- The Washington Times - Monday, December 17, 2001

Most of us remember Deep Blue, the chess-playing supercomputer from IBM that beat Gary Kasparov, the world chess champion. Having a taste for really massive computers, I called IBM and asked whether they were doing anything these days in the supercomputer line.
Yes, they say. They are putting together a couple of real monsters. I talked about them with Bob Germain, a manager at the company's Computational Biology Center. A physicist by training, he is now a de facto molecular biologist.
"Computational biology" is a curious phrase for anyone who studied biology more than a decade ago. Aren't biologists supposed to put bugs in bottles and dissect things? Why, you might ask, do they need the world's most humongous computer?
We'll get to that.
IBM is working with Lawrence Livermore National Laboratory, says Mr. Germain, on a machine called Blue Gene/L which, says IBM, will have more computational power than the 500 biggest computers in existence today. It will go live in about 2004. (In another column we will worry about how it works.) IBM says it will run at 200 teraflops i.e., in a second it will do 200,000,000,000,000 calculations. Its bigger brother, also in the Blue Gene family and due out roughly in 2006, will be five times as fast.
This is fast beyond belief.
Which brings up a journalistic problem: How do you make a machine like that mean anything to people (like me) who are not computational physicists? Computers are so insanely fast these days, and get faster so fast, that the numbers numb the mind. What are these machines good for? What can you do with them?
Let's take a swing at the answer.
The bigger of the two machines is intended for the study of certain problems in molecular biology, specifically the folding of proteins. Much of the behavior of the body at the cellular level is managed by enzymes.These are protein molecules that control chemical reactions, which we are all a bundle of. No reactions, no us.
Enzymes work largely because of their physical shapes: An enzyme molecule may have a depression like a slot, for example, and something that it needs to stick to has a projection, so they fit together, which allows them to do whatever they are supposed to do. It's a bit like a three-dimensional jigsaw puzzle. So the shape of enzymes is important.
Now a protein, which enzymes are, is a very long chain of amino acids stuck together like beads on a string. Different amino acids in different orders give you different proteins. But the strings of acids don't just flop around in cells like pieces of rope.Because the acids have different electrical charges and such, the strings crumple and fold into particular shapes.
This folding has practical consequences, explains Mr. Germaine. "A single wrong amino acid in a protein can cause disease by causing the protein to fold the wrong way."He gave cystic fibrosis as an example.
In short, the body is crucially dependent on protein chemistry.So a whole lot of scientists want to understand how proteins fold and why.
As it turns out, the amount of computation necessary to simulate the folding of a fair-sized protein is enormous.Comparatively simple examples can keep a fast machine running for months.According to Mr. Germaine, even Blue Gene won't handle all of the protein problems scientists would like to throw at it. But, he says, it will come much closer than anyone has come before.
This is part of a bigger picture.A lot of scientists believe that the life sciences are about to grow explosively, if they aren't already doing it.The reason, they say, is that humanity is finally learning enough about how living things work to do useful things.
To cure a disease caused by problems within cells cystic fibrosis, cancer, Alzheimer's, sickle-cell anemia and so on you have to understand how cells work. Until then, somebody said, you are like someone trying to repair an engine without knowing what pistons are for. When the general level of understanding is good enough, problems that were hard become easy.
We're about there or so many believe. In large part progress occurs because instrumentation keeps getting better: the laboratory gadgetry that lets scientists look into cells.Oddly enough, huge computers are now biological instruments.
So that's why this particular supercomputer is important.

LOAD COMMENTS ()

 

Click to Read More

Click to Hide