[BioC] Can't reproduce assessment data included in affycomp package

[Rafael Irizarry]

Philippe,

Thanks for catching this. And sorry it took so long to reply. We have
been busy fixing the problem. Here is what happened: In 2004 we made a
change to one of the assessment functions. Specifically, the
definition of low, medium and high strate changed for the fold change
assessments. Unfortunately we did not rerun the assessment functions
on the rma and MAS results to fix the pre-computed assessment included
in the package. This is why you couldn't reproduce it. The website
also had a few entries that were run with the old functions. We have
now re-run the current assessment on everything and changed the
precomputed assessments in the package.

http://affycomp.jhsph.edu

is updated. You can see the old ones here:
http://affycomp.jhsph.edu/AFFY2/TABLES.hgu/0.html

Fortunately, things don't change much.

The affycomp package now contains all the new assessments. Those that
have been comparing  a method to the rma or MAS 5.0 precomputed
assessment should re-run the comparison.

Sorry for any trouble this caused.

Best wishes,
Rafael

On Wed, Apr 20, 2011 at 1:57 AM, Philippe Serhal
<philippe.serhal at umontreal.ca> wrote:
> I can't quite seem to reproduce the precomputed MAS 5.0 and RMA
> assessments provided in the affycomp package.  Although I have been
> unable to find the code that was originally used to generate them, I
> think it is fairly safe to assume it looked a little something like this
> (given a current working directory containing the HG-U133A spike-in data
> [1]):
>
> R> library (affy)
> R> library (affycomp)
> R> data <- ReadAffy ()
> R> eset <- rma (data)
> R> write.table (data.frame (2 ^ exprs (eset), check.names = F),
>  + file = "tmp.csv", sep = ",", col.names = NA, quote = F)
> R> tmp <- read.newspikein ("tmp.csv")
> R> a <- assessSpikeIn2 (tmp, method.name = "RMA_testing")
>
> I then compare this with the precomputed assessments:
>
> R> data (rma.assessment2.133)
> R> affycompPlot (rma.assessment2.133, a)
>
> Four of the six resulting figures [2] -- 1b, 2b, 4c, and 5e -- show
> identical results for the two assessments, while the other two -- 5c and
> 5d -- show significantly diverging results.  Thinking perhaps the
> assessments had been computed using an older version of RMA, I tried
> 'bgversion = 1' in the rma() call, but that didn't help.  In fact, this
> doesn't seem to have anything to do with the RMA pipeline, because if I
> repeat the process substituting 'mas5' for 'rma' and
> 'mas5.assessment2.133' for 'rma.assessment2.133' (and not
> exponentiating), I obtain similar results: 5c differs, but none of the
> others.
>
> Any idea what I'm doing wrong here?  Thank you in advance for any
> insight you may be able to provide.
>
> R version: 2.10.0
> affy version: 1.24.1
>
> [1] HG-U133A spike-in data:
> http://www.biostat.jhsph.edu/~ririzarr/affycomp/hgu133spikein.tgz
> [2] affycomp figure definitions:
> http://affycomp.jhsph.edu/AFFY2/comp_form.html#definitions
>
> --
> Philippe Serhal
> Functional and Structural Bioinformatics Lab
> Institute for Research in Immunology and Cancer (IRIC)
> Université de Montréal
>
> _______________________________________________
> Bioconductor mailing list
> Bioconductor at r-project.org
> https://stat.ethz.ch/mailman/listinfo/bioconductor
> Search the archives: http://news.gmane.org/gmane.science.biology.informatics.conductor
>