[R] hclust and cutree: identifying branches as classes

[Peter Langfelder]

On Mon, Sep 12, 2011 at 4:59 AM, Laurent Fernandez Soldevila
<l.fernandez4 at yahoo.fr> wrote:
> Good afternoon,
>
>
> After cuting a hierarchical tree using cutree(), how to check correspondances between classes and branches?
> This is what we do:
>
> srndpchc <- hclust(dist(srndpc$x[1:1000,1:3]),method="ward") #creation of hierarchical tree
> plclust(srndpchc,hmin=20000) #visualisation
>
> srndpchc20000 = cutree(srndpchc,h=20000) #returns 4 classes
> table(srndpchc20000 )
>
> srndclass20000 = cbind(srnd at data[1:1000,],srndpchc20000) #assigning classes to objects
> srndcents20000 <- aggregate(srndclass20000, by=list(srndpchc20000), FUN=median)
> matplot(1:36,t(srndcents20000[,-c(1,38)]))
>
>
> But how can we make sure that, for example, class 1 is the first branch in the tree plotted  by plclust() ?
Hi Laurent,

in the WGCNA package we have a function called plotDendroAndColors
that takes, as the minimum, the hierarchical clustering tree and the
clustering, and plots the tree with colors underneath.

you would use something along the lines of

plotDendroAndColors(srndpchc, srndpchc20000,
                                  "Classes", # Some label for the classes,
                                   dendroLabels = FALSE # To not plot
labels for each leaf
                                  )

An example of the plot, with several different tree cuts, can be seen, e.g., at

http://www.genetics.ucla.edu/labs/horvath/CoexpressionNetwork/BranchCutting/Example-Dendrogram-10.png

HTH,

Peter