[BioC] Heatmap.2 scale problems: Sacling inside the function gives different results than scaling outside!!!

[Steve Lianoglou]

Hi,

On Fri, Jul 23, 2010 at 6:44 AM, Benjamin Otto
<b.otto at uke.uni-hamburg.de> wrote:
> Hi Guys,
>
> thanks for confirmation. :-)
>
> I was pretty sure because I have a little modified script of heatmap.2() where I added some features such as scaling "before" clustering where I can choose between z-score or just median centering, or having the signal color legend below or beside the heatmap ... stuff like that. In fact it even has a little module which I snatched from the HeatPlus package (I remember it was called like that) that should provide the possibility to color the dendrogram-clusters according to a given parameter defining either number of clusters of cut-heigt.
>
> If somebody is interested just send me a notification. However I must warn you, that the modified code still has some minor bugs and is far from "package" ready, because I used to code it beside my projects and never had the time to concentrate on giving it a really final shape.

Sure, why not? :-)

Throw it up as a gist, perhaps, and those who like can hack it into shape:
http://gist.github.com/

I actually have my own version of heatmap.2 as well, because I'm
convinced the default arguments in the function definition that deal
with "symmetrical keys" are wrong.

Take for instance the definition of the symbreaks argument:
symbreaks = min(x < 0, na.rm=TRUE) || scale != "none"

I'm pretty sure it should be `symbreaks = min(x, na.rm=TRUE) < 0 ||
...` since the former will always evaluate to 0 (FALSE)
(likewise with symkey).

I emailed the maintainer about it but never got a response, so I've
just been using my own version since.

-- 
Steve Lianoglou
Graduate Student: Computational Systems Biology
 | Memorial Sloan-Kettering Cancer Center
 | Weill Medical College of Cornell University
Contact Info: http://cbio.mskcc.org/~lianos/contact