[BioC] question about RMA normalized gene expression values

[Mark Cowley]

hi James A,
i've come across a situation where the probeset distributions can be  
quite different: say you have 10 arrays, and 3 were poor hybs, eg they  
were very bright (which can happen if the RNA is over-fragmented),  
then while the probe distributions will be identical because of the  
qnorm, the probeset distributions will differ quite dramatically  
following the median polish.

If you're seeing quite distinct shapes in the distributions of RMA  
normalised data, i'd be doubting the quality of the original arrays.  
Can you get your hands on the CEL files? that'll tell you for sure.

HTH
mark

-----------------------------------------------------
Mark Cowley, PhD

Peter Wills Bioinformatics Centre
Garvan Institute of Medical Research, Sydney, Australia
-----------------------------------------------------

On 26/01/2010, at 5:59 AM, Bjoern Usadel wrote:

> Hi,
>
> generally in RMA the individual probe distributions are identical,  
> the probe-set distributions are usually only similar.
>
> You can verify this by sampling (w/o replacement) twice n/l  vectors  
> (i.e. probe sets) of length l from the same population of n values  
> (i.e probes) and then using average/median/median polish whatever  
> you like on the vectors and compare the distributions you get.
>
> Best Wishes,
> Björn
>
> James Anderson wrote:
>> Hi James,
>> Thanks for your reply. In the example you provided, the  
>> distribution is ALMOST identical, except in the region with large  
>> intensities. So I think the correct way should be that the  
>> distribution in probe set level from different arrays should be  
>> ALMOST identical, but theoretically, it could deviate from strictly  
>> identical, due to the median polish. The reason I asked this  
>> question is because I actually downloaded one dataset online which  
>> says it has been RMA normalized, but the distribution deviates from  
>> identical more than what I expect, that's why I am suspicious of  
>> the question: do different arrays have identical distribution in  
>> probe set level after RMA? Thanks again,
>> -Jim
>> --- On Wed, 1/20/10, James W. MacDonald <jmacdon at med.umich.edu>  
>> wrote:
>> From: James W. MacDonald <jmacdon at med.umich.edu>
>> Subject: Re: [BioC] question about RMA normalized gene expression  
>> values
>> To: "James Anderson" <janderson_net at yahoo.com>
>> Cc: bioconductor at stat.math.ethz.ch
>> Date: Wednesday, January 20, 2010, 1:04 PM
>> Hi James,
>> James Anderson wrote:
>>> Hi, Do different arrays have identical distribution in probe set  
>>> level after RMA normalization? Since RMA does quantile  
>>> normalization in probe level, the distribution in probe level  
>>> should be identical. Does the median polish summarization (which  
>>> summarizes the expression value from probe level to probe set  
>>> level) make the distribution different from each other?
>> You can test this for yourself.
>> biocLite("affydata")
>> library(affydata)
>> data(Dilution)
>> hist(Dilution)
>> a <- normalize(Dilution)
>> hist(a)
>> eset <- rma(Dilution)
>> plotDensity(exprs(eset))
>> Best,
>> Jim
>>> Thanks,
>>>
>>> -James
>>>
>>>
>>>
>>>           [[alternative HTML version deleted]]
>>>
>>> _______________________________________________
>>> 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
>> -- James W. MacDonald, M.S.
>> Biostatistician
>> Douglas Lab
>> University of Michigan
>> Department of Human Genetics
>> 5912 Buhl
>> 1241 E. Catherine St.
>> Ann Arbor MI 48109-5618
>> 734-615-7826
>> **********************************************************
>> Electronic Mail is not secure, may not be read every day, and  
>> should not be used for urgent or sensitive issues       	 
>> [[alternative HTML version deleted]]
>
> -- 
> --------------------------------------------------
> Björn Usadel, PhD
> Max Planck Institute of Molecular Plant Physiology
> AG Integrative Carbon Biology
> Am Muehlenberg 1
> 14476 Potsdam-Golm
> Tel.: +49 331 5678153
> email usadel at mpimp-golm.mpg.de
> http://tinyurl.com/IntegrativeCarbonBiology
>
> _______________________________________________
> 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