[BioC] General statistical question 2 colour data

Anja Schiel a.e.schiel at lumc.nl
Thu Mar 24 15:27:28 CET 2005

Dear all,

I have some basic statistical questions and which packages might provide
tools to answer those.

I am running R 2.0.1 + Bioconductor under Debian testing.

I did have a look at the archives but did not find any satisfying
answers (or I didn't search the right way) and since I have only a
limited statistical background I need some pointers/help.

I have 4 sets of mice (wt vs treated) that have been hybridized as
random pairs to a total of 8 oligo arrays (each pair twice as a dye swap
pair). Each array is spotted in duplo.

I would like to not only identify genes that are differentially
expressed between the two experimental groups (wt and treated) but I
also want to extract the effect of the treatment from the mouse
individual effect of each mouse-pair(this effect could then be used as a
factor in determining the required number of animals for a new
experiment). I assume that each hybridisation will differ from the other
pairs due to the mouse-(or patient-)effect. Is there an easy way to
extract this information?

Any hint to which package I should use or vignettes that discuss this
issue would be greatly appreciated (if I know where to look for I
usually get things running).

I do not have only on treatment but actually two treatments. The groups
were hybridised in a loop design. I was told that I could only perform
an ANOVA if I split the ratios (is this true?). I was also told this was
dangerous business if I could not guarantee that the correlation between
the individual green and red channels of each experimental group was
"good" enough.

My question is therefore, how can I check the correlation?
Should I run a normalisation on the individual channels of each array or
do I have to test the correlation on raw channel data (which differ quit
a lot, hence the normalization, at least thats what I thought).
What is a good correlation anyhow? Our Oligo-arrays are printed on home
made slides, so quality is always an issue, where do I have to draw a

And, if I have to work with ratios after all is there no other
statistical approach to the loop design?

I fear that these are very basic questions you probably get every now
and then but I would greatly appreciate any help.

Thanks, Anja

Anja E. Schiel, Ph.D.
Departments of General Internal Medicine and Human Genetics
Leiden University Medical Center
PO Box 9503
2300 RA Leiden
The Netherlands 
tel: -31-(0)71-5276067
fax: -31-(0)71-5276075

More information about the Bioconductor mailing list