R语言学习指南(6) 初探相关性热图

相关性热图，顾名思义就是根据数据之间的相关性系数来绘制热图，可分为2类：组内相关性热图与组间相关性热图

组内相关性热图，即由一组数据内部的相关性系数绘制而成，废话不多说直接看代码

相关性矩阵可视化

library(tidyverse)
library(corrplot)
cor(mtcars) %>% corrplot(method = "circle",order = "hclust",
type = "lower",tl.srt = 45,tl.col = "black")

正相关以蓝色显示，负相关以红色显示。颜色强度和圆圈的大小与相关系数成正比

full：显示完整的相关矩阵
upper：显示相关矩阵的上三角
lower：显示相关矩阵的下三角

将相关图与显着性检验相结合

cor.mtest( )函数计算P值

res1 <- cor.mtest(mtcars)

sig.level = -1显示所有p值

cor(mtcars)%>% corrplot(type="lower", order="hclust",tl.srt = 45,
                        tl.col = "black",
                        p.mat = res1$p,insig = "p-value",sig.level = -1)

显示不显著点的p值

cor(mtcars)%>% corrplot(type="lower", order="hclust", 
                        p.mat = res1$p,insig = "p-value")

将不显著的点用空表示

cor(mtcars)%>% corrplot(type="lower",order="hclust",
                        p.mat = res1$p,insig = "blank")

在p值> 0.05的点上打X

cor(mtcars)%>% corrplot(type="lower",order="hclust",
                        p.mat = res1$p, sig.level = .05)

在p值> 0.01的点上打X

cor(mtcars)%>% corrplot(type="lower",order="hclust",
                        p.mat = res1$p, sig.level = .01)

将p值转化为*添加于图上

cor(mtcars) %>% corrplot(type="lower", order="hclust",
p.mat = res1$p,insig = "label_sig",
           sig.level = c(.001, .01, .05),
           pch.cex = .9, pch.col = "white",
           tl.srt = 45,tl.col = "black")

ggplot2绘制组内相关性热图

corr.test( )会同时计算p值与相关性系数，通过for循环将p值转化为*

library("reshape")
library("psych")

p <- corr.test(mtcars,method="pearson",adjust = "fdr")
cor <- p$r
pvalue <- p$p
display <- pvalue
l1 <- nrow(display)
l2 <- ncol(display)
for(i in 1:l1){
  for(k in 1:l2){
    a <- as.numeric(display[i,k])
    if(a <= 0.001){
      a <- "***"
    }
    if( 0.001 < a && a <= 0.01){
      a <- "**"
    }
    if(0.01 < a && a < 0.05){
      a <- "*"
    }
    if(a >= 0.05){
      a <- ""
    }
    display[i,k] <- a
  }
}

构建ggplot2绘图文件并导出

heatmap <- melt(cor) %>% rename(replace=c("X1"="sample1","X2"="sample2",
                               "value"="cor")) %>%
  mutate(pvalue=melt(pvalue)[,3]) %>%
  mutate(display=melt(display)[,3])

write.table(heatmap,file ="heatmap.xls",sep ="\t",row.names = F)    

ggplot2可视化

ggplot(heatmap,aes(sample1,sample2,fill=cor))+
  geom_tile()+
  theme_minimal()+
  scale_fill_distiller(palette = "Spectral")+
  geom_text(aes(label=display),size=5,color="white")+
  scale_y_discrete(position="left")+xlab(NULL) + ylab(NULL)+
  labs(fill ="expr")

下一节再谈组间相关性热图