[BioC] Affy: Present calls in an eset

[Julian Selley]

I too have been looking at this issue and would be interested in any answer
people might have.

Firstly, my understanding is that it is a detection call ("Is the transcript
of a particular gene Present or Absent?" -- Affymetrix "Statistical
Algorithms Description Document". This is with reference to the background
level. The detection call is also referred to under a number of other
pretexts - "discrimination score and probability" for one.

I have been going through the MAS 5 documentation and various other stuff,
and all I can find to reference how this works, is at the "spot"
level(ncol*nrow position - forgive my naivety but I too started working in
this field just a month ago and am still coming to terms with
understanding). As to yet I have found nothing as to how to program it for
an expression set, yet quite obviously this is possible. Stats was never a
strong point of mine either.

If anyone wants a look, I have written a function to work at the "spot"
level. It's not at the stage of error checking or anything - it was a quick
hack.

Thanks,

Julian.



-----Original Message-----
From: bioconductor-bounces at stat.math.ethz.ch
[mailto:bioconductor-bounces at stat.math.ethz.ch] On Behalf Of A.J. Rossini
Sent: 08 October 2003 15:06
To: Arne.Muller at aventis.com
Cc: bioconductor at stat.math.ethz.ch
Subject: Re: [BioC] Affy: Present calls in an eset

<Arne.Muller at aventis.com> writes:

> I get your point with interpreting absent/present calls. Technically it's
a
> nice feature, becasue one can just discard the majority of the genes on
the
> chip for further analysis. In fact I think absent/present calls make sense
in
> terms of biology, since just a fraction of the genes are realy expressed
at a
> time. How to express this numerially is a different story (and I guess a
> difficult one).

Sure, but equivalently, I could implement the "feature" with "Tony's
Complete Crud Calls" (TCCC):

            rbinom(40000,1,0.2)

Change the 20% expressed to a different number if it makes you feel
better. 

I'm not being quite fair, but from my perspective, it's only slightly
less realistic than the MAS or Resolver calls -- the difference being
that the probability of response is an unclear function of the
expression level, with a trend where the probabilty of the call goes
up with the expression level (which is the technical component).

And then, you've got the possibility of low levels of expression which
are real but hard to pick out.

> Anyway, with MAS the calls are calculated anyway, can't they? So, I'd be
nice
> (at least for "completness") to add a "mascall" method to the exprSet
objects
> generated by affy. What do you think?

It would be great.  As far as I know, the MAS calls are protected as a
trade secret and non-disclosure agreements.  Maybe not, but I happen
to know for sure that the rosetta resolver calls are! (or I'd have
implemented them for amusement)

> By the way, if you ignore the call, do you set an arbitrary intensity
cutoff
> later in your analysis, or do just reley on the statistics (anova  p-value
or
> whatever)?

If you have biological replicates, then either, depending on what you
are looking for.  I've never found an approach which works all the
time.  If you have technical replicates, then the calls are necessary,
but note that they don't have biological meaning, just technical
(technology) meaning.

best,
-tony

-- 
rossini at u.washington.edu            http://www.analytics.washington.edu/ 
Biomedical and Health Informatics   University of Washington
Biostatistics, SCHARP/HVTN          Fred Hutchinson Cancer Research Center
UW (Tu/Th/F): 206-616-7630 FAX=206-543-3461 | Voicemail is unreliable
FHCRC  (M/W): 206-667-7025 FAX=206-667-4812 | use Email

CONFIDENTIALITY NOTICE: This e-mail message and any attachme...{{dropped}}

_______________________________________________
Bioconductor mailing list
Bioconductor at stat.math.ethz.ch
https://www.stat.math.ethz.ch/mailman/listinfo/bioconductor