R heatmap.2 in gplots - can't replicate the plot when dendograms are passed:

Question

I am trying to get a plot generated via heatmap.2, but would like to understand how the clustering is done. So, I tried to replicate the dendograms ahead of the function call, and plot with these. However, the final figures are different. any clue?

HEATMAP 1:

h<-heatmap.2(dat,col=redgreen, trace="none",
          xlab="Samples", ylab="Genes" ,scale="none" )

vs.

HEATMAP 2:

Rowv <- hclust( dist(dat))   #defaults to  method="euclidean" and method="complete") 
Colv <- hclust(dist(t(dat))) #same as above

heatmap.2(dat, Rowv=as.dendrogram(Rowv), Colv=as.dendrogram(Colv),
          col=redgreen, trace="none",
          xlab="Samples", ylab="Genes",scale="none" )

These two would generate different heatmaps. Any clue why?

Answer 1

Try

Rowv <- as.dendrogram(hclust(dist(X)))

Maybe as.dendrogram makes the difference here. R is still a mess^W^W a mystery to me.

Answer 2

I'm pretty late but I'll try. I don't know about heatmap.2 , but heatmap reorders the dendrogram by default.

To get exactly the same results as those you get when not providing the precalculated dendrogram you would have to say (for the row dendrogram):

Rowv <- hclust( dist(dat))
Rowv <- reorder(as.dendrogram(Rowv), rowMeans(dat)) # reorder by the row means of dat <br/>

heatmap(dat, Rowv=Rowv, Colv=Colv=as.dendrogram(Colv),col=redgreen, trace="none",xlab="Samples", ylab="Genes",scale="none" ) 
# leaving everything else as in your original message