[BioC] nested design in limma?

[Kevin R. Coombes]

It depends on the overall goal.

If you are only comparing two groups of samples (A and B) and never want 
to compare any of the data with anything else, then it is more efficient 
to use a design that runs A against B on the same slide. In that case, I 
believe some A samples should be run in the Cy3 channel and other A 
samples should be run in the Cy5 channel. Otherwise, you end up 
confounding a technical factor (dye) with the biological contrast of 
interest (A versus B).  As a consequence, any differences you find could 
potentially be explained by dye bias.  In general, I advise people 
against technical replicates and urge them to find more biological 
samples. So, I do not advocate "dye-swapping" in the form of running the 
same sample twice, but I do advocate "dye-swapping" by running some 
samples of each type with each color.

If you have more than two types of samples (either now or in the future) 
or you are interested in finding subtypes within the samples or 
developing classification models or predictors that can be used with 
future samples, then you don't want to run A-versus-B within an array. 
Instead, you should use a reference design.  In that case, you can (and 
probably should) always run the reference in the same channel.  After 
all, you're not interested in how the common reference compares with 
anything; you'll end up comparing the A-vs-reference slides to the 
B-vs-reference slides to see how A differs from B.

A longer discussion can be found in the book "Design and Analysis of DNA 
Microarray Investigations" by Richard Simon et al.

Best,
	Kevin

Georg Otto wrote:
> Gordon Smyth <smyth at wehi.edu.au> writes:
> 
>> Just as an aside, I am continually amazed at how common technical 
>> dye-swaps are. As far as I can see, they just complicate the analysis 
>> to no advantage, yet they have captured the imagination of many 
>> biologists. My guess is that this an attempt to balance the dyes, 
>> although this can be better achieved without introducing technical replication.
> 
> 
> This is an interesting point, since I am going to design and analyse
> two-colour experiments soon and I am interested in what people think
> about it. Do you recommend dye-swapping with biological replicates? Or
> no dye-swapping at all?
> 
> Best,
> 
> Georg
> 
> _______________________________________________
> Bioconductor mailing list
> Bioconductor at stat.math.ethz.ch
> https://stat.ethz.ch/mailman/listinfo/bioconductor