[BioC] GCRMA explained in words

[Zhijin (Jean) Wu]

TO put it simple:

RMA and GCRMA share the same idea of assuming that the observed intensity
of a probe is a sum of specific (S) and non-specific (B,background)
components, and both compute posterior mean of specific component
given the observed sum.

The main differences are:
-RMA assumes that background follows normal distribution.
It also assumes that background for all probes are from the same normal
distribution. Therefore RMA uses all probes to estimate the one set of
parameters of that normal distribution.

-GCRMA assumes log normal distribution for background. It also assumes
that the parameters of that log normal distribution depends on the
probe sequence (a little more complicated than just GC content). Therefore
it uses "similar" probes to estimate those parameters. Which probes are similar is
determined by their sequences.

-GCRMA computes posterior expectation of log(S) but RMA
computes posterior expectation of "S"

Hope this helps,
Jean Wu

On Tue, 7 Feb 2006, Richard Friedman wrote:

> Dear Bioconductor list,
>
> 	I am trying to get a understanding
> and verbal description of GCRMA.  I know that
> it involves fitting for non-specific binding in
> a GC-content dependent way, but it is not clear how.
> If I were to explain RMA in words (thanks to Benilton and other
> list members)  I would say that
> the probe frequency vs intensity plot is fit with
> a Guassian Model for noise and an exponential model
> for signal. Is it correct to say that in GCRMA, the fit is
> the same as in RMA but with set of probes of a given GC
> content fit separately? If not, how would one put it with
> comparable simplicity? I've read the papers several times
> and some of the presentations on Dr. Irizarry's web-site,
> and I am still not sure that I understand how the fit
> is done.
>
> Thanks and best wishes,
> Rich
> ------------------------------------------------------------
> Richard A. Friedman, PhD
> Associate Research Scientist
> Herbert Irving Comprehensive Cancer Center
> Oncoinformatics Core
> Lecturer
> Department of Biomedical Informatics
> Box 95, Room 130BB or P&S 1-420C
> Columbia University Medical Center
> 630 W. 168th St.
> New York, NY 10032
> (212)305-6901 (5-6901) (voice)
> friedman at cancercenter.columbia.edu
> http://cancercenter.columbia.edu/~friedman/
>
> "42 is the answer. Dylan got it wrong. 'Blowin'
> in the wind' is not the answer. It isn't even
> a number' " - Rose Friedman, age 9
>
> _______________________________________________
> Bioconductor mailing list
> Bioconductor at stat.math.ethz.ch
> https://stat.ethz.ch/mailman/listinfo/bioconductor
>