[BioC] justRMA() using different versions of affy

[cstrato]

Dear Ben,

Thank you for pointing me to the earlier discussion. As you already 
mentioned this issue has only a minimal effect.

Currently I am trying to compare the results when using only 
median-polish, i.e. w/o background correction and quantile 
normalization. Is the following code correct:

affy <- ReadAffy()
eset <- expresso(affy, bg.correct=FALSE, normalize=FALSE, 
pmcorrect.method="pmonly", summary.method="medianpolish")
affy.mp <- 2^exprs(eset)

When you say "getting it wrong" do you mean that people who have 
computed their data using an older version of "affy" (or are still using 
an older version) were getting it wrong?

BTW, what do you mean with "The "RMA" brand name has been somewhat 
diluted..."
At least my implementation of RMA in XPS has implemented the correct RMA 
algorithm, as verified earlier with older versions of affy. The same 
holds true for APT.

Best regards
Christian


Ben Bolstad wrote:
> I believe this issue has been touched on in the past:
>
> https://stat.ethz.ch/pipermail/bioconductor/2008-December/025678.html
>
> The only thing that should manifest in numerically different expression
> values is that the quantile normalization code handles numerical ties
> slightly differently than it did before. 
>
> As for whether your XPS code, APT and anything else for that matter is
> now getting it "wrong" . That is a decision I leave to independent
> developers. I personally past the point of caring whether other
> implementations of RMA have exact numerical agreement given the same
> data input beyond those I wrote myself (eg the BioC/affy/preprocessCore
> implementation and RMAExpress). The "RMA" brand name has been somewhat
> diluted in its usage in the real world, being applied to all sorts of
> things.
>
> Best,
> Ben
>
>
> On Mon, 2009-06-29 at 23:17 +0200, cstrato wrote:
>   
>> Dear all,
>>
>> Maybe this question has been answered long time ago, but I have just 
>> tested function justRMA() with different versions of package affy and 
>> found the following behavior:
>>
>> The results obtained with version affy_1.14.2 for BioC 2.0 and earlier 
>> versions are different from the results obtained with version 
>> affy_1.16.0 for BioC 2.1 and later versions. Interestingly, affy_1.16.0 
>> is the first version to use the new package preprocessCore.
>>
>> When comparing these versions of affy to the newest version of the 
>> Affymetrix Power Tools (apt-1.10.2) or to xps it turns out that the 
>> results are identical to the old implementation affy_1.14.2, but differ 
>> from newer implementations of package affy.
>>
>> Although the difference is less than 0.5% I would nevertheless like to 
>> know what causes this difference. I have always assumed that only 
>> certain source code has been moved from affy to preprocessCore but did 
>> not expect to see a difference in the results obtained.
>>
>> Do you know what causes this difference and why?
>> Is there a difference in the background calculation, quantile 
>> normalization or median-polish summarization?
>> Or is there a difference in the usage of probes for background 
>> calculation and/or quantile normalization?
>>
>> Best regards
>> Christian
>> _._._._._._._._._._._._._._._._._._
>> C.h.r.i.s.t.i.a.n   S.t.r.a.t.o.w.a
>> V.i.e.n.n.a           A.u.s.t.r.i.a
>> e.m.a.i.l:        cstrato at aon.at
>> _._._._._._._._._._._._._._._._._._
>>
>> _______________________________________________
>> Bioconductor mailing list
>> Bioconductor at stat.math.ethz.ch
>> https://stat.ethz.ch/mailman/listinfo/bioconductor
>> Search the archives: http://news.gmane.org/gmane.science.biology.informatics.conductor
>>     
>
>
>
>