APPLICATIONS

To understand how biomolecules like DNA move across protein pores - a problem that for decades proved too complicated to simulate - a team of renowned British scientists is turning to a wide array of technologies, including visualization and computational solutions from Silicon Graphics. Their resulting project, designed to create a Simulated Pore Interactive Computing Environment (SPICE), was honored last week at SC|05, an international supercomputing conference where the researchers tied for first place in the HPC Analytics Challenge. Led by scientists from University College London (UCL), the global SPICE project team found a way to compute the free energy profile of a translocating biomolecule along the vertical axis of a pore. Simulating translocation of DNA, RNA and polypeptides may someday help medical researchers control that activity, potentially leading to new treatments for numerous medical conditions and illnesses. The SPICE team's approach is so advanced that traditional means had rendered the problem unsolvable. The ongoing project requires fully atomistic simulations, which illustrate molecular and chemical interactions within the smallest unit of matter. "The time scale for DNA translocation molecules is on the order of tens of microseconds," said Professor Peter Coveney, one of the SPICE project's lead investigators and director of the Centre for Computational Science at UCL. "Simulating timescales like these for models of 275,000 atoms or larger just isn't possible with standard approaches to molecular dynamics. Even assuming that computer power will continue to accelerate along the curve described by Moore's Law, we're still a couple of decades away from the day when this type of simulation will be routine." To achieve the project's aims, the team is using two HPC grids - the TeraGrid in the USA and the UK's National Grid Service - along with SGI visualization resources located at UCL. Key to visualizing the simulation is a Silicon Graphics Prism system with six Intel Itanium 2 processors and four ATI FireGL graphics pipes connected via OpenGL Vizserver, which allows remote team members to collaborate by interactively viewing the simulation. SPICE team members also leveraged two Silicon Graphics Prism systems located in the University of Manchester booth at SC|05. Other SGI resources, including a multi-node SGI Altix installation, are linked via the UKLight optical network so SPICE team members have high-speed access to remote grid and visualization resources. "Without the use of a sophisticated grid infrastructure and high-performance visualization resources, we would not be able to pursue this research," said Coveney. "Throughout all our work, SGI's hardware and software has played a key role in making our efforts so successful." The SPICE team is led by UCL's Coveney, who collaborates with Shantenu Jha, and Matt Harvey. The trio in turn is collaborating with Stephen Pickles of UCL, Robin Pinning of the University of Manchester, Peter Clarke of the University of Edinburgh, Bruce Boghosian of Tufts University, Charlie Catlett of TeraGrid, Charles Laughton of Nottingham University, Rob Pennington of National Center for Supercomputing Applications (NCSA) TeraGrid, Sergiu Sanielevici of the Pittsburgh Supercomputing Center, Jennifer Schopf of Argonne National Lab, and Richard Blake CCLRC Daresbury. "The work of these researchers reveals a unique and inspiring kind of ingenuity - one that uses collaboration to push beyond the limitations of current technology to do something that has been widely viewed as impossible," said Tim Butchart, UK country manager and EMEA marketing director, SGI. "We're delighted to see SGI visualization and computation solutions play a pivotal role in the SPICE team's success, and we congratulate them on winning the HPC Analytics Challenge at SC|05." As it does annually, SC|05 last week presented awards that recognize the innovative work of conference participants and HPC leaders. The HPC Analytics Challenge is a new award competition honoring select technical and commercial applications that use leading-edge, advanced analytics techniques to solve complex, real-world problems. The growing need to leverage large amounts of underutilized data has led to the use of sophisticated methods for analysis and high-end visualization in conjunction with high performance computing, bandwidth and networking capabilities. The two teams that won awards were chosen from among 21 participants in the competition. Tying with the SPICE team was a team of researchers whose project is titled, "Real Time Change Detection and Alerts from Highway Traffic Data." Other SC|05 Awards Several other teams earned accolades at SC|05 for work completed on Columbia, NASA's 10,240-processor SGI Altix supercomputer. For instance, one of the other five finalists in the HPC Analytics Challenge leveraged Columbia's formidable computational power to create a new, more effective visualization tool for interrogating large, unsteady data sets. In another category that featured two teams tied for first place, the Best Technical Paper Award went to "High Resolution Aerospace Applications using the NASA Columbia Supercomputer," authored by Dimitri J. Mavriplis of the University of Wyoming, Michael J. Aftosmis of NASA Ames Research Center, Marsha Berger of the Courant Institute. In addition, the SGI Altix family took top honors in the Tour de HPCycles panel session. In the SC|05 competition fashioned after the cycling world's Tour de France, a panel of HPC end users voted to award jerseys for a range of categories. SGI Altix earned the coveted Yellow Jersey, the award for best overall supercomputer.