[BioC] DSPAM Re: Loading multiple Affy chips simultaneously

[lgautier at altern.org]

> I just realized I forgot to give some basic information on my system:
> R 2.4.1
> affy 1.12.2
> gcrma 2.6.0
> Windows XP
>
> Paul
>
> -----Original Message-----
> From: Paul Boutros [mailto:Paul.Boutros at utoronto.ca]
> Sent: Saturday, January 27, 2007 1:59 PM
> To: 'bioconductor at stat.math.ethz.ch'
> Subject: Loading multiple Affy chips simultaneously
>
> Hello,
>
> I'm trying to analyze & merge some rat data that was generated on two
> separate Affymetrix chips (RAE230-A and RAE230-2).  One chip (RAE230-2) is
> a
> superset of the other.  Approximately equal numbers of animals were
> analyzed
> with the smaller array (n=9) as the larger (n=10).
>
> I would like to:
> a) pre-process (via GC-RMA) the two datasets separately
> b) subset the summarized expression levels of the larger chip
> c) merge the subsetted dataset with the smaller one
> d) perform statistical analysis on the overall data (e.g. with limma)
>
> My code loads the data like this:
> data.batch1 <- ReadAffy(filenames=cel.files.batch1,
> phenoData="phenodata_batch1.txt");
> data.batch2 <- ReadAffy(filenames=cel.files.batch2,
> phenoData="phenodata_batch2.txt");
>
> Then does some diagnostics such as:
> par( mfrow = c(1,2) );
> hist(data.batch1, main = "Batch 1");
> hist(data.batch2, main = "Batch 2");
>
> Which generates the warnings:
> Attaching package: 'rat2302cdf'
>
>         The following object(s) are masked from package:rae230acdf :
>          i2xy
>
>         The following object(s) are masked from package:rae230acdf :
>          xy2i
>
> I was not sure if this is a problem or not.  I later go on to use the
> functions AffyRNAdeg and gcrma, and I am wondering if this masking will
> interfere with their correct operation?  In other words, do I need to
> process the two batches sequentially, or can I do it in parallel as I've
> shown here?


The "xy2i"-thing is a hair-rising concept: each cdf package defines
a function i2xy and xy2i that will convert index to xy coordinates on
the chip, and xy to index respectively. As you see, the function is
masked through successive loadings of different cdf packages
(and I will not need to elaborate on the consequences whenever the
different chips have different number of row and columns).

The "affy" package stuck to the safer function "indices2xi" and
"xy2indices", but there is no warranty that extension to the package
do as well. Scanning the code in the packages "AffyRNAdeg" and "gcrma"
(and all the packages they depend on), looking for xy2i and i2xy is one
thing you may want to do.
I have tried very hard to have this dangerous thing dropped... may this
email help to have it moved away for good.


Hoping this helps,


Laurent




> Many thanks for any help,
> Paul
>