[BioC] map probe to refseq and select 3' biased one

[Francois Pepin]

Hi Weiwei,

You're talking about probesets not probes, correct?

Otherwise, you would end up throwing away most of the probes on every
probeset and that would be a fairly bad idea.

You could look at a handful of probes for cases when several probesets
exist for a single gene and see if there are any systematic 3' biases,
but I would expect otherwise. As an alternative you could filter using
variance or interquartile range. It's maybe not as elegant as choosing
the 3' one, but it should work fairly well.

Francois

On Sun, 2007-06-24 at 18:08 -0400, Weiwei Shi wrote:
> Intially I thot I could tell from e.g. hgu133aREFSEQ. However, I did not find.
> 
> The reason I do this, is following MAQC I paper when they align
> different probes from different platforms. Currently I am focusing on
> affy products (still not the same platforms). I was notified that they
> selected 3'bias in this many2one situation.
> 
> I believe there exists some RefSeq database which has that field to
> tell you which probe (for the same gene) is the 3' end (though not
> sure of it). Since today is Sunday I don't want to bother people, so I
> try to post here to see if there is a solution or not.
> 
> -w
> 
> On 6/24/07, Sean Davis <sdavis2 at mail.nih.gov> wrote:
> > Weiwei Shi wrote:
> > > Hi,
> > >
> > > Does anyone know how to map affy probe to refseq and select 3'biased
> > > one, so probe to gene can be one to one?
> > >
> > Are you talking about an affy probeset, or just the probes?  Even if you
> > map to refseq, there are multiple refseqs per gene in many cases, and
> > some of them have the same 3' exon.  Unfortunately, one simply has to
> > deal with this many-to-many situation.
> >
> > To answer your question, though, in general mapping probes or probesets
> > to other sequences will involve using blast, blat, or some other
> > alignment tool.  Going down this road, particularly with affymetrix
> > arrays, is a fairly large undertaking; I would not advise it unless you
> > have a particularly compelling reason to do so.
> >
> > Sean
> >
> 
>