Computer predicts protein's shape from sequence alone
25 September 2002
Using computer simulations, scientists have correctly predicted how a protein folds into its final shape purely from its genetic code .
A protein's form dictates its function. Forecasting what these molecules of life look like from their gene sequence is one of the most important challenges facing biology following the decoding of the human genome and that of other animals.
It would have, as the late biochemist Peter Kollman of the University of California once said, "a tremendous impact in all of biotechnology and drug design".
Proteins are made of chains of small molecules called amino acids - instructions encoded in genes in DNA govern which amino acids, and in what sequence. But scientists still don't know what most of our proteins look like, or what they do, because the amino-acid sequence of a protein is just the starting point.
A protein molecule acquires its shape as its amino-acid chain folds into a compact, three-dimensional blob. Currently researchers use laborious experimental techniques to reveal the positions of each atom in such blobs.
It ought to be possible to predict a protein's structure using computers to simulate how chains fold, knowing how amino acids tend to attract or repel one another. But, because many proteins contain hundreds or even thousands of amino acids, this problem is too complex for today's computers.
So Carlos Simmerling of the State University of New York at Stony Brook and colleagues started small. They worked on a tiny artificial protein called Trp-cage. This consists of just 20 amino acids, and was built by hand last year by Jonathan Neidigh and co-workers of the University of Washington in Seattle. It folds into a compact, well-defined shape like a much larger protein; most short chains remain loose and floppy.
When the Washington team worked out the mini-protein's shape using nuclear magnetic resonance spectroscopy, Simmerling's predictions fitted the measurements almost exactly.
Many researchers have attempted to simulate protein folding before, but none has followed the process right through to the final shape in quite this level of detail. The problem gets rapidly harder as protein size increases, so researchers are still a long way from being able to predict the structures of most natural molecules. But the Trp-cage success suggests that we do at least now have the right tools for the job.