[BioC] Re: follow-up: How to "merge" cel.containers ?

[Laurent Gautier]

On Mon, May 06, 2002 at 12:01:58PM +0200, julien.sylvestre@wotan.ens.fr wrote:
> - The union.container.cel function works quite well
> - Sorry but I still have a small problem and can't find info in your help pages or pdf doc : 
> It seems that generateExprSet has no default method and needs the user to specify one.

Yep... it has none.
(and the 'generateExprSet.methods' described in the documentation needs a fix)
currently they are:
"avgdiff", "playerout", "liwong"


> As previously told, liwong doesn't work for a matrix dimension incompatibility reason.


Yep... sorry but I did not have time to investigate it (especially since it seems data specific
and I do not have your data (and you are probably not ready to give all of them away)).

Otherwise, Rafael told me that Cheng Li advices to have at least 25 arrays to use the method.

(I'll send you a snippet of code to run and identify more precisely where the problem comes from). 


> What is the name short name for R. Izarry et al. method ? I don't like avgdiff too much so I would like to use this one or LiWong's.


Unlike for 'normalize' the two approaches (via Plob and via Cel.container) are not completely merged (yet)
for expression value computation from the probe level data.
Slight difference remain between them. The next version of affy will offer a better integrated system for that.

Using the default set by Rafael for express consists in using the approach of background correction
by Ben Bostad ( bg.adjust() ) to get a value for each probe from the pm and mm values, 
then applying Li and Wong model fitting to it. By looking at how works bg.adjust() and
generateExprVal.method.liwong() you can come up with the same through the Cel
(I am missing time to write the code now).



L.