[BioC] Two-color factorial design, common reference

[Gordon Smyth]

At 02:19 AM 26/06/2004, Sean Davis wrote:
>Sorry for another anova-like question, but I have a set of two-color
>experiments with two variables (time (4 and 8 hours) and treatment(two
>levels)), but they are hybed to a common reference (untreated line).  Does
>this design look correct (for use in limma)?

No it's not.

Hint: you can treat your common reference design exactly as if it was an 
affy experiment. See Sections 7.2 and 7.3 of the Limma User's Guide. This 
is a 2x2 factorial experiment. Go back through the archives of this list 
for answers on affy factorial experiments.

One thing to remember: for an affy or common reference experiment, the 
number of columns in the design matrix is equal to the number of distinct 
groups in your experiment. You have 4 groups.

Gordon

>Ref 4hr  8hr    TrtA  Trtb   Treated
>-1   1    0     1      0    1
>-1   1    0     1      0    1
>-1   1    0     0      1    1
>-1   1    0     0      1    1
>-1   0    1     1      0    1
>-1   0    1     1      0    1
>-1   0    1     0      1    1
>-1   0    1     0      1    1
>
>I am interested in knowing genes differentially expressed in the treated
>cells, treatment A vs. treatment B, 4hr vs. 8hr.  I would do contrasts
>something like:
>
>MakeContrasts(Treated,4hr-8hr,Trta-Trtb,levels=design)
>
>When I do the 4hr-8hr contrast, for example, does this "control" for the two
>different treatments?  In other words, if there are different magnitude of
>effects due to the treatments, but the genes changing are the same at 4 and
>8 hrs, will the 4hr-8hr still be significant?
>
>Thanks,
>Sean