17

我正在尝试创建一个树状图,我的样本是否有 5 个组代码(作为样本名称/物种/等,但它是重复的)。

因此,我有两个问题需要很大的帮助:

  • 如何在叶标签(而不是样本编号)中显示组代码?

  • 我希望为每个代码组分配一种颜色并根据它为叶子标签着色(可能会发生它们不在同一个进化枝中,这样我可以找到更多信息)?

是否可以使用我的脚本来执行此操作(ape 或 ggdendro):

sample<-read.table("C:/.../DOutput.txt", header=F, sep="")
groupCodes <- sample[,1]
sample2<-sample[,2:100] 
d <- dist(sample2, method = "euclidean")  
fit <- hclust(d, method="ward")
plot(as.phylo(fit), type="fan") 
ggdendrogram(fit, theme_dendro=FALSE)  

替换我的 read.table 的随机数据框:

sample = data.frame(matrix(floor(abs(rnorm(20000)*100)),ncol=200))
groupCodes <- c(rep("A",25), rep("B",25), rep("C",25), rep("D",25)) # fixed error
sample2 <- data.frame(cbind(groupCodes), sample) 
4

2 回答 2

18

这是一个使用一个名为“ dendextend ”的新包来解决这个问题的解决方案,该包正是为这种事情而构建的。

您可以在以下 URL 的“使用”部分中的包的演示文稿和插图中看到许多示例:https ://github.com/talgalili/dendextend

这是这个问题的解决方案:(注意如何重新排序颜色以首先适合数据,然后适合树状图的新顺序的重要性)

####################
## Getting the data:

sample = data.frame(matrix(floor(abs(rnorm(20000)*100)),ncol=200))
groupCodes <- c(rep("Cont",25), rep("Tre1",25), rep("Tre2",25), rep("Tre3",25))
rownames(sample) <- make.unique(groupCodes)

colorCodes <- c(Cont="red", Tre1="green", Tre2="blue", Tre3="yellow")

distSamples <- dist(sample)
hc <- hclust(distSamples)
dend <- as.dendrogram(hc)

####################
## installing dendextend for the first time:

install.packages('dendextend')

####################
## Solving the question:

# loading the package
library(dendextend)
# Assigning the labels of dendrogram object with new colors:
labels_colors(dend) <- colorCodes[groupCodes][order.dendrogram(dend)]
# Plotting the new dendrogram
plot(dend)


####################
## A sub tree - so we can see better what we got:
par(cex = 1)
plot(dend[[1]], horiz = TRUE)

在此处输入图像描述

于 2013-09-16T16:01:56.873 回答
12

您可以将hclust对象转换为 adendrogram并用于?dendrapply修改每个节点的属性(颜色、标签等属性),例如:

## stupid toy example
samples <- matrix(c(1, 1, 1,
                    2, 2, 2,
                    5, 5, 5,
                    6, 6, 6), byrow=TRUE, nrow=4)

## set sample IDs to A-D
rownames(samples) <- LETTERS[1:4]

## perform clustering
distSamples <- dist(samples)
hc <- hclust(distSamples)

## function to set label color
labelCol <- function(x) {
  if (is.leaf(x)) {
    ## fetch label
    label <- attr(x, "label") 
    ## set label color to red for A and B, to blue otherwise
    attr(x, "nodePar") <- list(lab.col=ifelse(label %in% c("A", "B"), "red", "blue"))
  }
  return(x)
}

## apply labelCol on all nodes of the dendrogram
d <- dendrapply(as.dendrogram(hc), labelCol)

plot(d)

在此处输入图像描述

编辑:为您的最小示例添加代码:

    sample = data.frame(matrix(floor(abs(rnorm(20000)*100)),ncol=200))
groupCodes <- c(rep("A",25), rep("B",25), rep("C",25), rep("D",25))

## make unique rownames (equal rownames are not allowed)
rownames(sample) <- make.unique(groupCodes)

colorCodes <- c(A="red", B="green", C="blue", D="yellow")


## perform clustering
distSamples <- dist(sample)
hc <- hclust(distSamples)

## function to set label color
labelCol <- function(x) {
  if (is.leaf(x)) {
    ## fetch label
    label <- attr(x, "label")
    code <- substr(label, 1, 1)
    ## use the following line to reset the label to one letter code
    # attr(x, "label") <- code
    attr(x, "nodePar") <- list(lab.col=colorCodes[code])
  }
  return(x)
}

## apply labelCol on all nodes of the dendrogram
d <- dendrapply(as.dendrogram(hc), labelCol)

plot(d)

在此处输入图像描述

于 2013-09-14T15:05:47.807 回答