Duke software dramatically speeds enz...(more like they would in real life. Moreo...)

"Thus it not only addresses flexibility. It embraces it," Donald said." "So we might be able to quickly discover things that would take a very long time through purely experimental techniques. It should, in principle, be possible to redesign any enzyme simply by inputting the protein's shape into the algorithm and telling it what you want it to do."

The new report is the latest milestone in a nearly 10-year effort to develop reliable enzyme design algorithms, Donald said. In the process, his group has had a long collaboration with Amy Anderson of the University of Connecticut to biochemically test the algorithm's accuracy using the Gramicidin S Synthetase enzyme system, which produces the natural antibiotic in Bacillus brevis bacteria.

In the new PNAS report, Ivelin Georgiev, Donald's computer science graduate student who is one of K*'s designers, used the latest version of the algorithm to redesign the first step in the biochemical assembly line needed to make the antibiotic.

"Redesigning that first step is a big achievement," Donald said. "We are now beginning work on redesigning the half dozen subsequent steps downstream."

The algorithm includes a "dead-end elimination" feature that can run though all possible chemical interactions and flexible molecular architectures to weed out scenarios that cannot work. Calculating just one redesign might take up to a week in the 230-processor computer cluster housed in Donald's lab, he said.

After all the calculations were completed, biochemist Cheng-Yu Chen, another of Donald's graduate students, confirmed the algorithm's predicted designs in Donald's biochemical wet lab, using bacteria to synthesize some "quite big and tricky proteins," Donald said. "It was not at all trivial to do that, and testing the functions of
'/>"/>