Argonne software help decode German E. coli strain

genomes and the new requests began to pile up — reaching more than 200 genomes an hour at one point. Its operators wanted to prioritize the E. coli work, so they turned to a resource designed for just such a possibility.

Magellan is a DOE test cloud computing project designed to boost research by making additional servers available on demand for scientific computing. The program, partially funded by the Recovery Act, has two sites — one at the Argonne Leadership Computing Facility and one at the National Energy Research Scientific Computing Center in California — but is designed to give researchers across the nation access to computing power in times of need.

The Argonne team duplicated the RAST server on Magellan, rapidly increasing the available computing power. “Our system is designed to use clusters, so we engineered it so that a piece of Magellan became part of the cluster that we use for RAST,” explained Bob Olson, an Argonne computer scientist who maintains RAST.

It worked — so well that even more submissions poured in. Argonne and Virginia Tech teams worked around the clock that weekend to keep the servers running smoothly.

“They found exactly what they were looking for,” Overbeek said. “The difference between this new strain and older ones came down to just a few genes. Apparently, the new strain included a combination of virulence factors present in other studied strains.”

The operation was a perfect case for Magellan, Olson said, because each genome submission is an independent problem. Simply adding more processors to handle the extra jobs is easy—unlike many other computations, which often must solve successive problems; processors must wait to start their jobs until another processor finishes.

Overbeek remembered the early days of annotating genomes in the mid-1990s, when it took four or five scientists more than a year to analyze just one genome. “Now we can spit them out in a few hours,” he said, and the team has already tested the next generation of RAST — a version so fast that it cuts the time for annotating a typical E. coli genome from eight hours to just fifteen minutes.

“RAST is really revolutionary,” Overbeek said. “It’s turned a problem that used to be insurmountable into one that is trivial.”

There is even an iPhone app to submit and receive genomes from RAST servers.

Developed at Argonne, RAST is funded through the National Institutes of Health (NIH) and run by the Pathosystems Resource Integration Center (PATRIC) at the Virginia Bioinformatics Institute. PATRIC keeps a publicly available database of sequenced genome information.