The project makes use of an IBM eServer p595 supercomputer along with IBM WebSphere and DB2 data management software valued at more than $2.5 million under IBM’s Shared University Grant program. Led by the University Health Network’s (UHN) Dr. Igor Jurisica, it will also involve collaboration with Big Blue’s T.J. Watson research centre in New York state. The bulk of the work is taking place at the Life Sciences Centre in the recently opened Medical and Research Sciences (MaRS) building in Toronto.
Jurisica said the supercomputer will be grappling with interactions of around 50,000 human protein samples and about 330,000 proteins from other animals. The proteins are plotted like nodes on a graph, which are then analyzed to look for substantial similarities and differences. This could help improve the care given to those who contract ovarian, colon and prostate cancer, he said. The high-performance IT equipment is necessary in part because the complexity, as well as the volume of data renders it invisible on traditional systems, he said.
“Even on multi-core computers, (coming up with an analysis and predictions) can take months,” he said. “We’re trying to come up with heuristics that will give reasonably approximate answers. Even if you’re using large high-resolution displays, however, you cannot really effectively display a graph on your screen with 50,000 edges and several tens of thousands of nodes. You would basically see a black box.”
Don Aldridge, IBM Canada‘s general manager for Higher Education, Research and Life Sciences, said the company has been working with Jurisic since he was a graduate student at the University of Toronto. His track record there, as well as his involvement with IBM’s Centre for Advanced Studies (CAS), ensures Big Blue is making a sound investment, he said. This is the third grant IBM has given Jurisic and his team.
“There’s a certain amount of enlightened self-interest involved,” Aldridge said, “because we have folks who are specifically looking at the intersection of biology and computer science.”
Jurisic said the compute requirements of his work surpass expectations very quickly, which makes it hard to forecast what the team should be spending on over time. When he first got involved in protein crystallography in 1998, for example, they didn’t realize how the number of proteins to be screened and analyzed would exponentially increase.
“You don’t want to come up with a four-year plan because we wouldn’t necessarily have the resources to build it, and by the fourth year the technology would change and we would need to adjust things accordingly,” he said. “In this case DB2 and other middleware is supporting what we’re doing from day zero and scaling with it.”
IBM has formed several other partnerships to create IT systems that would investigate cancer causes and cures, including one with Memorial Sloan-Kettering Cancer Center (MSKCC), the Molecular Profiling Institute and the CHU Sainte-Justine Research Center in Montreal.