R:我能用一个使用dendextend的for循环来‘设置’我的树状图吗？

Question

以iris数据集为例，我以以下方式使用dendextend进行Pearson集群：

library(RColorBrewer)
library(dendextend)
data(iris)
newmat <- iris[,1:4]
rownames(newmat) <- paste(iris$Species, rownames(iris))

dmat <- 1 - cor(t(newmat), method="pearson")
dmat <- as.dist(dmat)
clust.obj <- hclust(dmat, method="complete")
dend.obj <- as.dendrogram(clust.obj)
numsamples <- length(rownames(newmat))
maxdist <- max(get_nodes_attr(dend.obj, "height"))

groups <- levels(iris$Species)
cols <- colorRampPalette(brewer.pal(length(groups), "Set1"))
myPal <- cols(length(groups))

vals1 <- grep(groups[1], labels(dend.obj), value=TRUE)
vals2 <- grep(groups[2], labels(dend.obj), value=TRUE)
vals3 <- grep(groups[3], labels(dend.obj), value=TRUE)
vals1B <- grepl(groups[1], labels(dend.obj))
vals2B <- grepl(groups[2], labels(dend.obj))
vals3B <- grepl(groups[3], labels(dend.obj))

dend.obj <- dend.obj %>%
  set("leaves_pch", 19) %>%
  set("leaves_cex", 1) %>%
  set("branches_lty", 2) %>%
  set("by_labels_branches_col", value = vals1, TF_values = c(myPal[1],Inf)) %>%
  set("by_labels_branches_lwd", value = vals1, TF_values = c(2,Inf)) %>%
  set("by_labels_branches_lty", value = vals1, TF_values = c(1,Inf)) %>%
  set("by_labels_branches_col", value = vals2, TF_values = c(myPal[2],Inf)) %>%
  set("by_labels_branches_lwd", value = vals2, TF_values = c(2,Inf)) %>%
  set("by_labels_branches_lty", value = vals2, TF_values = c(1,Inf)) %>%
  set("by_labels_branches_col", value = vals3, TF_values = c(myPal[3],Inf)) %>%
  set("by_labels_branches_lwd", value = vals3, TF_values = c(2,Inf)) %>%
  set("by_labels_branches_lty", value = vals3, TF_values = c(1,Inf)) %>%
  set("labels_colors", ifelse(vals1B, myPal[1], ifelse(vals2B, myPal[2], myPal[3]))) %>%
  set("leaves_col", ifelse(vals1B, myPal[1], ifelse(vals2B, myPal[2], myPal[3])))

png(filename="test.png", height=1200, width=400)
mar.default <- c(5,4,4,2) + 0.1
par(mar = mar.default + c(0, 0, 0, 4))
plot(dend.obj, main="test cluster", xlab="Distance", horiz=TRUE, cex.main=1, cex.axis=1, cex.lab=1)
legend(maxdist, numsamples, groups, cex=1, pch=19, col=myPal)
dev.off()

它产生了这个簇，我发现它对颜色和所有东西都很有用：

​

​

问题是，现在我想将其封装到一个函数中。并且groups的长度可以改变。所以我需要在for循环中完成dend.obj的set部分。

类似于：

for (i in 1:length(groups)){
    set("by_labels_branches_col", value=grep(groups[i],labels(dend.obj),value=TRUE), TF_values=c(myPal[i],Inf)) %>%
    set("by_labels_branches_lwd", value=grep(groups[i],labels(dend.obj),value=TRUE), TF_values=c(2,Inf)) %>%
    set("by_labels_branches_lty", value=grep(groups[i],labels(dend.obj),value=TRUE), TF_values=c(1,Inf))
}

这显然不起作用..。这里的ifelse也是如此，这真的很棘手。

任何帮助都将不胜感激！我不知道如何解决这个问题。谢谢!

Answer 1

好的，如果有人感兴趣，我在最后所做的，似乎工作正常，是执行for循环并将命令存储在execval变量中，然后使用以下命令运行该变量：

eval(parse(text=execval))