Software used:
BPEL : Provides the workflow design tool and enactment engine
GridSAM : Provides the web service job execution interface to run the simulations

The UCL Theoretical Chemistry group led by Professor Sarah L. Price has a long track record in polymorph simulation. This interest is driven by the need for understanding the different possible crystal structures that a complex molecule may adopt and has important applications in drug and formulation design, understanding the behaviour of pigments in paint and the engineering of complex materials. In the EPSRC funded e-Materials project, the group ported their approach for polymorph simulation to a grid setting.

Early on in that effort it became clear that the workflows during that search needed to be flexibly amended and the group experimented with BPEL for that purpose. Prof. Price's team now uses the OMII-BPEL environment¹, GridSAM and the entire OMII stack to control the parallel execution of up to 8000 jobs on the UCL Condor pool during any one search. The figure below shows the BPEL process that controls job-submissions to the UCL Condor by invoking web services provided by GridSAM.

This process is one of the leaves in a complex hierarchy of processes. Such hierarchical composition is enabled, as each BPEL process is also a web service. In order to achieve significant speed-up of the simulation on a grid infrastructure hundreds and thousands of these submission processes can be handled by the OMII-BPEL run-time concurrently. In grid settings nodes also occasionally fail or become unreachable and BPEL compensation handling primitives support recovery from these failures, typically through re-submission to other nodes. The environment persists process state in PostgresQL, the OMII database in order to survive accidental or scheduled termination of the BPEL run-time, a crucial feature to support long-running workflows. The OMII-BPEL environment is fully integrated with the OMII stack. As a result the execution of BPEL processes is secured through WS-Security primitives, a crucial feature required to protect the IP of drug discovery processes in a grid setting.

Through this grid-enabled approach, Prof. Price's group has brought down the time for obtaining simulation results from up to three months (on a dual CPU Silicon graphics machine) to mere 4 hours on the 1,200+ node UCL Condor pool and at the same time increased the precision of the searches. This dramatic improvement has changed the way chemists perform simulations and enabled the simulation of more complex molecules. Prof. Price's group is now able to perform exploratory and low precision searches interactively. The provision of this environment has enabled a number of scientific breakthroughs. Using the approach Prof. Price's group was able to predict a new polymorph of the drug Aspirin², which was later experimentally confirmed. The BPEL environment was used to successfully confirm the structure of a new polymorph of a peptide-coupling agent³. Most publicity, including an article in Nature News⁴ has been generated by the successful prediction of the structure of the conformational polymorph IV of the nootropic agent Piracetam, which was recently discovered by re-crystallisation under pressure⁵.

¹Emmerich, W.; Butchart, B.; Chen, L.; Wassermann, B.; Price, S.; J Grid Comp 2005, 3(3-4), 283-304
²Ouvrard, C.; Price, S. L. Cryst. Growth Des. 2004, 4, 1119.
³Nowell, H.; Frampton, C. S.; Waite, J.; Price, S. L. Acta Crystallogr. Sect. B-Struct. Sci. 2006. In press.
⁴http://www.nature.com/news/2005/050919/full/050919-7.html
⁵Nowell, H.; Price, S. L. Acta Crystallogr. Sect. B-Struct. Sci. 2005, 61, 558.