RESEARCHERS SEEK LOW-TECH SOLUTIONS FOR HIGH-TECH DRUG RESEARCH

FAYETTEVILLE, Ark. — University of Arkansas researchers plan to create a computer program that will help speed the process of predicting the properties of molecules-a process that currently costs drug companies billions of dollars a year.

Peter Pulay, Distinguished Professor of chemistry and biochemistry, and Amy Apon, associate professor of computer science and computer engineering, have received $488,000 from the National Science Foundation to create a computer program that will allow chemists to store and rapidly retrieve large amounts of data on clusters of computers.

Chemists at drug companies spend much of their time screening molecules for potential pharmaceutical benefits. This used to be done "in vitro," or in a test tube, which takes lots of time and material. Today’s technology allows researchers to screen molecules "in silico," or using a computer, Pulay said. But drug researchers want to screen as quickly as possible, and that can prove to be difficult and expensive.

"The faster the data retrieval is, the more it costs," Apon said. Indeed, drug screening comes at a high price; the average initial screening of molecules for a potential drug costs companies about $200 million.

This is because predicting the physical properties of even middle-sized biological molecules using quantum mechanical calculations requires a lot of computation, Pulay said. The molecular information can take up terabytes of memory-a terabyte equals 1,000 gigabytes, or 1,000,000,000,000 bytes, enough information to fill about 200,000 books.

Researchers seek knowledge of molecular structure because structure determines function, and changes in the structure can enhance or inhibit certain molecular properties. This relationship between structure and function may help biomedical researchers create designer drugs or discover the underlying mechanisms of a particular disease. If scientists know the electronic structure of a molecule, they also may be able to predict how it will behave in different situations--and might be able to build a synthetic molecule that could mimic its natural counterpart.

But calculating the structures of all but the smallest molecules remains a daunting task, because the electron energies must be calculated with respect to one another to get an accurate structural picture. Because of the number of electrons buzzing around in each atom, the required calculations can take thousands of hours of computer time.

Pulay uses computer clusters, single computers wired in parallel, to make such calculations. However, computer clusters have limitations based on the time and cost of one computer communicating with remote computers, Apon said.

"You have to find a way to hide the communication costs," she said. That is what she and Pulay will work to do over the next three years.

Once created, the computer processing program will have applications in other areas, including large image processing calculations, atomic structure calculations and large simulations such as wind tunneling, Apon said.