[BioC] why S-Score is not poppular like T-test pvalue and BH FDR , which is OK to use?
rkennedy at vcu.edu
Wed Nov 30 20:03:04 CET 2005
The S-Score is an algorithm for analyzing microarray data, originally
developed in Michael Miles' laboratory at UCSF (by Li Zhang, currently
at MD Anderson Cancer Center). Dr. Miles has since relocated to our
institution, and we are working as a group on the S-Score.
As Adaikalavan notes, the S-Score has not been heavily publicized, which
may account in part for the lack of widespread use. However, since the
release of Dr. Miles' Journal of Neuroscience paper (Kerns et al, J
Neurosci 2005, 25(9):2255-2266), we have been seeing increasing
interest. The S-Score does have some rather interesting features, with
the main one being that it performs tests of hypotheses directly from
the probe level data. The usual steps in analysis require that a probe
set summary measure be generated, and statistical tests are performed on
the summaries. Under conditions of no differential expression, the
S-Scores follow a standard normal distribution; this means that
hypothesis testing can be done directly with the S-Scores, without an
intermediate expression summary measure.
There are, of course, some limitations with the S-Score as well. The
software had previously been available only as a compiled stand-alone
program for Windows (at http:www.brainchip.vcu.edu/expressionda.html) so
it took a few steps to combine it with an environment such as R.
However, we have been working on an R implementation which, as Seth
noted, should be posted in the near future. This initial release is
just a straight port of the Windows version to R; it does not add any
new features, but does make it easier to interface with the other
Bioconductor packages. Another principal limitation of the S-Score is
that the software performs comparisons two chips at a time. We are now
working on a multivariate version for direct multi-chip comparisons, but
have not set a definite timeframe for its release.
The main reference for the S-Score is Zhang et al, J Mol Biol 2002,
317:225-235 that describes some of the original development; Dr. Miles
has also used the S-Score in several peer-reviewed publications. We are
in the process of submitting an article on additional validation studies
of the S-Score (which did rather well compared to other methods) and an
applications note for the R implementation.
Hope this has been helpful - we are definitely interested in making the
S-Score more widely available.
Kellie J. Archer
Department of Biostatistics
Virginia Commonwealth University
Saurin Jani wrote:
>Have any one used S-Score algorithm in order to
>analyze microarray data and extracted diff. expressed
>genes ? Or is it useful to use use...RMA -> do t-test
>get p-valule -> and do FDR (BH) --> extract diff.
>why S-score is not in much of use? like T-test pvalue
>and BH FDR..which is OK to use?
>Thank you so much in advance,
>Bioconductor mailing list
>Bioconductor at stat.math.ethz.ch
More information about the Bioconductor