[BioC] Seeking help to remove control spots in single channel normalization

Prashantha Hebbar prashantha.hebbar at manipal.edu
Thu Jan 8 13:39:05 CET 2009


Dear Lee,

I am thankful to your response. Yes, as you said control information will be
there in ControlTypes column of featured extract data. But the problem is
when I give following command for background correction

Ggene<-backgroundCorrect(RG$G,method='normexp')    #I referred this command
in mail list

Ggene variable will not have RG$genes information (i.e. gene list info). If
I get RG$genes information in next steps (after inter array normalization)
then only it is possible to remove control spots. 

So, now my question is If I use directly RG in my background correction (for
example: Ggene<-backgroundCorrect(RG,method='normexp')), will it create any
problem? In this case, RG$G-RG$Gb and RG$R-RG$Rb will be done right? So,
analysis is meaningful right?

If you make me clear in this I think I will not have any further problem in
removing Control spots after inter array normalization. Sorry for making you
to write me again.

Regards,

Prashantha

-----Original Message-----
From: Leon Yee [mailto:yee.leon at gmail.com] 
Sent: Monday, January 05, 2009 6:37 PM
To: Prashantha Hebbar Kiradi [MU-MLSC]
Cc: bioconductor at stat.math.ethz.ch
Subject: Re: [BioC] Seeking help to remove control spots in single channel
normalization

Hi,

Prashantha Hebbar Kiradi [MU-MLSC] wrote:
> Dear Friends,
> 
> I am working on single channel normalization for Agilent 244K chip data
using Limma package. I refered the communication happend between Gordon and
Abhilash about single channel normalization. I able perform it. But, not
able to get the gene list soon after the normalization as we get in dual
channel analysis.
> 
> I am able to get the gene list in topTable stage. But I do not want gene
list at the end of the analysis. Because I want remove the control spots
soon after the normalization as we do for dual color.
> 
> Following are the steps which I followed to perform single channel
normalization,
>> library(limma)
>> target<-readTargets("/home/mlscrh2/MData/target.txt")
>> RG<-read.maimages(target$FileName, source="agilent",
path="/home/mlscrh2/PrakrathiData", columns = list(G="gMeanSignal",
Gb="gBGMeanSignal",R="gMedianSignal",Rb="gBGMedianSignal"),annotation=
c("Row", "Col", "ProbeUID","ProbeName", "GeneName"))
>> Ggene<-backgroundCorrect(RG$G,method='normexp')
>> MA.q <- normalizeBetweenArrays(Ggene, method="quantile")
>> design = c(1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1)
>> fit = lmFit(MA.q,design)> ebFit <- eBayes(fit)
>> a <- topTable(ebFit,genelist = RG$genes,adjust="fdr",n=300000)
> 
> I tried to incorporate genelist in the stage of background correction and
normalization, but ends up with an error.
> So can you please suggest me, How to remove control spots soon after
normalization in single channel analysis?

The Feature Extraction file of Agilent array will contain a column 
called "ControlType", so if you use read.maimages with
annotation= c("Row", "Col", "ProbeUID","ProbeName", "GeneName", 
"ControlType"), you can filter out the control spots by "ControlType":
0 means non-control.

HTH

Leon


######################################################################
Attention:\ This e-mail message is privileged and confid...{{dropped:7}}



More information about the Bioconductor mailing list