[BioC] Re: Single channel normalization

[Natalie Thorne]

Hi Christian,

We have just added to limma some functions and documentation for doing
single-channel normalization. Limma is an R package that has been included
in the Bioconductor release package.  Gordon Smyth is the maintainer
for limma and the package is frequently updated. You should go to
http://bioinf.wehi.edu.au/limma/ for the latest release which now includes
the single-channel(sc) normalization functionality and function for
plotting the single-channel densities.

This latest release has also just been posted to Bioconductor to the
Developmental Package of limma at http://www.bioconductor.org/ and will be
available there shortly.

In R use help.start() and go to packages...limma...Accommpanying
documentation...usersguide (in .html or .pdf)

Section 14 of the usersguide is on single-channel normalization for
two-color arrays.  This gives an example of normalizing the
single-channels of the ApoAI data set which is part of sma.  
See also the help examples for the function plotDensities,
normalizeWithinArrays, and normalizeBetweenArrays.  Some of these
examples are given below.


library(limma)
library(sma)
help.start()
data(MouseArray)
							# just bkg corr
MA.n <- MA.RG(mouse.data)				# no normalization

MA.q <- normalizeBetweenArrays(MA.p, method = "q")	# quantile norm

G.q <- normalizeBetweenArrays(RG.MA(MA.n)$G,method="q") # only green sc's
							# takes a matrix			

tmp<-cbind(RG.MA(MA.n)$R,RG.MA(MA.n)$G)[,c(1,3,8,9,12)] # select sc's
tmp.q <- normalizeBetweenArrays(tmp, method = "q")	# takes a matrix		

MA.p <- normalizeWithinArrays(MA.n, mouse.setup)	# default p-loess
MA.pq <- normalizeBetweenArrays(MA.p, method = "q")	# pq norm

MA.MpAq <- normalizeBetweenArrays(MA.p, method = "Aq")	# MpAq norm
							# Yang & Thorne 03


plotDensities(MA.n)					# default

plotDensities(MA.n,arrays=c(1:6),			# same as default
groups=c(rep(1,6),rep(2,6)),col=c("red","green"))

plotDensities(MA.n,arrays=NULL,groups=NULL,		# diff cols
col=c("blue","purple"))

plotDensities(MA.n,singlechannels=c(1,2,7))		# indexing sc's

plotDensities(MA.n,singlechannels=c(1,2,7),		# diff cols
col=c("pink","purple"))

plotDensities(MA.n,singlechannels=c(1,2,7),		# groups=2
col=c("pink","purple","blue"))				# col too long

plotDensities(MA.n,singlechannels=c(1,2,7),		# controlling
groups=c(1,2,3),col=c("pink","purple","blue"))		# col and groups

plotDensities(MA.n,singlechannels=c(1,2,7),groups=c(1,1,1),
col=c("purple"))

#  All single-channels, three groups (ctl,tmt,reference), three colors.
plotDensities(MA.n,singlechannels=c(1:12),
groups=c(rep(1,3),rep(2,3),rep(3,6)),col=c("darkred","red","green"))

#  Densities after single-channel MpAq (preferred) normalization.
plotDensities(MA.MpAq)


     *   *    Natalie Thorne, Phd Student.
   *     *    
   *   *      Genetics and Bioinformatics Division,
     *	      The Walter And Eliza Hall Institute
   *   *      Of Medical Research (WEHI). 
   *     *    PO Royal Melbourne Hospital, 3050
     *   *    PH:    +61 3 9345 2631
       *      Fax:   +61 3 9347 0852 
     *   *    URL http://bioinf.wehi.edu.au
   *     *    
   *   *      Mathematics and Statistics Department,  
     *        The University of Melbourne   
   *   *      Fax:   +61 3 9344 4599
   *     *    
     *   *    
       *      CRC for discovery of genes for common 
     *   *    human diseases.
   *     *
   *   *      Email : thorne at wehi.edu.au

On Tue, 9 Sep 2003, cmprobst wrote:

> Hi, Natalie
> 
> I would like to ask you some question about single channel normalization, mainly the how-to.
> 
> I am asking you for suggestion in how to plot the density functions as
was done in the paper with Jean Yang. 
> 
> And also, if it is possible to apply the quantiles.normalization (and if
yes, how) in a matrix of m spots per n hybs. 
> 
> TIA
> 
> Christian