[BioC] effect of normalization on class prediction?

[Jenny Drnevich]

Hello,

I have a question that may be so basic in class prediction that I can't 
even find any discussion on it anywhere: What effect does between-array 
normalization of microarray data, particularly those methods that normalize 
to a particular group of arrays, have on class prediction algorithms? Based 
on my scant knowledge of class prediction with microarray data and the few 
details I found in the literature, prediction algorithms are generated from 
a set of data that is variously divided into training and test tests, and 
possibly subdivided as well... but this entire set of data is likely to be 
normalized together. How should a future, independent set of arrays, or in 
particular, a single array, be handled in regards to normalization before 
the prediction algorithm is applied? Do others see that this is a potential 
problem, or am I imagining it?

I hope this question is in the scope of the Bioconductor list, but if not, 
would someone please direct me to a person or place that could answer my 
question?

Thank you,
Jenny


Jenny Drnevich, Ph.D.

Functional Genomics Bioinformatics Specialist
W.M. Keck Center for Comparative and Functional Genomics
Roy J. Carver Biotechnology Center
University of Illinois, Urbana-Champaign

330 ERML
1201 W. Gregory Dr.
Urbana, IL 61801
USA

ph: 217-244-7355
fax: 217-265-5066
e-mail: drnevich at uiuc.edu