It’s the most powerful supercomputer on the planet – a worldwide network of PCs dedicated to researching the mysteries of protein folding.
There's a table of the world's most powerful supercomputers, and IBM's Roadrunner supercomputer is currently on top of the www.top500.org chart. With 129,600 cores, the mammoth machine has a quoted number-crunching rate of 1.46 petaflops. This is incredible stuff, especially when you consider that most of us haven't really thought about anything in terms of petaflops before. However, even Roadrunner looks decidedly weedy compared with the power of Stanford's Folding@home project. Its computational power has now surpassed the five petaflops mark.
Comprising a huge network of Internet-connected PCs, Folding@home is really the world's most powerful supercomputer. You may well have crunched through a few work units for Folding@home, and you probably know that it benefits medical research. But what exactly is its purpose, and has it produced any worthwhile results?
What is a protein?Let's start with the science behind Folding@home, which is protein folding. In simple terms, a protein starts out as a long string of amino acids, but it can't perform its biological function until this string is transformed into a three-dimensional shape. This transformation is called folding, and once the protein has folded, it can then perform its biological function.
However, proteins can also misfold and take on a different three-dimensional structure. A simple example of this can be seen when cooking an egg. Crack an egg into a pan, and you can see the proteins of the egg white in their natural runny state. As you apply heat to them, the proteins start to unfold. After this, the amino acids from the different proteins that comprise the egg white will then mix, causing the egg white to change texture and become solid. However, proteins changing like this inside the body can have drastic consequences, particularly as the process is, as in the case of frying an egg, often irreversible.
A scientist who has been researching protein folding for many decades is Professor Harold A Scheraga, based at Cornell University near New York. "There are genetic diseases," explains Scheraga, "in which a particular amino acid might have been mutated to another one. That will change the interactions between the different parts. Because one amino acid is different, it doesn't fold into the proper form, so it doesn't perform its biological function.
"Sickle Cell Anaemia is a good example of that. One particular amino acid is changed, and the haemoglobin - that's a protein in the red blood cell that binds oxygen - starts to aggregate one molecule with another to form a big glob instead of folding properly. It distorts the structure of the red blood cell into the form of a sickle. Since haemoglobin has to bind oxygen in an un-deformed red blood cell, and carry it around in the blood, the red blood cell isn't functioning well, so somebody who has that disease won't get enough oxygen."
As Scheraga points out, this is just one example of a situation where misfolding leads to diseases. Another prominent misfolding disease is Alzheimer's, which also results from proteins misfolding and aggregating with other molecules to form big globs. "In Alzheimer's, the misfolding forms big clumps in the brain," says Scheraga.
Studying how and why proteins misfold has been a major challenge for medical research scientists for decades, and Scheraga used some of the early forms of computational modelling of the process. "I was one of those who started the field," says Scheraga. "We had limited computer time, and we could work with small chains. What is small? Five amino acids, and then we got up to ten and so on."
Scheraga now has his own supercomputer in his lab, featuring 800 CPUs, and his colleagues have now developed the code to calculate the total energy of a string of up to 1,000 amino acids. This is a lot of computer power, but it's still not enough. He has to apply for additional CPU time at national computing centres in the USA and even in Germany.
Issue: 107 | December, 2009