Affycomp III: A Benchmark for Affymetrix GeneChip Expression Measures
In April 2011,
Philippe Serhal tried to reproduce the
Affycomp II assessment of RMA, an effort we realized was
doomed to failure.
That assessment is important because (along with
MAS 5.0) it constitutes a conceptual base-point of this benchmark tool.
However, despite the hopeful subtitle, our assessments, produced on entry
submission and never updated, had become incommensurable.
All but the most recent had become slightly inconsistent with today's
affycomp package, hence incomparable, or at least not easily comparable.
- While Affycomp II dates from 2003, the modern affycomp includes
non-trivial though minor changes made in 2004 to assessLS and
assessMA2, refining the spec for low/med/high-
The package example assessments (mas5.assessment,
rma.assessment, and variants -- see
affycompData) were however still old, from 2003.
- The changes of 2004 (should) alter not only assessment, including the
example assessments, but in turn certain comparison of expression measures.
- Only the most recent (i.e. for entries submitted in 2009 or later -- there
were no successful entries for competition in 2008) of the
AFFY2 assessments, used the modern package.
- Earlier assessments (i.e. for entries submitted in 2007 or before -- in effect,
for the entries submitted in 2003 - 2004), were produced, on entry submission, by a
moving affycomp package that was possibly experimental and defintely not tracked.
In response, Affycomp III ...
- The affycomp package example assessments have been re-built (affycompData, 1.1.0).
- The site's affycomp software is updated (1.28.3).
- There are
new assessments and tables (for example,
MAS 5.0), affecting:
- mainly the $LS and $MA2 slots of assessments
(see differences > .01)
- columns 6-8 and 11-14 of the tables (thus, rankings
based on any of these scores)
- Figure 4c and Figures
new comparison tool uses the new assessments.