A massive new push to predict the three-dimensional (3D) structure of proteins based on their amino acid sequences was launched last week by a coalition of European labs and bioinformatics groups.

With the aid of a high-performance Cray computer at Silicon Graphics and newly developed software, the group plans to analyze all the 200,000 protein sequences available in public databases and model as many as possible, predicting how they fold up into their final 3D form by comparing sequences with those of the 5000 known protein structures. "We hope we'll get some information for up to 65,000 proteins," says team member Manuel Peitsch at Glaxo Wellcome's research center in Geneva. "Structural features of sequences showing high levels of similarity to sequences of known protein structures can be predicted, but we want to move to the prediction of structures showing lower levels of similarity," says another participant, biologist Michael Sternberg at the Imperial Cancer Research Fund labs in London.

The $50,000 project, dubbed the 3D Crunch, will put its raw results out on the World Wide Web. Then it will be up to protein crystallographers and nuclear magnetic resonance analysts to see how close the computer was able to come to reality. "Matching their work against our predictions should help us determine which parts of our strategy are effective," says Sternberg.

Volume 280, Number 5368, Issue of 29 May 1998, p. 1353