小提琴图是【箱线图】与【核密度图】的结合，箱线图展示了分位数的位置，核密度图则展示了任意位置的密度，通过小提琴图可以知道哪些位置的数据点聚集的较多，因其形似小提琴而得名。如下图所示，其外围的曲线宽度代表数据点分布的密度，中间的箱线图则和普通箱线图表征的意义是一样的，代表着中位数、上下分位数、极差等。

最近看单细胞的文章比较多，所以常见小提琴图来展示marker基因。

今天，我们就来学习小提琴图的绘图技巧。

library(ggplot2)

library(ggsignif)

library(patchwork)

用一个简单的测试数据，包含两列：Group包含7个值，剩下一列是Values。

data <- read.table("data1.txt",sep="\t",header=T)

ggplot(data, aes(x = Group, y = Values, fill = Group)) +

geom_violin(position = position_dodge(width = 1), scale = 'width') +

scale_color_manual(values = c("#d1d2d2","#fbd3b9","#a3c6a3","#ccdecc","#c3dbed","#a1c9e5","#417bb9"))+

theme_bw()+

theme(panel.grid = element_blank(),

      axis.text.x = element_text(angle = 45, hjust = 1),

      legend.position = "none")

通过geom_violin就实现了一个比较简单的小提琴图。

ggplot(data, aes(x = Group, y = Values, fill = Group)) +

geom_violin(position = position_dodge(width = 1), scale = 'width') +

geom_jitter(size=0.3) +

scale_color_manual(values = c("#d1d2d2","#fbd3b9","#a3c6a3","#ccdecc","#c3dbed","#a1c9e5","#417bb9"))+

theme_bw()+

theme(panel.grid = element_blank(),

      axis.text.x = element_text(angle = 45, hjust = 1),

      legend.position = "none")

加入抖动点的效果。

ggplot(data, aes(x = Group, y = Values, fill = Group)) +

geom_violin(position = position_dodge(width = 1), scale = 'width') +

#geom_jitter(size=0.3) +

geom_boxplot(position = position_dodge(width = 1), outlier.size = 0.7, width = 0.2, show.legend = FALSE) +

scale_color_manual(values = c("#d1d2d2","#fbd3b9","#a3c6a3","#ccdecc","#c3dbed","#a1c9e5","#417bb9"))+

theme_bw()+

theme(panel.grid = element_blank(),

      axis.text.x = element_text(angle = 45, hjust = 1),

      legend.position = "none")

也可以加入箱图一起。

也可以像前面柱状图一样，加入统计检验。

my_comparisons <- list( c("Mut", "Mut_Gain"), c("Mut", "Mut_LOH"), c("Mut_LOH", "Mut_Gain") )

ggplot(data, aes(x = Group, y = Values, fill = Group)) +

geom_violin(position = position_dodge(width = 1), scale = 'width') +

#geom_jitter(size=0.3) +

geom_boxplot(position = position_dodge(width = 1), outlier.size = 0.7, width = 0.2, show.legend = FALSE) +

scale_color_manual(values = c("#d1d2d2","#fbd3b9","#a3c6a3","#ccdecc","#c3dbed","#a1c9e5","#417bb9"))+

theme_bw()+

theme(panel.grid = element_blank(),

      axis.text.x = element_text(angle = 45, hjust = 1),

      legend.position = "none")+

stat_compare_means(comparisons = my_comparisons)