[R] Identifying or searching for labels in a hclust/dendrogram/heatmap

[michael watson (IAH-C)]

Hi Sean
 
Thanks for the help, but I really wanted to do this in R :)
 
Any suggestions?
 
Mick

________________________________

From: Sean Davis [mailto:sdavis2 at mail.nih.gov]
Sent: Thu 09/03/2006 1:01 PM
To: michael watson (IAH-C); r-help
Subject: Re: [R] Identifying or searching for labels in a hclust/dendrogram/heatmap






On 3/9/06 7:47 AM, "michael watson (IAH-C)" <michael.watson at bbsrc.ac.uk>
wrote:

> Hi
> 
> Sorry if this is in the help :-S
> 
> I've looked at example(dendrogram) and though it gives some indication of what
> I want, it doesn't do all.
> 
> OK, so here is what I want to do: draw a tree, and then have an action, on
> user-click, to either draw a sub tree or a plot of the data.  I also want
> users to be able to search for a particular label and have it highlighted on
> the tree, say in red, where all the other labels are black.
> 
> Now, the only tree I can use the "user-click" with is an hclust object, with
> the identify.hclust() function.  As far as I know, neither the dendrogram
> objects or the output of heatmap is cut-able in this way.  So I have the
> sub-tree and plot drawing set up on user-click using an hclust object and
> identify.hclust - good :)
> 
> Therefore I am working with hclust objects (which is a shame as the dendrogram
> and heatmap objects look prettier) but I can cope.  How do I then go on and
> highlight a single label, or group of labels, when I have plotted the hclust
> object?  Can I highlight labels on dendrograms and heatmaps 2?
> 
> I can just imagine that after a user has drawn the tree, they will want to
> know where their genes of interest actually lie within that tree.
> 
> I'm happy to share the completed code when I have sorted all of this out

Mick,

I personally would make use of the ctc package (bioconductor) to output data
to Treeview (and/or cluster).  R may be able to do this, but interaction
with graphics still leaves a bit to be desired in R (at least as compared to
a dedicated, specialized Java App), particularly once you get beyond about
50 genes.  Treeview is just much better set up for this kinda stuff.  You
could probably automate the process given, for example, a limma MAList or
fit object, or something like that.

That said, I'd love to see what other people come up with to accomplish what
you are asking.

Sean