Pittsburgh Supercomputing Center
News Release



NEWS RELEASE
December 11, 2001




Pittsburgh Supercomputing Center and Compaq Team Up to Support CASP5

By making large-scale computational resources available, PSC and Compaq expect to advance the science of structural biology.

PITTSBURGH — The Pittsburgh Supercomputing Center (PSC) and Compaq Computer Corporation are collaborating to provide large-scale computational resources in support of CASP5, an important process by which the international community of molecular and structural biologists assesses its ability to accurately predict the three-dimensional structure of proteins. On Dec. 7, the National Science Foundation Partnerships for Advanced Computational Infrastructure program allocated computing time on PSC's Terascale Computing System (TCS) to support this PSC-Compaq collaboration.

Deriving protein structures from genomic sequence data is the major challenge of contemporary computational biology. Genes are the blueprints for proteins, and the next step in harnessing the flood of data from genome research is using it to deduce the 3-D structure of proteins. For practical purposes, these problems are insoluble without advanced software and powerful computers. Among research groups worldwide, a multitude of methodologies have arisen, many based in comparing unsolved sequences with databases of known structures, others aimed at deducing a protein's structure solely from its sequence data — the so-called "protein-folding problem."

CASP (Critical Assessment of Techniques for Protein Structure Prediction) is a community-wide experiment held every two years that assesses these methodologies. "CASP has become vitally important to the structure-prediction community," says David Deerfield, who directs the biomedical research program at PSC.

Through CASP, researchers participate in blind-test experiments in which they determine the 3-D structure of sequences for which the structures are already known but haven't yet been made publicly available. Independent assessors then judge the quality of the predictions. Participation has steadily grown, from 35 research groups at the first CASP, 10 years ago, to 158 groups two years ago at CASP4. "Through CASP," says Deerfield, "the field moves toward consensus on such questions as which methods are most accurate, which most cost-effective, where can future effort most effectively be directed?"

One of the major barriers to structure-prediction progress, says University of Maryland computational biologist John Moult, president and founder of CASP, has been lack of access to large-scale computational resources. "Many of these methods require very large amounts of computational capability in order to achieve good results," says Moult, "and in the past some groups have been at a disadvantage because they lack access to top-flight computational resources. We desperately need innovative new methods to remove the present bottlenecks to progress. And many of those methods will require major computing resources. CASP should not be limited by computational capability."

To that end, PSC and Compaq have teamed up to provide computing time on the TCS and on a Compaq 100-processor cluster of AlphaServer systems, which is compatible with the TCS. Comprising 3,000 Compaq AlphaServer EV68 processors, the TCS has a peak capability of six teraflops (six trillion operations per second) and is the most powerful system in the world committed to unclassified research.

As part of this collaboration, PSC and Compaq will make available a computational grid environment to facilitate execution of CASP calculations. The grid environment will provide a common means of access to computational resources at PSC and Compaq.

Both Compaq and PSC will provide consulting and engineering support to help CASP5 participants run their structure-prediction software on the TCS AlphaServer architecture.

Along with computational cycles for CASP5 participants on its 100-processor AlphaServer cluster, Compaq will also help provide computational and data storage capabilities to collect, rapidly assess and publish the results of CASP5. Compaq will also serve as a sponsor and technology provider for the CASP5 meeting to be held in Asilomar, California in December 2002.

"Compaq is delighted to significantly expand its efforts with PSC in providing scientific researchers with the high-performance computing environment they need to achieve the next level of computational breakthroughs required to accurately predict the structure of proteins," said Bill Blake, vice president of High Performance Technical Computing at Compaq. "We see this collaboration as a way to support and encourage researchers in developing or improving their algorithms for 3-D structure prediction, and, in the longer term, accelerate the research community's understanding of structure-function relationships for improved drug discovery."

Small grants will be awarded for initial development, testing of algorithms, and optimizing on the Compaq AlphaServer system architecture. A panel of scientists — Deerfield of PSC, Michael Levitt, chair of computational structural biology at Stanford University School of Medicine, and Jeffrey Skolnick, director of computational science and structural biology at the Danforth Plant Science Center in St. Louis — will review requests for larger follow-up allocations supporting actual participation in the CASP5 experiment, which begins in April 2002.

Information on CASP and the application process for computing time is at:
http://predictioncenter.llnl.gov/

The Pittsburgh Supercomputing Center is a joint effort of Carnegie Mellon University and the University of Pittsburgh together with the Westinghouse Electric Company. It was established in 1986 and is supported by several federal agencies, the Commonwealth of Pennsylvania and private industry.

PSC Public Information Contact::

Michael Schneider
Pittsburgh Supercomputing Center
412-268-4960
schneider@psc.edu


© Pittsburgh Supercomputing Center (PSC)
Revised: December 12, 2001
URL: http://www.psc.edu/publicinfo/news/2001/casp-12-11-01.html