[BioC] day-of-hybridisation variability

[kfbargad at ehu.es]

Dear list,

does anyone know of any studies performed to quantify variability due 
to day of hybridisation of affy chips?

we had an experiment where we had 12 conditions and three replicates 
per condition and due to lab reasons, we had to hybridise on three 
different days, as follows:

rep1 for each condition on day 1
rep2 for each condition on day 2
rep3 for each condition on day 3

this way I "can control" variability introduced by day of hyb.

when doing an ANOVA on all 12 conditions I got over 600 d.e.g , 
and for some pairwise comparisons, I obtained just over 1000, which is 
not bad.

But I say "can control" because what worries me is that if I group the 
arrays by day of hyb, I now have 12 reps of "day of hyb" and an 
ANOVA test on the three days gives me 1900 d.e.g!

How can I explain such an amount of variability, is it because I have 
twelve reps of "day" and only three reps of biological condition? I 
would expect this variability to be random, so that very few genes 
would have a d.e. call. I know that my conditions produce changes, 
because I can see 1000 d.e.g on some of the comparisons, but can 
really "day" produce more changes than my experiment?

Thanks in advance for your comments,

David