[BioC] [heatmap] different distance functions for columns and for rows

[Sean Davis]

On 5/31/06 2:25 AM, "Ron Ophir" <ron.ophir at weizmann.ac.il> wrote:

> I see. I missed it. You mean something like this
> 
> x<-matrix(rnorm(1000,10,2),100,100)
> hc<-as.dendrogram(hclust(dist(x,method="euclidean")))
> hr<-as.dendrogram(hclust(as.dist(cor(x,method="pearson"))))
> x11();hv <- heatmap(x,Rowv=hr,Colv=hc)

Couldn't have (and, indeed, didn't) said it better myself.

Sean



>>>> Sean Davis <sdavis2 at mail.nih.gov> 30/05/2006 22:02:16 >>>
> 
> 
> 
> On 5/30/06 3:56 PM, "Ron Ophir" <ron.ophir at weizmann.ac.il> wrote:
> 
>> I would like to run clustering on columns (samples) and rows (genes)
> and
>> get a plot similar to the output of heatmap(). However I would like
> to
>> apply euclidean distance to samples and pearson distance to genes.
>> Is it possible to do it with heatmap(). If not, is there other
> function
>> that can do it?
> 
> Ron,
> 
> You can pass Rowv and Colv dendrogram objects, as output by hclust,
> for
> example.  By doing so, you can use different distance functions (or
> even
> different clustering algorithms) for the rows and columns.
> 
> Sean