[BioC] different gcrma output on Linux and windows

[Vladimir.Morozov at ari.altanapharma.com]

Hi

I   have got different GCRMA results on Windows and Linux. Basicly I run 
the standart GCRMA on three replicate U133A chips and get correlation 0.86 
between the  windows and linux results.

I have the same package versions on linux and windows:
> package.version("gcrma")
[1] "1.1.3"
> package.version("hgu133acdf")
[1] "1.4.3"
> package.version("hgu133aprobe")
[1] "1.0"
> 


On windows
> R.Version()
$platform
[1] "i386-pc-mingw32"
.......


>data <- 
ReadAffy(filenames=c(".............../GP00067.CEL","............/GP00025.CEL","....../GP00057.CEL"),phenoData="PC3.pdata")
> eset.gcrma <- gcrma(data)
Computing affinities.Done.
Adjusting for optical effect...Done.
Adjusting for non-specific binding...Done.
Normalizing
Calculating Expression
>write.table(exprs(eset.gcrma),file="PC3.eset.gcrma",quote=FALSE,sep="\t", 
row.names = TRUE, col.names = TRUE)

copy PC3.eset.gcrma to PC3.eset.w.gcrma on Linux
on linux
> package.version("gcrma")
[1] "1.1.3"
> 
> R.Version()
$platform
[1] "i686-pc-linux-gnu"
........
> data <- 
ReadAffy(filenames=c("......./GP00067.CEL","..../GP00025.CEL","...../GP00057.CEL"),phenoData="PC3.pdata")
> eset.gcrma <- gcrma(data)
Computing affinities.Done.
Adjusting for optical effect...Done.
Adjusting for non-specific binding...Done.
Normalizing
Calculating Expression
#read the windows expression values
> eset.gcrm.w <- read.exprSet("PC3.eset.w.gcrma")
> cor(exprs(eset.gcrma),exprs(eset.gcrm.w))
            GP00067.CEL GP00025.CEL GP00057.CEL
GP00067.CEL   0.8633871   0.8613008   0.8612103
GP00025.CEL   0.8629481   0.8615257   0.8608711
GP00057.CEL   0.8639332   0.8619907   0.8625088

#just correlation function check
> cor(exprs(eset.gcrma),exprs(eset.gcrma))
            GP00067.CEL GP00025.CEL GP00057.CEL
GP00067.CEL    1.000000    0.999486    0.999442
GP00025.CEL    0.999486    1.000000    0.999133
GP00057.CEL    0.999442    0.999133    1.000000

Alspn u133B chip result are different.Thanks in advance for any 
suggestions

Vladimir


	[[alternative HTML version deleted]]