[BioC] RMA/QuantileNormalization results difference between oligo and aroma.affymetrix for Hugene

[Benilton Carvalho]

Dear Mikhail,

I was able to reproduce the issue you reported. The  probeNames()
method in 1.10.3 is missing a sort by fid.

setMethod("probeNames", "GeneFeatureSet",
                 function(object, subset=NULL){
                   res <- dbGetQuery(db(object), "SELECT fsetid FROM
pmfeature ORDER BY fid")[[1]]
                   as.character(res)
                 })

I'll get this fixed now.

b

On Mon, Mar 22, 2010 at 11:04 AM, Benilton Carvalho
<beniltoncarvalho at gmail.com> wrote:
> what's the array you're looking at?
>
> sessionInfo()?
>
> thanks,
> b
>
> On Mon, Mar 22, 2010 at 10:54 AM, Mikhail Pachkov <pachkov at gmail.com> wrote:
>> Dear Benilton,
>>
>> I have got a problem obtaining probe indices along with probe names. My script:
>>
>> library(oligo);
>> workingDir = getwd();
>> celfiles<-list.files(path=workingDir,pattern=".CEL$|.cel$");
>> rawdata=read.celfiles(celfiles);
>>
>> pms = pm(rawdata)
>> rmadata=rma.background.correct(pms)
>> qndata=normalize.quantiles(log2(rmadata))
>>
>> res <- dbGetQuery(db(rawdata), "SELECT fsetid,atom,fid FROM pmfeature")
>> pid=paste(res[,1],res[,2],res[,3],sep=":")
>> rownames(qndata)<-pid
>>
>> colnames(qndata)<-sampleNames(rawdata)
>>
>> However during analysis of the data it looked like probe names were
>> determined wrong. I have tried to use pmindex() to extract "fid" of pm
>> probes which seems to be a list of numbers sorted in ascending order.
>> I do the following:
>>
>> pnames=probeNames(rawdata)
>> length(pnames)
>> [1] 818005
>>
>> pmidx=pmindex(rawdata)
>> length(pmidx)
>> [1] 818005
>>
>> # first value in probe names
>> pnames[1]
>> [1] "7896737"
>>
>> # first value in pm indices
>> pmidx[1]
>> [1] 1056
>>
>> If I check pgf file for probe with index "1056", it belongs to
>> probeset "7981328" not "7896737" as it given in pnames.
>>
>> My question: How to obtain probeset-probe_id pairs in correct order
>> for annotating expression values in "pms" matrix?
>>
>> Best regards,
>>
>> Mikhail
>>
>