[BioC] Affymetrix quality control and normalization

laurent buffat laurent.buffat at it-omics.com
Tue Jul 27 12:17:43 CEST 2004


For the quality control

Look at affyPLM package and plot the residual of a linear probe model :

And look the "pseudo-image". If the residual are randomly distributed, when
your hybridation is probably good.

Hopes it's help.

L. Buffat

-----Message d'origine-----
De : bioconductor-bounces at stat.math.ethz.ch
[mailto:bioconductor-bounces at stat.math.ethz.ch] De la part de Dipl.-Ing.
Johannes Rainer
Envoyé : mardi 27 juillet 2004 08:36
À : James MacDonald
Cc : bioconductor at stat.math.ethz.ch
Objet : [BioC] Affymetrix quality control and normalization


we got a new affymetrix station and are now becoming a affymetrix core
. as i am fairly new in the one color micro array field i wanted to know how
other people work with affymetrix chips.
at the moment i am normalizing the chips with RMA. i compared these results
MAS5 and GC-RMA background correction. from this comparsions it looked that
worked best (also with only two chips used in the comparsion), GC-RMA made
strange adjustements (i found genes down regulated after GC-RMA background
correction, where they should (must) be up regulated). with MAS 5 i get to
regulated genes, to big variance in the low intensity range... so the method
am using now is RMA.
now to the questions:

a) quality control: how to define when a chip has not worked, when excluding
chip from the analyis? i am looking at the moment at the histogram (big
range or not?) and at the 3'  5' ratio, but where is the limit for this
when was the RNA degraded?

b) RMA with a low number of chips, is this possible? i thins the more chips
(biological replicates) i have the better the model fitting workes.

c) defining regulated genes: i am currently using a M (fold change) cut off
of 1
(2 fold), better solutions?

thanks, jo

Bioconductor mailing list
Bioconductor at stat.math.ethz.ch

More information about the Bioconductor mailing list