[BioC] two questions about limma

[Gordon Smyth]

>Date: Tue, 6 Mar 2007 11:12:45 +0800 (CST)
>From: zhaodj at ioz.ac.cn
>Subject: [BioC] two questions about limma
>To: bioconductor at stat.math.ethz.ch
>Message-ID: <2959.159.226.67.50.1173150765.squirrel at mail.ioz.ac.cn>
>Content-Type: text/plain;charset=gb2312
>
>Dear members,
>I encounter two problems when analysing the microarray data employing the
>package limma. I am using R ver.2.4.1 and limma ver.2.9.13.
>
>
>-------Question 1-------------
>After the backgroundCorrect step, the software gives a series of warning
>messages as follows:
> > RG.b<-backgroundCorrect(RG,method="normexp",offset=0)
>Corrected array 1
>Corrected array 2
>Corrected array 3
>Corrected array 4
>Corrected array 5
>Corrected array 6
>Corrected array 7
>Corrected array 8
>Corrected array 9
>Corrected array 10
>Corrected array 11
>Corrected array 12
>Corrected array 13
>Corrected array 14
>Corrected array 15
>Corrected array 16
>Corrected array 17
>Corrected array 18
>Warning messages:
>1: Produced NaNs in: log(x)
>2: Produced NaNs in: log(x)
>3: Produced NaNs in: log(x)
>
>I find that some NaNs occur when running the code "out <- normexp.fit(x)".
>I don't know how to deal with the NaNs.If the NaNs do not influence the
>selection of the differentially expressed genes, I will ignore them. If
>not, I seek for your suggestion.

This may be a serious problem, but it's impossible to tell without 
seeing your data. Have you checked your data before and after 
correction? You need to check whether there are actually any missing 
values in the background corrected values. Do the MA plots looks 
sensible? Perhaps the problem is caused by exact zeros in your 
foreground intensities (which theoretically shouldn't be there).

>-------Question 2-------------
>Another problem is from the duplicateCorrelation. The function glmgam.fit
>is from the package statmod.
> >corMA.p<-duplicateCorrelation(MA.p,design,ndups=2)
>1: Too much damping - convergence tolerance not achievable in:
>glmgam.fit(dx, dy, start = start, tol = tol, maxit = maxit, trace = trace)
>2: Too much damping - convergence tolerance not achievable in:
>glmgam.fit(dx, dy, start = start, tol = tol, maxit = maxit, trace = trace)
>3: Too much damping - convergence tolerance not achievable in:
>glmgam.fit(dx, dy, start = start, tol = tol, maxit = maxit, trace = trace)
>4: Too much damping - convergence tolerance not achievable in:
>glmgam.fit(dx, dy, start = start, tol = tol, maxit = maxit, trace = trace)
>
>I don't know the meaning of this warning. Can I ignore them and proceed to
>the subsequent analysis of the differentially expressed genes? Will the
>two shortcomings mentioned in the warnings affect the reliability of the
>differentially expressed genes? If it does,please give some advice.

This is not a problem. See ?duplicateCorrrelation

Best wishes
Gordon

>Thanks to all of you!