[BioC] ad hoc tools for affy tiling arrays?

[Sean Davis]

On 2/1/06 7:11 AM, "Dario Greco" <dario.greco at helsinki.fi> wrote:

> hello,
> thanks a lot for your messages.
> actually, in case we will approach to the tiling arrays, we would like to
> screen the transcriptome, and not the genome.

On tiling arrays, there is no distinction between transcriptome and
genome--you are measuring every probe along the genome.  Some will have
signal and may represent transcription, while some will not.  You will have
to map each probe to genomic features (or lack thereof) before you know what
you are measuring.

> we would actually like to find non coding (novel ?!?!) RNAs having a
> different expresison pattern in several samples...or at least this would
> be the ultimate dream!

You will likely find many novel transcripts, based on our experience and
from that of others in the literature.  Determining differential expression
is potentially a very difficult problem.  You are measuring at discrete
probes, so you have to first determine what constitutes transcription, then
whether it is represents a novel transcript, and finally if you are seeing
differential expression.  Each of these is an area of research
in-and-of-itself.  

> so, in this situation, what would be the best analytical approach, in your
> opinion?

This cannot be summarized in any meaningful way in an email.  There are just
too many details to consider:  What controls are present?  What samples are
being used?  Is this one array or multiple to cover the genome?  How dense
are the probes?  Do you need to compare across arrays for the same sample or
not?  Are the samples of similar quality?  What genome annotation do you
want to use?  How do you want to "lump" probes into transcriptional units?
What computational resources do you have?

The first step is to get the Integrated Genome Browser from Affy and
visualize your data.  After that you can begin to decide what you will
need/want to do.  I would definitely recommend running some test slides to
make sure that your experimental design is reasonable before you spend the
presumably large amount of money on a full set of slides.  Plan on several
weeks to months to analyze the data.  Read carefully the few papers
available on tiling array analysis of the transcriptome to get ideas
regarding the complexity.

I'm certainly not an expert in this area, so it will be good to get others'
opinions.  However, I have done enough with tiling arrays to appreciate
their complexity and am just trying to pass some of this along.

Sean