[BioC] about low B values and inconsistent experiments: array quality issues?

[J.delasHeras at ed.ac.uk]

This is not strictly a BioConductor question, but a more general one... 
I hope it's not too off-topic.

I am using Limma/LimmaGUI for the analysis of my two-colour expression 
arrays (cDNA).

I have become a "fan" of B values as a tool to rank genes according to 
their likelihood that they're differentially expressed. Rather than 
deciding an arbitrary cut-off point (B, P or fold-based) I choose how 
many genes to work depending on how many I can reasonably deal with, 
taking into account their ranking according to B.

I have just come to a problem, and I was wondering what other more 
experienced people here think.

The story is like this:

1) Using a human 10k array, I did several comparisons between cell 
lines. I was getting B values of up to 14-15. Everything look good.

2) We got a new array, a 22k one. Hey! more genes! great!
This time my experiments were comparisons between treated/untreated in 
the same cell line. The treatments were various siRNA knockdowns and 
overexpression of transgenes. The highest B values were around 4...
I assumed the much lower values were a reflection of the variation 
between experiments (every slide was hybridised with a separate 
treatment)... By Western we could see how our transgene was being 
expressed at varying levels in different experiments, so it sort of 
made sense... although the low B values still surprised me a bit.

3) I decided to repeat one of the experiments from 1)... this is a 
comparison between two straight cell lines, no transfections or any 
other treatment. Now, while using the 10k array I obtain a max value of 
B=14.43, when I tried it on my new 22k arrays, the largest B value was 
a mere 3.69.

All hybs were performed in the same way, with the same buffer. RNA 
extracted and labelled the same way... same scanner...

All slides look clean, before and after hyb.

Has anybody experienced something like this?
I'm beginning to think that the new arrays are not that good... Does 
anybody know what would be a good way to check that a set of arrays are 
good enough?

I have access to a scanner that can detect DAPI. I stained two old 
slides and scanned them. The 10k one gave pretty uniform spots. The new 
22k one varied wildly (I can email pictures upon request) from spot to 
spot... but I haven't yet checked whether there's also big differences 
*between* individual slides. I imagine if the spotted DNA varies in 
amount a lot, where sometimes you have very little, the ratios could be 
more variable than we wish, and perhaps that brings the B values down? 
I really don't know.

I have a whole bunch of these arrays and they contain many genes we are 
interested in... but they'll be pretty useless if I cannot trust the 
results and makes their interpretation difficult.

I would appreciate any help, any pointers as to what I should check to 
ascertain the quality of any slides we may purchase. Or any comments 
pointing out to other possible factors I have not considered.

Thanks so much!

Jose

-- 
Dr. Jose I. de las Heras                      Email: J.delasHeras at ed.ac.uk
The Wellcome Trust Centre for Cell Biology    Phone: +44 (0)131 6513374
Institute for Cell & Molecular Biology        Fax:   +44 (0)131 6507360
Swann Building, Mayfield Road
University of Edinburgh
Edinburgh EH9 3JR
UK