The moment Hideaki Fujitani unlocks the heavy door and enters the room, the buzzing noise — which sounded like a simple hum from the outside — gets much louder.
“The noise is from the cooling fans,” he shouts as he tries to tidy up his hair, ruffled by the air blowing from two bulky black boxes with screen doors and from beneath the ventilation panels on the floor. “The machine is at work 24 hours a day.”
The “machine,” occupying a small room in a building on the University of Tokyo’s research campus in Tokyo’s Meguro Ward, is a computer system. It’s not like any other computer system, though: Equipped with those powerful cooling fans and a ventilation system, it is a supercomputer, which Fujitani believes will save people’s lives.
Fujitani, professor at the university’s Laboratory for Systems Biology and Medicine (LSBM), is part of an interdisciplinary group of scientists working to develop drugs for people with recurrent or advanced cancer. Led by Tatsuhiko Kodama — named one of the world’s 10 most influential scientists of 2011 in the Dec. 22 issue of British Nature magazine — the group is currently working on the development of drugs from antibodies. An antibody is a protein produced by the body’s B cells, and it circulates in the blood. As part of the immune system, antibodies recognise and stick to antigens, which are foreign molecules that form part of viruses and bacteria. Once bound to an antigen, an antibody can neutralise it.
A modified antibody could “dock” with an antigen specific to cancer cells and prevent the cancer cells from growing, and in the most desired scenario, it could kill them. Kodama’s group has already identified the target proteins and strategies for how to attack them. With the help of supercomputers, the scientists say they can make antibody-based drugs far more effective and less harmful to patients than other cancer drugs, which often attack healthy cells as well and result in severe side effects.
“In the future, many people in advanced stages of cancer will be able to have their illnesses cured, and live longer,” Fujitani says confidently.
Fujitani’s job, in particular, is to simulate exactly how the antibody and antigen interact and stick to each other at an atomic level. He also wants to find out in what situations they can be bound together most powerfully, like finding the right key to fit a keyhole, he says.
To do that, he has calculated, using molecular dynamics, the three-dimensional moves of 30,000 to 40,000 atoms that make up the antigen, antibody and water around them. “Even when the proteins (in the antigen and antibody) bind, they are easily blown apart, as they are affected by the movement of the water around them. So we need to calculate the dynamics involved in all of them.”
The size of molecules he studies is about 20 angstroms (one angstrom is one 10 billionth of a meter) and atoms move in about a femtosecond, or one millionth of one billionth of a second. To understand what is really going on, though, computers must work really fast and hard.
“A molecule moves in about 1 femtosecond, gradually changing the shape of proteins over microseconds,” he says. “To see the dynamics, you need to solve about one billion equations. If one CPU is able to solve one equation per second, it would still take 32 years to solve all the problems. That’s why we need the fastest supercomputer with lots of CPUs.”
Made up of 612 CPUs, the LSBM machine, purchased from Fujitsu, boasts 34.67 teraflops (the capacity to make 34.67 trillion calculations per second) in a benchmark computing program, which is widely used to gauge the speed of supercomputers. That’s equal to the computing capability of the Earth Simulator (ES) in Yokohama, which was ranked the world’s fastest computer in 2002.
By the time the university bought the system in 2010, however, the computer with the same speed as ES had become much smaller — and far more energy efficient — requiring only one fortieth of the power needed for ES, Fujitani says.
Soon after getting the supercomputer up and running in the summer of 2010, Fujitani began running simulations on it for about a month, and by the end of October, he got results for the necessary proteins’ structure. When he compared the results with an X-ray image of a real antibody frozen at minus 230 degrees, the crystal structures of the two matched perfectly — verifying the accuracy of his simulations, he says.
Computer-aided drug design has made great advances lately, Fujitani says, though he pointed out that, in as recently as 2006, the reliability of computers for use in drug development was a subject of big debate among experts. Fujitani, who started biocomputing research at Fujitsu in 2002 before moving to the University of Tokyo in April 2010, says that for a long time he had great difficulty getting support from his former employer and Japanese pharmaceutical companies for the idea of using computers to design drugs: “(LSBM’s) Dr. Kodama was the only one who became serious about it.”
Fujitani says his research would have been impossible without the support of the national government, as private-sector drugmakers do not have the resources to invest a huge amount of money in supercomputers. Kodama’s team, as well as the ¥500 million supercomputer, receive Cabinet Office funding for innovative R&D programs. Fujitani will also be leading a project to use the K computer (see main story) for drug design, which would enable even speedier all-atom simulations.
“The K will be 240 times faster than the machine here, so we can do the calculations much more quickly, and run different programs simultaneously,” he says, noting that he will be in charge of simulating four or five drug tests on K, including ones for leukemia and lung cancer. “What takes a month to simulate here will be done in three or four days.”
But he is skeptical of the ongoing discussions in Japan among academics on the development of the so-called exa-scale computers. “There is no talk yet of what you want to use such computers for,” he says.
As an example of forward-thinking approaches, Fujitani cites the work of D.E. Shaw Research (DESRES) in New York, a computational biochemistry research lab headed by billionaire scientist David Shaw. Also the founder of a major hedge fund named D.E. Shaw & Co., Shaw foresaw the importance of biocomputing and set up the DESRES in 2002. Now chief scientist at DESRES, he has poured in his own money into hiring some 100 scientists and building 11 supercomputers designed to do molecular dynamics simulations only. Shaw has also tied up with major pharmaceutical firms, to push for computer-aided drug design.
“(DESRES) is a typical example of how you can build fast computers without making them as big as K,” says Fujitani. “(Shaw) himself is very enthusiastic about creating new businesses. Japan must be more strategically-minded in creating new fields and new industries. Otherwise, it has no chance of survival.”