Researchers reach the Holy Grail of computational chemistry

Tom Cundari and Angela Wilson
Through math and computer modeling, University of North Texas chemists Tom Cundari and Angela Wilson and their research groups have found more efficient ways to understand the thermodynamic properties of molecules, which is vital in designing new materials such as drugs.
Friday, December 22, 2006

Explorers have gone the world round seeking the Holy Grail, but two University of North Texas professors have found their version in a computer laboratory.

Angela Wilson and Tom Cundari reached what they call computational chemistry's Holy Grail in December 2005, after several months of preparation but only five days of computer simulations.

Through math and computer modeling, the professors merged aspects of several existing quantum mechanics theories into a single hypothesis. They then found more efficient ways to understand the thermodynamic properties of molecules, which is vital in designing new materials such as drugs.

"We combined some of the ideas others have used with a very good way of obtaining accurate energies," explains Wilson, an associate chemistry professor whose contributions were in theory and method development. "Others had not combined those two thought patterns. We decided to, and it worked."

Their achievement appeared in the March 2006 issue of the "Journal of Chemical Physics," and it's attracting attention to UNT as a national center of excellence in the emerging field of computational chemistry.

In computational chemistry, the results of theoretical chemistry are translated into computer programs to calculate the structures and properties of molecules and solids, and the programs are then applied to real chemical problems.

The accomplishment of Wilson and Cundari, attained under the auspices of UNT's year-old Center for Advanced Scientific Computing and Modeling that they co-direct, enabled them to use quantum mechanics to predict energies to within one kilocalorie per mole with less effort than previous techniques. (A mole is an amount of a chemical substance.) The intent is to learn how much energy is associated with the arrangement of the atoms within a molecule and how much it takes to make a chemical reaction happen, or to make a molecule form or break apart.

Wilson says the accomplishment means that, as computational chemists, she and Cundari "have reached the Holy Grail, and now we can actually address larger chemical molecules at much higher levels of accuracy than other groups have been able to do."

Cundari, a professor of chemistry whose focus is on practical applications, explains why:

"This allows us to design and predict the feasibility of chemical processes from the ground up," he says. "It's not just better bean counting, but a level of accuracy that you need to design new chemical processes."

In about a decade, Cundari foresees the new approach resulting in, for example, faster idea-to-market delivery of new medicines.

"One thing we are trying to understand is the thermodynamics of drugs interacting with enzymes, which determines how effective a drug is, and that effectiveness is reduced to thermodynamic differences between useful and toxic," he says. "There's a very small energy difference between good for you and bad for you."

He adds that every molecular state has an energy associated with it

"You want a good molecular state to have a favorable energy. It sounds very Zen-like," he says. "In chemistry, the differences between a good state and a bad state are very small, close to one kilocalorie per mole. What we are doing has a difference of one kilocalorie or less."

With the new method, dubbed the correlation consistent Composite Approach, or ccCA, pharmaceutical companies eventually could bring out new drugs "faster, safer and cheaper," Cundari projects.

Chemists currently look at tens of thousands of compounds searching for those that will produce the desired result with the fewest side effects.

With ccCA, however, "we can now predict the properties of those compounds with high accuracy and confidence," Cundari says.

"This approach allows us to become better handicappers, like in a horse race. We can isolate the best material to be tested, with the least side effects, the most effective, and so on," he says.

Not that this will happen immediately. Wilson and Cundari worked with small molecules, such as water, which has two hydrogen atoms and one oxygen atom. Scaling up to larger, more complex molecules means vastly more number crunching, but they're working on it.

As of May 2006, in fact, the professors and their team of graduate students had achieved the critical one kilocalorie per mole with hydrocarbon compounds containing up to 15 carbon atoms, and "we think we can do better than that," Wilson says.

"Our computer and mathematical methods seem to be working quite well," she says.

She acknowledges that other, more cumbersome methods can be used to obtain similar results, but says her team's work is the first practical technique.

"We thought others weren't going about the methodology the right way," she says. "They worked, but there were a lot of 'fudge factors' involved. We did it with pure mathematics, without fudge factors."

Wilson notes that on an octane molecule containing eight carbon atoms, a Pacific Northwest National Laboratory group used 1,400 computer processors computing simultaneously for a day to reach one kilocalorie per mole. The UNT team took five days - but significantly used only four processors, or basically four desktop computers. That, of course, is much more realistic computing power for the typical pharmaceutical lab wanting to follow their lead.

To explain the significance, she draws an analogy of how much time the method that she and Cundari developed saves. If all of the computations had been done using just a single computer processor, previous state-of-the-art methods would have required 3.3 years to complete them, while their method would have taken just 20 days, she says.

"A 20-day calculation is doable by most researchers," Wilson says. "A 3.3-year calculation is not. And most people don't have access to 1,400 processors."

Kirk Peterson, a chemistry and materials science professor at Washington State University in Pullman, Wash., says what Wilson and Cundari achieved is "more reliable and quicker" than previous methods of reaching the magic kilocalorie per mole barrier.

"This is one of the Holy Grails of computational chemistry, and for some chemists this is the Holy Grail to do this for a large variety of molecules. If this can be scaled up, it will be very useful in designing new drugs and materials," he says.

Wilson says she and Cundari have been happy with their results, but not satisfied.

"We think we can do it larger, we think we can do it faster and we think we can do it more wisely. We haven't really gone into the program and used all its bells and whistles," she says.

Unlike chemists who mix a little of this and a little of that in lab beakers, Wilson and Cundari arrived at ccCA through computational chemistry, where test tubes are replaced by quantum mechanics and number crunching on a sophisticated network of linked computers. UNT has built one of the strongest and largest computational chemistry programs in the country within the last five years, with seven professors and about 50 students.

The work has certainly not been without its challenges, Wilson says, noting the demands of marrying theories with better methods that take advantage of the positive aspects of each theory while reducing inaccuracies. Solving the thousands of mathematical functions that describe the motions of electrons around atoms and molecules was another challenge, she says.

"It was difficult because to reach this level of prediction in terms of computational methods you have to do a very, very good job of addressing the quantum mechanics involved," she says.

That, though, is one of her joys. The first UNT faculty member to receive the National Science Foundation's highly competitive CAREER Award, Wilson says she absolutely loves mathematics and using math to understand chemistry and physics.

"It excites me to understand important problems using all of it," she says.

But the computations in ccCA were still far too massive for Wilson and Cundari to handle alone. Helping greatly, along with other graduate students, was postdoctoral research associate Nathan DeYonker, whom Wilson says has "great chemical intuition."

"Nate had the ability to see through the mountain of data he was generating and find patterns," she says.

For how large a molecule can ccCA obtain critical data?

Wilson says the answer to that depends on the computer resources available, but a Grand Challenge Grant she received during the 2006 fall semester from the U.S. Department of Energy will provide thousands of hours of computer time on some of the world's most powerful computers, located at the DOEs laboratories in Richland, Wash.

UNT News Service Phone Number: (940) 565-2108