Information Technology Reference
In-Depth Information
The data challenge was very useful to identify the limitations and
bottlenecks of the EGEE infrastructure. The WISDOM production system
developed to submit the jobs on the grid accounted for a small fraction of
the failures as well as the grid management system. On the other hand, the
resource brokers were observed to signii cantly limit the rate at which the
jobs could be submitted. Another signii cant source of inefi ciency came
from the difi culty for the grid information system to provide all the rel-
evant information to the resource brokers when they distributed the jobs
on the grid. As a consequence, job scheduling was a time-consuming task
for the WISDOM users during all the data challenge due to the encoun-
tered limitations of the information system, the computing elements, and
the resource brokers.
14.3.2.3
Results
Postprocessing of the huge amount of data generated was a very demand-
ing task as millions of docking scores had to be compared. At the end of
the large-scale docking deployment, the best 1000 compounds based on
scoring were selected thanks to postprocessing ranking jobs deployed on
the grid. They were inspected individually. Several strategies were
employed to reduce the number of false positives. A further 100 com-
pounds were selected for postprocessing. These compounds had been
selected based on the docking score, the binding mode of the compound
inside the binding pocket, and the interactions of the compounds to key
residues of the protein.
Several scaffolds were identii ed in the 100 compounds selected for
postprocessing. The urea, thiourea, and guanidino scaffolds were the
most frequently observed in the top 1000 compounds. Some of the com-
pounds identii ed were similar to already known plasmepsin inhibitors,
like the urea analogs that were already established as micromolar inhibi-
tors for plasmepsins (Walter Reed compounds) [12]. These results were
already an indication that the overall approach was sensible and that
large-scale docking on computational grids had the potential to identify
new inhibitors, such as the guanidino analogs. To coni rm these results, a
new step had to be implemented: the rei nement of the compound selec-
tion using molecular dynamics computations.
14.3.3
Molecular Dynamics on the Grid
14.3.3.1
Introduction
While docking methods have been signii cantly improved in the last few
years by including more thorough compound orientation searches, addi-
tional energy contributions, and/or rei ned parameters in the force i eld, it
is generally agreed that docking results need to be postprocessed with
Search WWH ::
Custom Search