[BioC] question about RMA and filtering

[Phguardiol at aol.com]

Hi,
I have identified a set of genes that are differentially expressed between 2 
experiments run in duplicates. The aim of the study is to identify genes and 
pathways differentially expressed between a wild type and a mutated cell line. 
I have used the affy package and the RMA function with quantile normalization.
Because of the small number of replicates I have used the LPE test for this 
purpose and the B&H FDR option. Now my concern is that in some cases I will 
have a wide range of fold changes between my 2 experiments for these genes 
starting by ~ 1.1 or -1.1. In addition some fold changes will be > 5 or -5 but the 
expression values will remain below a certain threshold of let say 100. 
I wonder if most of the people from this list would agree to take the 
following criteria for the final selection: absolute value of the fold change >= 1.5 
and at least one of the 2 group experiments with average signal >= 100. I took 
this last value because when I use MAS5.0 the background is always below this 
value in my experiments. Is a signal in RMA of 100 similar to a signal in 
MAS5.0 of 100 ? In other word, is my threshold based on MAS5.0 data useable with 
RMA ? 
My last question is: shall I filter my data with these criteria before 
running the test for significance or after the test has been run ?
thanks for your help
Philippe   

	[[alternative HTML version deleted]]