An IBM supercomputer directly related to the famous Deep Blue that battled chess grand master Gary Kasparov in 1997 has recently been recruited to aid Canadian researchers in the fight against cancer.
The Life Sciences Discovery Center of the University Health Network (UHN) in Toronto is using an IBM eServer p595 supercomputer to analyze protein interactions.
These interactions are vital to the understanding of cancer biology at the molecular level. The process is crucial to charting a tumor's progression and ultimately developing treatment.
The UHN center received the supercomputer along with IBM's WebSphere Information Integrator server application and DB2 database software under Big Blue's Shared University Grant program. The package is worth more than US$2.5 million.
"One of the biggest challenges in our research is being able to sort through a large volume of data," said Igor Jurisica, a scientist with the Ontario Cancer Institute of the UHN. Jurisica is also a professor of computer science and biophysics at the University of Toronto.
With a lesser computer, Jurisica estimates his calculations would take months but with the p595, they can be accomplished in weeks or days, shaving wait times by 50 per cent. "With 120,000 known proteins and 25,000 to 30,000 we're looking at millions of possible interactions. You can see we have quite a real big hair ball."
The p595 supercomputer used by the UHN "is a descendant of Deep Blue" that famously trounced Kasparov in a series of man versus machine chess matches in 1997, according to Don Aldridge, IBM Canada's general manager for higher education, research and life sciences.
Deep Blue was a 1.4-ton IBM RS/6000 SP high performance computer that had 256 Power2 processors. It was specifically created to play chess and could explore 200,000,000 positions per second. At the time of the match Kasparov could explore approximately three positions per second.
Deep Blue technology was eventually used to tackle other real world problems such as forecasting weather patterns.
"In the case of Deep Blue it was a natural progression to evolve from predicting chess moves to predicting hurricanes because in both cases you are dealing with multiple variables," said Dave Marks, consulting director for the research and consulting firm IDC.
According to Marks, the desktop-sized p595 supercomputer inherited a lot of properties from Deep Blue "but is much faster in terms of computing capacity and consumes less power."
Aldridge said p595 can be configured in different ways to deal with specific situations, but that the UHN is "stretching it in ways we would not have thought possible."
Learning is a two way street, said Aldridge. "When scientists like Igor [Jurisica] come to the IBM Lab it gives our researchers a chance to work with our equipment in real world situations."
Marks said IBM is working on another supercomputer project dubbed Blue Gene with the mission of studying molecular functions and protein folding.
Proteins react with each other in varied ways. Jurisica's team is trying to determine which patterns are disruptive or prone to cause tumors.
Molecular profiles of cancers help detect signs that pathological tests might have missed at the tissue level, Jurisica said. "For instance, it is known that women with mutations in their BRCA1 and BRCA2 genes are predisposed to breast and ovarian cancer."
However, he cautioned that it is not a black and white situation. "By tracing protein patterns and factoring in variables such as medical history, habits and environment, we could determine whether the person is simply predisposed or if cancer is already present."
This determination, Jurisica said, could spell the difference between providing early aggressive treatment and needlessly lowering the standard of life exposing the person to unneeded medication.