[BioC] gcrma in affylmGUI

James Wettenhall wettenhall at wehi.edu.au
Fri May 28 17:13:41 CEST 2004

Hi John,

> affylmGUI ROCKS!!!

Thanks!  It's great to get some positive feedback as well as the
many inevitable bug reports!

Hopefully, I will implement gcrma in affylmGUI within the next 
few days, but for now, here's a "hacker's method".

After you've loaded your affy data into affylmGUI, click on the
"Evaluate" menu and select "Evaluate R code".  Then paste in the
following commands:

NormalizedAffyData <- gcrma(RawAffyData)
NormalizedAffyData.Available <- TRUE
NormMethod <- "gcrma"
          text="Available (gcrma)",

Then from the "Run" menu, select "Show Text Results Only"
[ These commands will be evaluated in the environment, 
"affylmGUIenvironment". ]

Just normalize with RMA (even though you really want gcrma),
then just paste in the first two lines of code from above:
NormalizedAffyData <- gcrma(RawAffyData)

But then as a penalty for "cheating", affylmGUI will confuse you
by referring to your normalized data as "RMA" data 
(in the tree/drill-down diagram) when in fact it is "GCRMA".

You only need to type in "library(gcrma)" once per affylmGUI
session (in fact once per R session) so you could even get away
with just one line of code:
NormalizedAffyData <- gcrma(RawAffyData)

> Venn diagrams in affylmGUI- what is used as the cutoff for 
> significance in the Venn Diagrams, and can that cutoff be 
> manipulated/changed?

The venn diagrams in affylmGUI use the classifyTestsF function
from limma which uses a default p-value cutoff of 0.01.
Unforunately, I have not yet implemented a way to adjust this
from the GUI via a dialog box.  I have put this on my "TO DO"
list!  Hopefully next week, you will be able to download a new 
version of affylmGUI from http://bioinf.wehi.edu.au/affylmGUI

I could try to explain a "hacking" way to do this, but then that 
would just delay my implementation of the "nice way", so for 
now, I'll at least refer you to the affylmGUI hacker's guide:

See also the limmaGUI developer's guide (cDNA version):
and the limmaGUI (cDNA version) hacker's guide:

The limma package is loaded automatically whenever you load
affylmGUI.  Once the limma package is loaded, you can view
help on classifyTestsF from within R, by typing :

> Final ?- If an obvious defect/scratch is seen on a chip in 
> affylmGUI-what is your opinion on the most appropriate method 
> to dealing with it within R short of tossing out all the data? 
> This is the first time I've actually looked at my chips 
> visually, some of the defects are significant, but I don't 
> want to toss the baby out with the bathwater.

This is not an easy question to answer in general.  If you have
a reasonable number of replicate chips, it may be OK to include
at least one chip with a spatial artifact such as a scratch,
because although the chip by itself may suggest incorrect gene
expression levels/ratios for genes with probes in the scratch
region, hopefully the replicate chips won't have the same
spatial artifacts, so the linear modelling will in a way take
care of this.  Because of the inevitable inconsistency between
chips, the linear model will effectively down-weight your
t-statistics etc. to tell you that you shouldn't be too
confident about your expression measurements for those genes in
the scratch-area (i.e. the results may not be reproducible).
But if the other (non-scratched) chips are reasonably
consistent, it's still possible to see those genes
high-up on your list.

That's just one way of looking at spatial artifacts.  There are
many other ways of assessing and dealing with chip-quality
issues and spatial variation and I'm really not the best person 
to ask about this, so I'd suggest posting to the Bioconductor 
list with a subject line, something like: "affy quality weights" 
and see what sort of response you get.

It's very likely that someone will suggest a method which is
not yet available within affylmGUI (unless you want to "hack").
I'm certainly interested in adding more array-quality weighting
options into affylmGUI and I'd love to hear some suggestsions 
about this from the affy quality experts!

Best regards,

James Wettenhall                                  Tel: (+61 3) 9345 2629
Division of Genetics and Bioinformatics           Fax: (+61 3) 9347 0852
The Walter & Eliza Hall Institute         E-mail: wettenhall at wehi.edu.au
 of Medical Research,                     Mobile: (+61 / 0 ) 438 527 921    
1G Royal Parade,
Parkville, Vic 3050, Australia

More information about the Bioconductor mailing list