[BioC] normalization of duplicate spots

[Jianping Jin]

Dear BioConductor list:

I have a set of one color (cy3) data in which each gene was spotted 
duplicate in separate blocks (grids). I looked at how well the duplicated 
genes were correlated by using either "duplicateCorrelation" (limma) or 
regular "Pearson" correlation test. Both tests gave out similar values 
around 0.55.

Duplicate spots should be highly correlated in intensities. The correlation 
coefficient of 5.5 is not that great as I understand. Limma has some ways 
to process those technical replicates, such as "print-tip loess" and 
incorporating estimated spatial correlation into analysis using generalized 
linear model. But what if I just wanted to preprocess data with something 
like vsn or IQR normalization without using linear model fit? Should I 
normalize the grids, e.g. grids 1 and 3, grids 2 and 4 (which contain 
corresponding duplicate spots) on each chip to make duplicate spots more 
correlated?

Usually I just get mean values of the duplicate spots for such a small data 
sets (722 genes on each slide). Does anybody know a better way to handle 
these duplicates than using means?  I will greatly appreciate if anyone 
could share her/his experience with me!


Thanks in advance!

Jianping

xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
x Jianping Jin Ph.D.            x
x Bioinformatics scientist      x
x Center for bioinformatics     x
x 3133 Bioinformatics Building  x
x CB# 7104                      x
x University of North Carolina  x
x Chapel Hill, NC 27599         x
x Tel: (919)843-6105            x
x Fax: (919)843-3103            x
x E-mail: jjin at email.unc.edu    x
xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx