[BioC] Normalization with few spots

[Teresa CAsals]

Hello

I have received some cDNA arrays to analyze with very
few genes on them.

There are only 32 different genes, each of which has
been spotted on each array 8 times. There are also
some (not many more controls and spikein spots).

It is a reference design with unbalanced dye-swap
based on biological replicates intended to compare
three mutant types to a wild type. I didn't suggest
it, just received the data after the experiment was
performed (so I may be able to say what it died of :-)
The design is as follows 

Array	Cy3	Cy5	
1	Mut-1	Wild
2	Mut-1	Wild
3	Wild	Mut-1
4	Mut-2	Wild
5	Mut-2	Wild
6	Wild	Mut-2
7	Mut-3	Wild
8	Mut-3	Wild
9	Wild	Mut-3

My questions are:

1-How should I normalize the data? I ususally use
marrayNorm with print-tip lowess, but I think this may
not be adequate having so few spots.

Another question refers to dye-swap normalization. I
have read in some bioconductor courses slides that a
self normalization may be adequate for dye-swap
experiments. 
In this case a normalized estimate of the log ratio is
obtained M values (1/2 (M-M')).
My questions are

2- Doesn't it imply some information loss? I mean is
it truth that for for every two arrays I only get an
estimate? I may be missing something but I don't know
what...

3-How should I manage the assimetry in dye swap? It
seems unreasonable having first to average slides 1
and 2 and the combine it with three...

Any help or reference will be great

Thanks

Teresa Casals





		
______________________________________________
Renovamos el Correo Yahoo!: ¡100 MB GRATIS!
Nuevos servicios, más seguridad