Speaker
Description
3. Impact
We have used the production grid set up by the EGEE-II project. We have submitted 35,840 energy minimizations as individual jobs on the grid. This means that each job had gone through the submission processes, and thus paid the overhead inherent to the grid architecture and internal processes: from the submission through the user interface (UI), via the scheduling step on the resource broker (RB) to the execution on the computing element (CE), a cluster with several worker nodes (WN). The whole computing task was launched through 12 RBs, which have scheduled all the jobs on 23 CEs. The total cumulated computing time was about 1,275 days, with a job duration of 51 minutes on average. The full calculation was completed after 4 days and 16 hours, running up to 1039 jobs simultaneously. This was 271 times faster than using a single machine.
1. Short overview
How proteins find their targets amongst millions (or more) of competing sites is still largely an unsolved problem. Understanding this process in detail is however central to understanding the mechanisms underlying gene expression. A better understanding of site-specific targeting is also a vital step towards rational re-engineering of proteins for therapeutic purposes.The problem becomes even harder when a complex of several proteins binds to DNA, as in the case of the nucleosome core particle.
4. Conclusions / Future plans
Using the EGEE grid to obtain a first indication of the binding specificity of the nucleosome turned out to be rather efficient. The results have demonstrated the sustainable status of the EGEE grid for large-scale experiments with a real laboratory workflow. We are planning to continue our study with an improved model that will require 140,000 energy minimizations, corresponding to roughly 16 years of sequential CPU time.
Provide a set of generic keywords that define your contribution (e.g. Data Management, Workflows, High Energy Physics)
Bioinformatics, Molecular simulation, Large scale experiment
URL for further information:
http://gbio-pbil.ibcp.fr