With $2 Million Grant, UB Team to Launch a "Transparent" Virtual Community for Protein-Structure Scientists

BUFFALO, N.Y. -- The concept of a virtual community, where like-minded individuals share their knowledge for the benefit of all who participate, might appear out-of-step with the culture of research scientists, who traditionally have guarded jealously their data from rival groups until publication.

But University at Buffalo scientists working in the field of protein-structure determination have been awarded a highly competitive, $2 million National Science Foundation Information Technology Research grant to do just that and more over the next four years.

The project is designed to use new technologies, such as grid computing, data mining and collaborative environments, to enhance protein-structure determination, a key tool in the rational drug-design process, where the discovery of new pharmaceuticals is based on precise knowledge of protein structures.

The Buffalo team is affiliated with the Strategically Targeted Academic Research (STAR) Center in Disease Modeling and Therapy Discovery at UB, sponsored by the New York State Office of Science, Technology and Academic Research.

The funding provides support to the UB research team to develop software that will allow crystallographers to tap into new sources of computational and scientific power to automatically enhance SnB, the protein-structure determination software package now used throughout the world, which was developed by scientists at UB and Hauptman-Woodward Medical Research Institute (HWI) in Buffalo.

When it first became available publicly in 1995, SnB itself represented a quantum leap in structure determination, allowing researchers to solve structures with virtually no input from the user.

Last year, the formula on which it was based, developed by Nobel Laureate Herbert Hauptman, president of HWI, was designated one of "The Top Ten Algorithms of the 20th Century" by Computing in Science & Engineering Magazine.

Through refinements and new editions of the software developed by UB and HWI researchers, this "black box" approach now is capable of solving the structures of molecules of thousands of non-hydrogen atoms. It also can solve critical substructures that allow for the determination of proteins with tens of thousands of atoms, which only a few years ago would have been regarded as impossible.

As emerging research in the fields of genomics and proteomics has begun to reveal, unraveling the structures of larger, complex proteins that may be important targets for new pharmaceuticals requires even more from SnB.

"SnB has had an enormous impact on the crystallographic community," said Russ Miller, Ph.D., UB Distinguished Professor in the Department of Computer Science and Engineering, principal investigator and director of UB's Center for Computational Research. "But its ultimate potential is unknown. This grant will allow us to make advances in structure determination by exploiting new computational paradigms."

One of those new paradigms, he explained, is the computational grid, a state-of-the-art platform in which hundreds, even thousands, of computers from across the nation and the world can be tapped to solve a single problem. The grid, he explained, harnesses not just the machines' computing power, but storage, device and personnel resources so they behave as a single environment.

"Creating a grid-enabled version of SnB puts us in a position to really tackle larger structures and at the same time create collaborative environments between users of programs who can benefit from each other's expertise without working together in a traditional sense," said Miller.

The project also enables Miller and his colleagues to take advantage of new techniques in data mining, allowing a software "agent" to sift through enormous amounts of data that are deposited in an SnB data warehouse at CCR and identify critical patterns that can lead to more efficient structure determinations.

"Once the data are in the repository, the software agent will continually mine them in an effort to determine adjustments in parameter settings," said Miller, "so the next time someone uses the package to solve a structure with characteristics similar to those in the warehouse, SnB will be even more efficient.

"Essentially, it will be able to improve itself," he said.

Once the software has been written that creates the appropriate collaborative and data-mining environments, the UB and HWI developers will begin work to create a virtual community.

"Initially, we will approach individual investigators to act as beta testers," said Miller. "We hope that users begin to understand and appreciate the fact that they will have to actively choose to participate in the community, but can still use SnB if they choose not to participate."

While scientists often are concerned about sharing data in the early stages of analysis, Miller noted that many are involved in virtual communities, such as efforts to write open-source software and contribute to certain types of databases.

"This is a slightly different type of virtual community, but if they want the benefits, they will participate," he said. "Crystallographers want to be able to solve protein structures with 'black boxes,' and this will provide them with an opportunity to improve the effectiveness of their studies while engaging in the cutting edge of computational science."

To create the grid-enabled version of SnB, CCR staff are harnessing key resources there that include a Dell server cluster that is one of the world's largest commodity-based clusters and a Pentium4 Dell cluster with Myrinet interconnectivity that is one of the most powerful machines in the world, as well as other machines from SGI, IBM and Sun.

Using a FakeSpace ImmersaDesk powered by SGI Onyx 2 and SGI 3300W visualization systems, CCR's visualization scientists are producing a collaborative environment that will allow SnB users around the globe to visually work together on structures.

In addition to Miller, investigators on the grant are Josephine Anstey, Ph.D., assistant professor of media study at UB who works on collaborative environments; Charles M. Weeks, Ph.D., senior research scientist at HWI and UB professor of structural biology, who originally developed SnB with Miller and Hauptman, and Aidong Zhang, Ph.D., UB professor of computer science and engineering, who studies and develops techniques for data mining.