[BioC] Affy Exon Array Analysis

[Justin Borevitz]

Hi Reinhard and BioC
We've been using affy tiling arrays in Arabidopsis for alternative splicing.
The method is straight forward once exons or transcription subunits (parts
of known alternative exons) are defined.  We first look for differential
gene expression according to the treatment, time or your contrast.  Use your
favorite package, limma, siggenes etc.  Take the residuals from the model.
Now all genes have been corrected for alternative expression, look again at
the contrast for alternative transcription subunits, using your favorite
method.  In our case its about 25k genes and 120k subunits tested for
alternative splicing.  The key is one is limited to get an idea for
alternative splicing within a gene in a single condition given the large
variance in probe effects, what does that mean biologically anyhow.  What
one has power for is alternative splicing across conditions.

Finally there is a semantic argument to be made, what does alternative gene
expression mean if there is alternative splicing? One could put both gene
expression and splicing in a single model, here though one exon would have
to be a reference, others are tested for an exon:treatment interaction
effect.  If that reference exon was alternatively spliced, it would be taken
as the gene expression difference and all the other exons alternatively
spliced.  Any how I just like to fit the models separately, its easier to
think about and explain.  Gene expression is difference in overall means
using all probes(exons). Alternative splicing is a difference across probes
in a subunit.

Hope it helps.. We can provide code but it's in a format for tiling arrays
not yet an affybatch
Justin
-----
Justin Borevitz
Ecology and Evolution (CLSC 915E)
University of Chicago
1101 E. 57th St.
Chicago, IL 60637 USA
(773) 702-5948 office
(773) 834-4055 lab
borevitz at uchicago.edu
http://naturalvariation.org

_____________________________
Dear Group,
Affymetrix recently launched arrays for exon analysis, containing
approx. 1.4 million probe sets. Athough some methods for detection of
alternative splicing events (MIDAS, PAC) have been described in
Affymetrix White Papers, these algorithms do not seem to be
implemented anywhere.
Does anybody have these or other tools for the analysis of Affyon
arrays (detection of alternative splicing events) available, maybe as
R code or so? Any help would be greatly appreciated.
Sincerely Reinhard

-- 
Dr. Reinhard Hoffmann