[BioC] Differences between lmfit model(from limma package) and Illumina-custom Model (from GenomeStudio)

[Gordon K Smyth]

Dear Abdul,

Please keep discussion on the list.

I don't believe that that the Illumina-custom analysis is to be 
recommended, but it is not described in sufficient detail to critique it 
properly.

Generally speaking, vendor-supplied statistical analyses are a convenience 
for some users, but they are not usually subject to peer-review and they 
can't be expected to be as good as the best academic statistical 
treatments of data from the same platforms.  This has been true for 
Affymetrix and other platforms besides Illumina.

Note also that your differential results will depend on preprocessing 
steps including background correction and normalization as well as on the 
statistical test used, and the limma and lumi packages provide 
preprocessing options not available in GenomeStudio.  I personally 
recommend the neqc() function in limma for this purpose.

Best wishes
Gordon

---------------------------------------------
Professor Gordon K Smyth,
Bioinformatics Division,
Walter and Eliza Hall Institute of Medical Research,
1G Royal Parade, Parkville, Vic 3052, Australia.
http://www.statsci.org/smyth

On Fri, 26 Apr 2013, Abdul Rawoof wrote:

> Dear Gorden,
>
> The Genome Studio software also have option for Differential Expression
> analysis (I am using Illumina-custom algorithm).
> I am using Quantile Normalisation+BH(FDR)+Pval 0.05 on both platform but
> the No. of Differential Expressed Genes I got is different from both
> platform.
> http://expression.genomecenter.ucdavis.edu/documents/illumina_normalization_081201.pdf
>
> Please find the attached file of my result from both platform.
>
> Sincerely Thaks,
> Abdul Rawoof
>
>
> On Fri, Apr 26, 2013 at 9:34 AM, Gordon K Smyth <smyth at wehi.edu.au> wrote:
>
>> Dera Abdul,
>>
>> We use GenomeStudio in order to export probe summary profiles and control
>> summary profiles which are then read into the limma package.
>>
>> GenomeStudio doesn't do any downstream statistical analysis, as far as I
>> know, comparable to limma. To see what limma does, see the User's Guide:
>>
>> http://www.bioconductor.org/**packages/release/bioc/**
>> vignettes/limma/inst/doc/**usersguide.pdf<http://www.bioconductor.org/packages/release/bioc/vignettes/limma/inst/doc/usersguide.pdf>
>>
>> Best wishes
>> Gordon
>>
>>  Date: Thu, 25 Apr 2013 10:20:58 +0530
>>> From: Abdul Rawoof <abdul87.edu at gmail.com>
>>> To: bioconductor at r-project.org
>>> Subject: [BioC] Differences between lmfit model(from limma package)
>>>         and Illumina-custom Model(from GenomeStudio).
>>>
>>> Hello everybody,
>>>
>>> I am doing WG-DASL data analysis using GenomeStudio and R/Bioconductor.
>>> Could anyone please able to tell me the major differences in both model
>>> that affect the result.
>>> If any documentation available please tell me.
>>>
>>> Thanks,
>>> Abdul Rawoof

______________________________________________________________________
The information in this email is confidential and intend...{{dropped:4}}