viable.Cells, group=Cell.type, color=Cell.type, linetype = "2"))+
  geom_line(mutated_0, mapping= aes(x = as.factor(Reagent.Conc),y = Avg.Viable.Cells, group=Cell.type, color=Cell.type, linetype = "0"))+
  
  #making the plot look prettier
  scale_colour_manual(values = c("#999999", "#E69F00")) +
  #scale_linetype_manual(values = c("solid", "dashed", "dotted")) + #for whatever reason, when I add this, the dash in the legend is removed...?
  labs(shape = "Reagent", linetype = "Reagent", color="Cell type")+
  scale_shape_manual(values=c(15,16,4), labels=c("0", "1", "2"))+
  #guides(shape = FALSE)+ #this removes the label that you don't want
  
  #Change the look of the plot and change the axes
  xlab("[Reagent] (nM/ml)")+ #change name of x-axis
  ylab("Relative viability")+ #change name of y-axis
  scale_y_continuous(breaks = scales::pretty_breaks(n = 10))+ #adjust the y-axis so that it has more ticks
  expand_limits(y = 0)+
  theme_bw() + #this and the next line are to remove the background grid and make it look more publication-like
  theme(panel.border = element_blank(), panel.grid.major = element_blank(),
        panel.grid.minor = element_blank(), axis.line = element_line(colour = "black"))

以下是我通过dput(df[9:32, c(1,2,3,4,5)])命令生成的数据框“mutated”的快照：

    structure(list(Biological.Replicate = c(1L, 1L, 1L, 1L, 1L, 1L, 
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 
1L, 1L), Reagent.Conc = c(10000, 2500, 625, 156.3, 39.1, 9.8, 
2.4, 0.6, 10000, 2500, 625, 156.3, 39.1, 9.8, 2.4, 0.6, 10000, 
2500, 625, 156.3, 39.1, 9.8, 2.4, 0.6), Reagent = c(1L, 1L, 1L, 
1L, 1L, 1L, 1L, 1L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 0L, 0L, 0L, 
0L, 0L, 0L, 0L, 0L), Cell.type = c("HCT", "HCT", "HCT", "HCT", 
"HCT", "HCT", "HCT", "HCT", "HCT", "HCT", "HCT", "HCT", "HCT", 
"HCT", "HCT", "HCT", "RKO", "RKO", "RKO", "RKO", "RKO", "RKO", 
"RKO", "RKO"), Mean.Viable.Cells.1 = c(1.014923966, 1.022279854, 
1.00926559, 0.936979842, 0.935565248, 0.966403395, 1.00007073, 
0.978144524, 1.019673384, 0.991595836, 0.977270557, 1.007353643, 
1.111928183, 0.963518289, 0.993028364, 1.027409034, 1.055452733, 
0.953801253, 0.956577449, 0.792568337, 0.797052961, 0.755623576, 
0.838482346, 0.836773918)), row.names = 9:32, class = "data.frame")

注意3：尽管一个列名为“Mean.Viable.Cells.1”，但这不是我正在绘制的平均值，而是之前计算的技术重复测量值的平均值。我正在从mutated_0、mutated_1和mutated_2的生物学重复中取平均值进行绘图。

- Fio Da

我建议您使用dput(head(df))命令来展示您的数据框，这样有助于其他用户重新创建数据框并使用您的代码。不要使用数据框图片。 - PesKchan

谢谢您的建议 @PesKchan！我添加了一小段数据，并确保它具有三种试剂和两种细胞类型。 - Fio

1

我希望你已经在下面得到了你的答案。 - PesKchan

1个回答

网页内容由stack overflow 提供, 点击上面的

可以查看英文原文，
原文链接

- stefan · Accepted Answer

利用ggnewscale包，可以这样实现：

在操作数据集之前将Cell.Type和Reagent转换为因子
不需要使用数据集mutate_0等。只需要一个汇总数据集，我通过Cell.type分割数据集以简化后续的代码。
为了获得所需的结果，分别为每个细胞类型绘制数据图。这就是为什么我按细胞类型分割数据的原因。
利用ggnewscale::new_scale添加第二个比例尺和图例来获取分离的图例，用于linetype和shape。此外，从美学中删除color并将其设置为参数。
至少对于您的数据片段，您必须在两个比例尺上都添加drop=FALSE以保留未使用的因子水平。
最后，为了减少代码重复，我使用一个帮助函数为每个Cell.type添加geoms和比例尺。

library(ggplot2)
library(dplyr)

mutated <- mutated %>% 
  mutate(Cell.type = factor(Cell.type, levels = c("HCT", "RKO")),
         Reagent = factor(Reagent, levels = c("0", "1", "2"))
  )

mean_mutated <- mutated %>%
  group_by(Reagent, Reagent.Conc, Cell.type) %>%
  summarise(Avg.Viable.Cells = mean(Mean.Viable.Cells.1, na.rm = TRUE)) %>% 
  split(.$Cell.type)
#> `summarise()` has grouped output by 'Reagent', 'Reagent.Conc'. You can override using the `.groups` argument.

layer_geom_scale <- function(cell_type, color) {
  list(
    geom_point(mean_mutated[[cell_type]], mapping = aes(shape = Reagent), color = color),
    geom_line(mean_mutated[[cell_type]], mapping = aes(group = Reagent, linetype = Reagent), color = color),
    scale_linetype_manual(name = cell_type, values = c("solid", "dashed", "dotted"), drop=FALSE),
    scale_shape_manual(name = cell_type, values = c(15, 16, 4), labels = c("0", "1", "2"), drop=FALSE) 
  )
}

# linetype by reagent
ggplot(mapping = aes(
  x = as.factor(Reagent.Conc),
  y = Avg.Viable.Cells
)) +
  layer_geom_scale("HCT", "#999999") +
  ggnewscale::new_scale("linetype") +
  ggnewscale::new_scale("shape") +
  layer_geom_scale("RKO", "#E69F00") +
  scale_y_continuous(breaks = scales::pretty_breaks(n = 10), limits = c(0, NA)) +
  theme_bw() +
  theme(
    panel.border = element_blank(), panel.grid.major = element_blank(),
    panel.grid.minor = element_blank(), axis.line = element_line(colour = "black")
  ) +
  labs(shape = "Reagent", 
       linetype = "Reagent", 
       color = "Cell type",
       x = "[Reagent] (nM/ml)",
       y = "Relative viability")