[BioC] limma single-channel analysis and intraspotCorrelation

Na, Ren Na at uthscsa.edu
Wed Nov 3 17:30:23 CET 2004

I have a data set of four slides (two pairs of dye swap),
slide#    cy3   cy5
slide1    wo    wr
slide2    wr    wo
slide3    mo    mr
slide4    mr    mo

wo: wildtype without treatment
wr: wildtype with treatment
mo: mutant without treatment 
mr: mutant with treatment

I have been thinking if it is possible to get lists of differentially expressed genes between wr and wo, genes between mr and mo, genes between mr and wr, genes between mo and wo, and genes responding to treatment differently in mutant compared to wild-type. I know factorial design is the better way to go. But based on the data I have now, can I use single channel analysis to get comparison of interest? What I tried and what I will do are like the following,

targets2 <- readTargets("Targets2.txt")
RG<-read.maimages(targets2$FileName, source="spot",wt.fun=wtarea(100))
RG$genes<- readGAL("Mouse24052004_635final2.txt")
RG$genes$Status<- controlStatus(spottypes, RG$genes)
# assign weight 0 to missing spots
w<-modifyWeights(RG$weights,status=RG$genes$Status, "miss",0)
MA<-normalizeBetweenArrays(MA, method="quantile")
targets2.sc <- targetsA2C(targets2)

# got error from intraspotCorrelation()
# what I will do
Then use topTable() to get lists of differentially expressed genes for each of five contrasts.

Here are my questions,

1) Is it reasonable?

2) When I run to function intraspotCorrelation, I got the error,
> corfit<-intraspotCorrelation(MA,design.sc)
Loading required package: statmod

Attaching package 'statmod':

        The following object(s) are masked from package:limma :

         matvec vecmat

Error in chol(ZVZ + lambda * I) : the leading minor of order 1 is not positive definite
In addition: Warning message:
reml: Max iterations exceeded in: remlscore(y, X, Z)
> traceback()
3: chol(ZVZ + lambda * I)
2: remlscore(y, X, Z)
1: intraspotCorrelation(MA, design.sc)
What could cause this error?

3) If I can not get lists of differentially expressed genes between mr and wr, genes between mo and wo, and the genes which respond to treatment differently in mutant compared to wildtype, I still like to get differential expression between mr and mo by doing the following,
fit<-lmFit(MA, design)
and also I can do the same thing to get differential expression between wr and wo.
But my concern is that if the number of replicates is too small and the result won't be reliable?

Thanks for any help and Thank Dr. Gordon Smyth for new limma user's guide


	[[alternative HTML version deleted]]

More information about the Bioconductor mailing list