如何使用 ggplot2 制作多面箱线图

Question

我有以下数据框：

samples_i <- c("LAIV D0", "LAIV D3", "LAIV D7", "LAIV D0", "LAIV D3", "LAIV D7", 
"TIV D0", "TIV D3", "TIV D7", "TIV D0", "TIV D3", "TIV D7")

irisTag_i <- structure(c(0, 0, 0, 11.2672863636364, 0, 0, 0, 0, 0, 0, 13.8881727272727, 
0, 0, 0), .Dim = c(2L, 7L), .Dimnames = list(c("HSP90B1", "DNAJB1"
), c("Neutrophil", "Tcell", "Monocyte", "Bcell", "NKcell", "PlasmaCell", 
"DendriticCell")))


SPVsR_i <- structure(c(0.1620678925564, -0.0609851972808482, -0.101082695275552, 
0.184268723991321, -0.0899021067853178, -0.0943666172060028, 
0.178289177586651, -0.0823892768809311, -0.0958999007057199, 
0.0331377432233005, 0.00289013805790048, -0.036027881281201, 
-0.0531973808347148, 0.0213528550009522, 0.0318445258337625, 
0.0179790366380429, 0.00347902775389391, -0.0214580643919368, 
-0.0136820170970586, 0.0142833182813199, -0.000601301184261278, 
0.0109856660204762, -0.00528600624634141, -0.00569965977413478, 
-0.0760171167711921, 0.0344372228755224, 0.0415798938956697, 
-0.114239469843063, 0.0217218301803764, 0.0925176396626868, -0.113283279031257, 
0.0424936766667866, 0.07078960236447, -0.14127024964406, 0.0595080054464686, 
0.0817622441975909, -0.0100499090500894, 0.0131491664210288, 
-0.00309925737093941, 0.101206058442775, 0.0231964804556542, 
-0.124402538898429, 0.00411785437964246, 0.0405556634613935, 
-0.044673517841036, 0.0720705616752313, -0.00782701824901867, 
-0.0642435434262126, 0.0753224665976433, -0.0323083061719772, 
-0.0430141604256661, -0.0654080281579984, 0.0124273486220488, 
0.0529806795359496, -0.0519970799923912, 0.00818146905729871, 
0.0438156109350925, 0.0200682008260364, -0.0466408267852637, 
0.0265726259592274, -0.0390251373720762, -0.0115216989414941, 
0.0505468363135703, 0.0321298528741327, -0.0151866963239294, 
-0.0169431565502034, -0.0310600302048482, 0.00718748395053659, 
0.0238725462543116, -0.0216937374381297, -0.00559429498828404, 
0.0272880324264137, 0.0288166559498562, -0.0173984873138801, 
-0.0114181686359761, -0.0176892628883129, -0.0235673738231865, 
0.0412566367114994, -0.00794904064609583, -0.000656094604392996, 
0.00860513525048881, -0.0538196455977893, 0.0200107051556589, 
0.0338089404421304), .Dim = c(12L, 7L), .Dimnames = list(c("1", 
"2", "3", "4", "5", "6", "7", "8", "9", "10", "11", "12"), c("Neutrophil", 
"Tcell", "Monocyte", "Bcell", "NKcell", "PlasmaCell", "DendriticCell"
)))

还有这段代码

par(mfrow=c(2,3), mai=c(1,0.4,0.4,0.1), omi=rep(0,4))
    for ( i in c(2:7)){
        # deliberately skip i=1
        # so we 2 x 3 can fit in image

        boxplot(SPVsR_i[,i]~as.factor(samples_i), outline=F, density=c(10,20,300),  las=2, yaxt="n", col=c(brewer.pal(3,"Blues"),
        brewer.pal(3, "Oranges")), density=c(10,20,300), at=c(1,2,3,5,6,7))
        title(colnames(irisTag_i)[i])
    }

我可以制作这张图片：

使用上面相同的三个数据框，我如何使用 ggplot2 来产生类似的结果？

我试过了，但失败了：

library(reshape2)
library(ggplot2)

nsamplesv <- cbind(SPVsR_i, samples_i,deparse.level=2)
nsamplesv_df <- as.data.frame.matrix(nsamplesv)
nsamplesv.m <- melt(nsamplesv_df,id.vars=c('samples_i'))
colnames(nsamplesv.m) <- c("samples",'celltype','score')

p <- ggplot(nsamplesv.m, aes(samples,score)) +geom_boxplot() +facet_wrap( ~ celltype)
p

生成此图像：

我不知道具体如何：

• 修复 y 轴，减少刻度和 3 位小数。
• 框不显示。每个框架应该有 6 个盒子。
• 想给盒子上色。 LAIV 为蓝色，TIV 为黄色。
• 将 x 轴旋转 180 度。

Answer 1

这是重塑数据的更好方法：

library(ggplot2)
library(dplyr)
library(tidyr)
library(RColorBrewer)

as_data_frame(SPVsR_i) %>% 
  bind_cols(data_frame(sample=samples_i)) %>% 
  gather(celltype, score, -sample) %>% 
  mutate(celltype=factor(celltype, levels=unique(celltype))) %>% 
  filter(!(celltype %in% c("Neutrophil"))) -> df

ggplot2 需要将美学映射到数据框或确保引用的变量与您使用的数据具有相同的长度。

我在这里模仿了您想要的输出，但请考虑尊重您的观众，而不是让他们歪着头阅读 X 轴标签。箱线图确实也应该有（IMO）主要的 Y 轴线，这样人脑就更容易解码这些值。这不是 100% 必要的，但（再次）重点是帮助理解。

你也没有注意到你的情节上的免费 Y 轴刻度，我没有在这里，但这也有点可怕。即使使用 Y 轴刻度标签，您也应该做一些事情来确保人们不会尝试以完全相同的比例来比较细胞类型（默认情况下/自动，因为他们正在努力解码情节，但额外的标签如果人们阅读了上述指南，将强制执行额外的处理步骤）。

p <- ggplot(df, aes(sample, score))
p <- p + geom_boxplot(aes(fill=sample))
p <- p + scale_fill_manual(values=c(brewer.pal(3,"Blues"), brewer.pal(3,"Oranges")))
p <- p + facet_wrap(~celltype, scales="free")
p <- p + labs(x=NULL, y=NULL)
p <- p + theme_bw(base_size=10)
p <- p + theme(strip.background=element_blank())
p <- p + theme(strip.text=element_text(face="bold"))
p <- p + theme(axis.text.x=element_text(angle=90, vjust=0.5))
p <- p + theme(panel.grid.major.x=element_blank())
p <- p + theme(panel.grid.major.y=element_blank())
p <- p + theme(panel.grid.minor.y=element_blank())
p <- p + theme(panel.margin=margin(20,20,20,20))
p <- p + theme(plot.margin=margin(20,20,20,20))
p <- p + theme(legend.position="none")
p