WISDOM: A Grid-Enabled Drug Discovery Initiative against Malaria - Grid Computing: Infrastructure, Service, and Applications - page 367

available in the Protein Data Bank. These are the structures of wild-type P.
falciparum DHFR (PDB code 1J3I) and of its C59R + S108N (1J3J) and
N51I + C59R + S108N + I164L (1J3K) highly resistant mutants. In addi-
tion, four crystal structures of P. vivax DHFR are available, the wild-type
structure (2BL9, 2BLB) and the S58R + S117N resistant mutant structures
(2BLA, 2BLC). Hence, we thought it was a great opportunity to make use
of these available resources to develop technology-based in silico screen-
ing on these targets. We chose P. falciparum DHFR (wt and the highly resis-
tant quadruple mutant) and P. vivax DHFR (wt and the double mutant)
structures ( Table 14.2 ).

14.4.2 Evolution of the Production Environment

Following the experience acquired during the i rst data challenge on
malaria and the i rst data challenge on avian l u [21], the WISDOM envi-
ronment was reorganized in two different and independent tasks:

• The submission of the jobs

• The follow-up of the jobs, and eventually their resubmission, as

well as the collection of the job status and their publication on a
Web site

These two processes can be started and run simultaneously, the second
one being fed from the information provided by the i rst one.
As seen during the deployment against avian l u [21], we removed the
automatic resubmission of jobs in case of a job failure, because we thought
that the overload generated by this automatic resubmission was a major
cause for the poor reliability we observed during the i rst WISDOM
deployment. Removing the automatic resubmission of jobs helped indeed
a lot to improve overall reliability, but this induced a large amount of work
for the users, who had to handle manual resubmissions of the instances
after all the jobs had i nished. The environment was further modii ed and
enhanced to become more a “launch and forget” system, to carry all the
tedious tasks and relieve the users whenever possible, such as automatic
resubmission of jobs, automatic storage of job results in a relational data-
base, automatic and real-time update of the experiment statistics and
status, viewable through a Web site, and so on. As a consequence, the
WISDOM environment was easily customized, tested, and successfully
used by the EELA [22] and EUChinaGRID [23] virtual organization.
The main objective was also to improve the fault tolerance of the system,
in implementing, for instance, a persistent environment that can be stopped
and restarted at any time without the risk of losing important information.
This also proved to be very useful as it enabled the whole maintenance of
the scripts and code and improved interactivity with the user, as the user
could also manage jobs in i ne detail; for instance, force the cancellation