Biomedical Engineering Reference
In-Depth Information
refi ne the Protein Data Bank (PDB) [10]. The goal of this application was to
recalculate 19,000 X-ray structures in the PDB. Indeed, structural biology,
homology modeling, and rational drug design require accurate 3D macromo-
lecular coordinates. However, the coordinates in the PDB have not all been
obtained using the latest computational methods. The study showed that they
can be improved in terms of fi t to the deposited experimental X-ray data as
well as in terms of geometric quality.
The re - refi nements of the structure models were performed on a hybrid
computing environment consisting of two virtual organizations of the EGEE
grid infrastructure and several clusters provided by bioinformatics institutes in
Europe within the framework of the EMBRACE project [11]. On a single CPU,
the entire calculations would have taken about 17 years. With our grid-and-
cluster computing approach, more than 90% of the total calculation was fi n-
ished in only two months—this shows the clear time advantage arising from the
usage of modern computing technology. All 16,807 successful re-refi nements
were complete after four months; the vast majority were done after three weeks.
By employing methods such as translation/libration/screw (TLS) motion
refi nement that represents the displacement of groups of atoms that behave
as (quasi) rigid bodies, 10,046 out of 15,034 structure models (67%) were
improved [9]. These results showed that re-refi nement of existing PDB entries
was worthwhile and, because the method is fully automated, little time invest-
ment was needed to re-refi ne a single structure model. PDB entries are now
routinely re - refi ned before they are used for molecular dynamics, homology
modeling, or drug design.
15.4 GRIDS TO SHARE DATA WHERE IT IS PRODUCED
15.4.1 Introduction
Everybody understands the importance of sharing data in modern-day science,
but one may wonder why it is important to share data where it is produced.
The important concept is to allow a data owner to keep control of who accesses
his or her own data. Indeed, in order to publish in peer-reviewed journals, a
scientist must demonstrate creativity and present ideas and results that are
beyond the present state of the art. One way is to work with experimental data
that have not been previously produced or analyzed. Making these data pub-
licly available represents a big risk of losing a competitive advantage and
therefore many researchers are reluctant to share their data out of fear they
are overtaken by competitors and exploited without due credit or attribution
in general. Not all scientifi c communities show this behavior. For instance, this
fear does not exist with high-energy physics experimental data because they
are unusable without an in-depth knowledge of the detector used to produce
them [1]. In the fi eld of molecular biology or astronomy, data are made avail-
able after one or two years so that the scientists have enough time for publish-
ing a few papers before releasing them for a wide use by the community.
Search WWH ::
Custom Search