heatmap and heatmap.2

capricy · 06-18-2012, 05:17 PM

I plot the same set of data with heatmap and heatmap.2 respectively, and got distinctly different profiles. Anybody could explain this? I know heatmap.2 has extra functions, such as color key; but it looks that its imaging algorithm might also be quite different from heatmap. Is it true? Thanks.

rboettcher · 06-19-2012, 04:02 AM

I noticed the same thing some time ago...

heatmap does not use the same colorkey for every row by default. Instead it performs some weird scaling...
This essentially means that the rows are NOT comparable by means of colour, since the scaling is performed independently for every row.

you can try this very easily with the following commands:

x = c(1:10)
a = cbind(x,x,x,x)
a[1,2] = 100
a[1,3] = 1000
heatmap(a)

Here's what the help says:
"By default (scale = "row") the rows are scaled to have mean zero and standard deviation one. There is some empirical evidence from genomic plotting that this is useful. "

set scale = "none" and your results will be comparable.

heatmap(a,scale="none")
heatmap.2(a,trace="none")

Best regards,
René

capricy · 06-20-2012, 05:47 AM

Thanks for help

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
R - problem in heatmap reading	Chuckytah	Bioinformatics	9	03-15-2012 12:51 PM
help with a heatmap with deseq - legend?	vebaev	Bioinformatics	3	03-03-2012 09:39 AM
RNA-seq, RPKM and heatmap???	sdwy2008	RNA Sequencing	10	07-14-2011 10:41 PM
R: heatmap color palette	akolman	Bioinformatics	5	06-16-2011 11:51 AM
'heatmap' tool	Livi81	General	0	07-14-2010 03:07 PM

06-18-2012, 05:17 PM	#1
capricy Senior Member Location: 63130 Join Date: Apr 2012 Posts: 125	heatmap and heatmap.2 I plot the same set of data with heatmap and heatmap.2 respectively, and got distinctly different profiles. Anybody could explain this? I know heatmap.2 has extra functions, such as color key; but it looks that its imaging algorithm might also be quite different from heatmap. Is it true? Thanks.

06-19-2012, 04:02 AM	#2
rboettcher Member Location: Berlin Join Date: Oct 2010 Posts: 71	I noticed the same thing some time ago... heatmap does not use the same colorkey for every row by default. Instead it performs some weird scaling... This essentially means that the rows are NOT comparable by means of colour, since the scaling is performed independently for every row. you can try this very easily with the following commands: x = c(1:10) a = cbind(x,x,x,x) a[1,2] = 100 a[1,3] = 1000 heatmap(a) Here's what the help says: "By default (scale = "row") the rows are scaled to have mean zero and standard deviation one. There is some empirical evidence from genomic plotting that this is useful. " set scale = "none" and your results will be comparable. heatmap(a,scale="none") heatmap.2(a,trace="none") Best regards, René

06-20-2012, 05:47 AM	#3
capricy Senior Member Location: 63130 Join Date: Apr 2012 Posts: 125	Thanks for help