Lab researcher's team shines in protein folding predictions

“This annual experiment is a delightful and terrifying phenomena for any protein scientist working on predicting protein structure,” Strauss said. “Participants in the experiment are sent a list of proteins whose structures have been solved but not made public. They are invited to stick their necks out and predict the structure of a true unknown. Then they and their peers go to the conference and find out the answer. They are assigned grades,just like in school.”

The computational technique, co-authored by the Strauss and Baker group, is known as Rosetta. The software is based upon the assumption that local sequence biases, but does not uniquely determine, the final structure. Predicted proteins are assembled from short fragments taken from a database of known structures with similar local sequences. Rosetta showed remarkable success in predicting the structure of folded proteins from its the linear sequence of amino acids.

“Some people call protein structure the second half of the genome project,” Strauss said. “Genes code for proteins, and a unique property of proteins is that they derive their function from their structure. Structures begin to tell us what proteins do, and how they do it, and we are developing ways to predict a protein ’s approximate structure from its sequence. Billions of dollars are being spent worldwide on measuring protein structure, so doing it computationally is obviously very desirable.”

There are between 10,000 and 100,000 important protein structures, but only a few thousand of these are currently known. Strauss specializes in predicting novel structures, ones that have not previously been seen.

“Nature recycles her structural designs,” Strauss said. “Once we know all the important structures, any new proteins will probably have structures similar to these, but so far we've only begun to get a sampling from direct measurement. So getting ahead of the experimentalist and predicting the structure of proteins with previously unseen designs --the novel folds – is a high value objective, not to mention the most fun and challenging. Predictions aren't nearly as good as measurements, and assessing prediction success is a relative kind of thing.”

“We ’re fascinated with the idea of tinkering with proteins,” Strauss said. “It ’s designing crescent wrenches at the molecular level.”