[BioC] Imitating Affy's own steps

[cstrato]

Since Henrik has kindly mentioned the comparisons I did between affy, 
APT and xps let me make some comments, too:

If you want to replicate as close as possible the results obtained with 
the Affymetrix software, i.e. the original "GCOS" software or the newer 
"Expression Console", respectively, then you need to use the command 
line tool APT, which you can freely download from Affymetrix. As Figure 
4 of my vignette shows, the results obtained with APT are almost 
identical to GCOS, while both affy and xps give slightly different 
results for a couple of probesets.

Although the reason for the difference between xps and APT is not 
completely clear to me there are three known differences:

1, Looking at the source code of APT you will see that all computation 
is done using "float", i.e. 32 bit, while in xps all computation is done 
using "double", i.e. 64 bit.

2, Some CEL-files contain [MASKS] and [OUTLIERS] sections which list 
probes to be excluded from further calculation. These probes are not 
excluded in xps.

3, For background computation a 4x4 grid is used. The center for each 
zone is determined slightly different in APT and xps.

BTW, you can find the code used to obtain Fig.4 in 
xps/examples/script4xps2apt.R.

Best regards
Christian


Henrik Bengtsson wrote:
> I won't answer you question, because I don't know the answer, but I'd
> like to add that there are a few methods out there with the same
> names/labels.   I don't know of many carefully done studies showing
> that the are numerically or near-numerically reproducible.  If any
> know of such, please share them here.  Though I recently discovered:
>
> Christian Stratowa, Introduction to the xps Package: Comparison to
> Affymetrix Power Tools (APT), April, 2010 (vignette of the xps
> package)
>
> which seems to do a quite serious assessment of the methods
> implemented in the xps package - it includes a comparison of
> Affymetrix's MAS5.0 implementation with that of the xps package and
> the affy package.
>
> /Henrik
>
> On Mon, Apr 12, 2010 at 9:22 AM, Timothy Wu <2huggie at gmail.com> wrote:
>   
>> On Mon, Apr 5, 2010 at 10:24 PM, James W. MacDonald
>> <jmacdon at med.umich.edu>wrote:
>>
>>     
>>> Hi Timothy,
>>>
>>> AFAIK, you have two choices. Using the affy package, you can use mas5() and
>>> mas5calls(). You can also use justMAS() from the simpleaffy package.
>>>
>>> I don't believe either one will give you exactly the same results. For the
>>> affy package see
>>>
>>> http://bmbolstad.com/misc/MAS5diff/Mas5difference.html
>>>
>>> for an explanation of the differences.
>>>
>>> Best,
>>>
>>> Jim
>>>
>>>
>>>       
>> Thanks for the reply. It looks pretty close.
>>
>> So is the expresso() command listed on the page
>> *
>> eset <- expresso(Data,
>> bgcorrect.method="mas",normalize=FALSE,pmcorrect.method="mas",summary.method="mas")
>> *
>>
>> be the same as the justMAS()?
>>
>> Thanks in advance.
>>
>> Timothy
>>
>>        [[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
>>
>>     
>
> _______________________________________________
> 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
>
>