[BioC] Factorial design with two-channel microarray data with Limma

James W. MacDonald jmacdon at med.umich.edu
Wed Oct 7 15:58:22 CEST 2009

Hi Eran,

I think you misunderstand what Gordon was getting at. Note that the mean 
value for any gene from either strain will already incorporate the 
difference between pulse and control. So if you use the simplest 
possible design matrix (the so called treatments contrast 
parameterization), then by definition you will get the comparison you 
are looking for.

You could use modelMatrix() to create your design matrix, but I find 
model.matrix to be a bit more intuitive in this case.

 > design <- model.matrix(~factor(rep(1:2, each=2)))
 > colnames(design) <- c("Intercept","104 - 58")
 > design
   Intercept 104 - 58
1         1        0
2         1        0
3         1        1
4         1        1
[1] 0 1
attr(,"contrasts")$`factor(rep(1:2, each = 2))`
[1] "contr.treatment"

In this case the Intercept column represents the mean of the 58 strain 
samples (which in turn represents P[58] - C[58], to use your notation). 
The 104 - 58 column represents the difference between the 104 and 58 
strain, or (P[104] - C[104]) - (P[58] - C[58]), which I believe is what 
you want.



Eran Tauber wrote:
> Dear Gordon, 
> Many thanks for your prompt reply. The User Guide and section 8.1.1 describes 
> a simple comparison between two strains hibridized to the same array. In my 
> experiment there are two factors, strain and treatment where for each strain, 
> the two treatment were hybridized to the same array (see original post). My 
> understanding was that my case resembles more example 8.7 in the manual, and 
> my goal is to identify the genes that respond differently in one strain 
> compared to the other (the interaction term, following the manual, question 3, 
> p.44). The problem is that section 8.7 demonstrates it with Affymetrix data 
> and I'm not sure how to apply it to a 2-channel experiment. Please advise, Eran
> _______________________________________________
> Bioconductor mailing list
> Bioconductor at stat.math.ethz.ch
> https://stat.ethz.ch/mailman/listinfo/bioconductor
> Search the archives: http://news.gmane.org/gmane.science.biology.informatics.conductor

James W. MacDonald, M.S.
Douglas Lab
University of Michigan
Department of Human Genetics
5912 Buhl
1241 E. Catherine St.
Ann Arbor MI 48109-5618

More information about the Bioconductor mailing list