[BioC] heatmap Clustering help, two class (Control vs Exp) Experiments,

[Saurin D. Jani]

Hi Steve,


May be something like this:

 d  d  u   | u   u  d n geneY
 u  u  d   | u   u  d n geneX
----------------------
c1 c5 c10  | e1 e4 e8 e2


where: d = down, u = up, n = normal , c= control , e = exprimental

original expression set has this order:
c1,c2,c3...c10 e1,e2,e3..e10

Thank you so much,
Saurin 


--- On Wed, 11/3/10, Steve Lianoglou <mailinglist.honeypot at gmail.com> wrote:

> From: Steve Lianoglou <mailinglist.honeypot at gmail.com>
> Subject: Re: [BioC] heatmap Clustering help, two class (Control vs Exp) Experiments,
> To: saurin_jani at yahoo.com
> Cc: "Bioconductor Bioconductor" <bioconductor at stat.math.ethz.ch>
> Date: Wednesday, November 3, 2010, 12:56 PM
> Hi Saurin,
> 
> On Wed, Nov 3, 2010 at 12:43 PM, Saurin D. Jani <saurin_jani at yahoo.com>
> wrote:
> > Hi,
> >
> > You are right but When I do this:
> >
> >
> heatmap.2(FeatureX,col=gmpalette,Colv=as.dendrogram(hclust(col.dist,method="average")),
> Rowv=as.dendrogram(hclust(row.dist,method="average")),scale="row",key=TRUE,keysize=0.60,symkey=FALSE,density.info="none",trace="none",margins=c(5,MapMargin),cexRow=1,cexCol=1,cex.sub=1);
> >
> > my control and exp. samples get mixed up..!! is there
> anyway I can pass a parameter ..not to do that just cluster
> samples on control and then exp. so, sorted view will be
> there.
> 
> But why would you cluster the samples to begin with, if you
> just want
> to reorder them in some (your) arbitrary way?
> 
> Assuming your data is properly nomralised, etc. and
> clustering your
> samples "mixes them up," then the heatmap is showing you
> visually that
> your treatment examples aren't "strikingly different" than
> your
> controls. Your data is trying to tell you that (apparently)
> all of
> these experiments kind of look (expression wise) like each
> other.
> 
> Maybe that's telling you something about the quality of
> your data, or
> its annotation?
> 
> Maybe you can try the plotPCA function in the affycoretools
> package as
> another way to see how your experiments "cluster
> together".
> 
> I'm not sure that it would change things, but what happens
> if you
> remove all probes w/ low variance across your entire
> dataset and
> re-cluster them?
> 
> > May be something like this: cluster control samples
> then exp. samples and then cluster based on Signal
> Intensity. so, I keep the order ctrl1,ctrl5,ctrl6,ctrl2,...
> and then Exp1,Exp5,Ex2,Exp10 ....so on...
> 
> But then this is kind of misrepresenting what one would
> expect to see
> in such a plot .. you could, of course, plot and save
> heatmaps over
> just your control data, then again with just your
> experiment, then
> photoshop them together, but ... what's the point?
> 
> I guess the question is: what are you trying to show in the
> heatmap
> you are trying to  produce?
> 
> Given that, people might be able to then suggest things you
> could try.
> 
> -- 
> 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
>