Friday, September 14, 2007

2000 Supercomputers Help University Of Idaho Scientists Explore Genetics And Bioinformatics

Date: September 7, 2000

Supercomputers Help University Of Idaho Scientists Explore Genetics And Bioinformatics

Science Daily MOSCOW, Idaho -- The mapping of the human genome is the tip of the iceberg that is the biological information revolution.

University of Idaho computer scientists and mathematicians are joining biologists to explore new ways to interpret the complex genetic information that describes all living things and their relationships.

Along the way, UI students returning to school this fall will find a new course few schools could hope to offer: building a new supercomputer.

A $500,000 National Science Foundation grant will allow James A. Foster, an associate professor of computer science, and a team of fellow scientists to build the university's expertise in bioinformatics.

The network-based supercomputer known as a Beowulf cluster computer will be built with the help of the UI class taught by Robert Heckendorn, an assistant professor of computer science. It will deliver high performance computing to scientists on the Moscow campus.

Foster's NSF grant follows a $90,000 National Institutes of Health senior fellowship that he won last year. Through the NIH grant, Foster worked with Holly Wichman, a UI professor of zoology, to understand evolutionary biology. He hopes to apply its rules to computer science, specifically genetic programming.

Foster and Wichman are founding members of the UI Initiative for Bioinformatics and Evolutionary Studies. The grant will also support a seminar for senior-level and graduate students on bioinformatics, an emerging science.

Fellow members of the initiative group are Paul Joyce and Steve Krone, UI mathematics professors. They recently won a $166,000 National Science Foundation grant to study methods to trace how selection and other factors can affect populations.

Foster's supercomputer actually will be the second owned by the UI. The first is used by zoologist Jack Sullivan, who studies population genetics. That machine, which was funded through a National Science Foundation Grant, is housed at the Smithsonian Institution. Dave Swofford, his partner in the phylogenetics study that is the basis for the grant, is based there.

Swofford is the leading developer of software to study phylogenetics or evolutionary relationships and needed the highest-speed connection to the supercomputer, Sullivan said. That is why the equipment is housed and administered at the Smithsonian in Washington.

The new supercomputer and grant will provide powerful computing resources to support the university's studies in bioinformatics and evolutionary studies, he said.

Ecologists and epidemiologists use mathematical models to trace the relationships among populations of plants and animals, including disease organisms. The goal is to sort through complicated or patchy information to understand links between individuals and groups.

Examples range from the evolution and spread of infectious diseases to invasions of exotic species like weeds, Krone and Joyce noted.

"From the genetic models, we want to look at how certain characteristics are passed down and change through time," Krone said. One way to approach this is by "looking back in time" to see how ancestral relationships and mutations cause the genetic characteristics seen in present-day populations. This backward-looking approach is called coalescent theory.

"Diversity is one general question that's at the bottom of a lot of things. Looking at coalescent theory is one way to calculate the amount of diversity in a population. We want to come up with models to describe and explain certain phenomena we see in nature," Krone said.

Joyce and Krone's interest is in the probability and statistical models that can be used to trace such relationships. Using mathematical methods to track genes while accounting for natural selection and other forces is complicated and fraught with randomness.

The same sorts of complexity and randomness that is part of biological evolution also holds sway for the computer programs Foster is interested in. "It's abstract enough that we could use it for our work building software that can adjust to new conditions or even repair itself," he said.


SIDEBAR: Students Get Hands-on Experience Building Supercomputer in UI Class

MOSCOW, Idaho -- The age of high-performance computing once symbolized by supercomputers costing millions of dollars apiece, may be closer to home than we think, a University of Idaho computer scientist says.

Robert Heckendorn, an assistant professor of computer science, will oversee students enrolling in a directed study class that will design and build a Beowulf cluster computer this fall. The new supercomputer is one of a class of low cost, high performance computers built with off-the-shelf components.

"We probably won't recognize them when they do go into use," said Heckendorn. Homes may be equipped with versions to customize heating or cooling systems, lights and appliances.

But until then, the new system the class will help design and build has the feel of the earliest days of home computing, when hobbyists worked in their basements or garages to assemble their own computers. "This is like building a hot rod," he said.

The students will work with a budget of $44,000, developing specifications for the equipment and recommending software and hardware for the supercomputer.

The budget comes from a $500,000 National Science Foundation grant won by James A. Foster, a UI associate professor of computer science, and a team of UI researchers to expand the university's expertise in bioinformatics.

"The students will put together a proposal as a class and make a recommendation about how to proceed. I will make the final decision but I don't want to steer the students by talking about it right now. I have an idea of the direction we'll take, but I may be wrong," he said.

The students will work on every step of the project, from determining the requirements the supercomputer must meet, though the purchase, assembly, software selection and installation. "They are involved from start to finish. It should be a great experience for them," Heckendorn added.

Beowulf cluster computers are the rage in high-performance computing these days. The clusters of computer processing and memory chips that form their heart can be assembled relatively cheaply from simple components. Off-the-shelf computers built for the consumer market can do the job.

"It's commodity computing. If you can only buy commodity computers and hook them together with the right stuff in the right way, you can get supercomputing power," he said. Although multi-million dollar specialty supercomputers still dominate the high end of the market, Beowulf-style supercomputers are gaining.

The class already has an experienced hand on the roster: Andrew Shewmaker, a UI senior from Kimberly, Idaho, studying computer science. "As far as I know he's the only one in the class who has worked with one," Heckendorn said.

Shewmaker learned about Beowulf supercomputing while interning at the Idaho National Engineering and Environmental Laboratory during the past three summers. "They've built two Beowulfs in the last two years," he said.

Although he wasn't involved in building the clusters, Shewmaker said, "I did get to help administer and use them."

"I am extremely excited about this project," Shewmaker said. "I am interested in building a Beowulf because I have read tons of information about it on the Internet and in books but I have yet to apply what I have read."

Note: This story has been adapted from a news release issued by University Of Idaho.

No comments: