An unfolding drama: Proteins unravelled

ONE of the greatest remaining challenges in science is working out how proteins curl up into their complex shapes. They do this in fractions of a second, always adopting the same three-dimensional form which determines their properties in the body.
When things do go wrong, and proteins misfold, diseases are the result: some forms of cancer, Creutzfeldt-Jakob disease and cystic fibrosis are all caused by slip-ups in the origami of the amino acids.
Knowing just how proteins fold could help both in understanding disease, and in designing drugs to treat it. Modern methods have made working out the sequence of the amino acids in any protein, once a chemical tour de force, into a simple process. But understanding how these sequences fold - what some scientists have called "the second half of the genetic code" - is much harder. It was once calculated that even with a simple protein containing 100 amino acids, the number of possible shapes is so enormous that it would take the fastest computers thousands of years to work them all out.
Faced with this, some doubters have questioned whether it will ever be possible to simulate the folding of proteins on a computer. But in the current issue of Science, two chemists from the University of California in San Francisco claim to have done it. Dr Yong Duan and Dr Peter Kollman used a Cray T3E, a supercomputer consisting of hundreds of central processors linked in parallel, to model the folding of a small protein, or peptide, with just 36 amino acids in it. The villin headpiece subdomain, as it is called, is found in the cells that line the gut. It is one of the smallest proteins that can fold autonomously, and its folded shape is known from structural studies using nuclear magnetic resonance. By programming the computer with the known data, the two chemists managed to simulate the folding process over a thousandth of a second. The first 150 millionths of a second showed the protein folding into a compact structure close to its known native state, then beginning to unfold again.
Though an exciting result, which suggests that protein folding may not after all be beyond the power of simulation, Dr Herman Berendsen, of the University of Groningen in The Netherlands, is cautious. In Science, he says that the real protein probably folds and unfolds hundreds of thousands of times until it stumbles into the most comfortable shape - the one with the lowest free energy. The simulation cannot capture that.
Another team, at the Technische Hochscule in Zurich, has tackled a simpler molecule, a beta-heptapeptide with 5,400 atoms rather than the villin peptide's 10,000, and simulated how folding begins at various temperatures - thus allowing direct comparison with experiment.
A recent conference in Santa Fe brought together biologists and physicists to discuss the question. Having cracked the structure of DNA and the genetic code, protein folding is the next challenge. "We're just beginning," admits Dr Steven Chu, of Stanford University, who won the 1997 Nobel Prize in Physics and has recently turned his attention to protein folding. Dr Berendsen agrees. We might have seen a glimmer of light, he says, but there is a long way to go.