 |
Chemistry without Chemicals Supercomputer helps theoretical chemists
By John Toon |
There are no test tubes, reagent bottles, fume hoods or hazardous waste containers in the laboratories of theoretical chemists Rigoberto Hernandez and David Sherrill.
|
The work of these researchers takes place instead
at the keyboard of a powerful supercomputer. Using models based on deep
knowledge of chemical processes, the computer simulates the hopping of
electrical charges and breaking of chemical bonds at a level of detail no
other technique could provide.
Though the results must still be verified by old-fashioned experimentation,
computational chemistry allows scientists to ask more complex questions and
get faster, more detailed answers without mixing the first chemical. The
technique provides clues to chemical engineering mysteries that cannot be
investigated any other way, and reduces trial and error in research.
At the Georgia Institute of Technology, researchers use this technique to
study protein folding, anti-cancer drugs, molecules key to the vision process
and polymerization.
"If you can do the experiments on the computer and try all the 'what-ifs'
that way, at the very end, the one reaction you really need can be done in
chemistry," says Hernandez, an assistant professor in the School of Chemistry and
Biochemistry. "By finding the very best solution on the computer, you can
limit the waste products by eliminating trial and error. This doesn't do away
with experimentation, but it gives the chemist another tool."
Chemistry in Changing
Environments Hernandez, however, studies non-equilibrium reactions in which the reacting
chemicals form a large part of the overall environment. In such systems, the
environment changes as the reaction proceeds, affecting the chemistry in ways
that are difficult to model and study. This complex interaction between
reaction and changing environment affects the outcome in important ways.
"In many cases, the final properties of the material are determined by
their history," he explains. "I want to understand how things are formed, not
just to characterize their properties once they are formed."
Protein folding provides an important example. After creation, proteins
fold through a complex chemical process that involves as many as 200 different
amino acid residues. Different folds give the proteins different properties.
In some cases, such as amyloidogenic proteins associated with "Mad Cow
Disease," a wrong fold creates a harmful protein: the scrapie form of prions.
"There is increasing evidence that the wrong folds are due to a kinetic or
dynamic process," Hernandez adds. "We as dynamicists are asking how a sequence
with a given structure goes to that folded structure. By understanding those
dynamics, we would be able to say something about altering the sequence and
preventing it from folding that way."
A supercomputer model accounting for complex changes in reactants and
environment may provide answers that researchers could find in no other way.
Among industrial applications is the way in which thermosetting polymer
materials form. Used in applications such as microchip packaging and
automobile parts, the starting materials include long-chain polymers,
cross-linking agents, fillers and dyes. In manufacturing, the mixture is
heated to liquid form, allowing the polymers to react and chemically link into
a final product whose properties depend on its thermal history – that is, how
quickly it was heated and allowed to cool.
Like protein folding, the polymerization reaction does not occur in
equilibrium because the process of cross-linking significantly affects the
reaction environment. A clearer understanding of the processing dynamics,
obtained through computer models, should give materials scientists better
control in selecting the starting materials.
"You would be able not only to predict what a particular composition is
going to do, but also determine what the initial conditions should be for a
desired outcome," Hernandez says. "That would allow you to design
thermosetting polymeric blends that would be optimal for your particular mold
or system."
Computational Quantum
Chemistry The enediyne family of anti-cancer drugs provides a vital weapon in the
battle against the dreaded disease. The highly reactive chemicals contain two
radicals that "steal" hydrogen atoms from the DNA of cancer cells, triggering
destruction of the cells. Although experiments have shown the basic steps of
the diradical reaction, detailed computer models could give new insights into
how to tune the reactivity by adjusting the chemical structures of the drugs.
Having that information would help scientists produce better anti-cancer
drugs.
These computational studies will require the development of new theoretical
techniques because current models can't accurately describe the highly
reactive diradicals, or more generally, any bond-making or bond-breaking
processes.
"To get a very detailed understanding of the whole process, going from
reactants to products and watching it happen in between, you usually are
breaking chemical bonds," Sherrill explains. "You would like to be able to
describe that theoretically, including the entire reaction path, not just the
beginning and the end."
Sherrill also works with Professor Mostafa El-Sayed and Clemens Burda, both
of Georgia Tech's Laser Dynamics Lab, in better understanding a family of
molecules essential to the vision process. These molecules twist when they
absorb photons of light, a process that has been studied – but not fully
explained – using high-speed laser dynamics.
"Apparently something happens during the course of the twisting that isn't
completely understood," Sherrill explains. "It is hypothesized that some
intermediate structure influences the twisting, but that structure has not
been observed. If we can find it through our modeling, it may apply to the
whole class of molecules."
Beyond its implications for vision research, the work could lead to
improvements in color photography and the capture of solar energy.
In a third major project, Sherrill helps Professor Christoph Fahrni in
designing a molecule that will bind to copper ions and emit light. Measuring
the phosphorescence would then allow scientists to see where copper
concentrates in cells, information important because of the role copper plays
in biochemistry.
Fahrni has begun designing candidate molecules in the laboratory even as
Sherrill models them on his computer. In this collaboration, theory suggests
directions for the experiments to take, and experiments calibrate the theory.
"If the theory gives good predictions, we can go ahead and tweak some of
the parameters theoretically and suggest better alternatives that can then be
made in the lab," Sherrill says. "This could help keep Christoph from having
to make dozens of structures that may not work."
Already, the work with collaborator Professor Alan Gabrielli of Southern
Polytechnic State University in Marietta, Ga., has led to unexpected results
that further the understanding of the phosphorescence of such probe materials.
New Supercomputer Accelerates
Efforts It promises a dramatic increase in the speed at which simulations run,
doing in a few hours what would have taken a week of number crunching on
smaller computer workstations. It also allows researchers to undertake more
complex and accurate simulations that they could not even attempt before.
"This will allow us to do some very high accuracy calculations on some
benchmark molecules that were inaccessible before," Sherrill says. "It will
make a tremendous difference."
Because the reactions Hernandez and Sherrill study are so complex, limits
on computing power have forced them to rely on approximations that do not take
into account all the variables that could influence the outcome. The
supercomputer will alter that trade-off, enabling simulations with fewer
compromises.
By speeding up the simulations, the machine will also change the way in
which the researchers work, allowing them to be more productive and creative
in following up unexpected results.
"If you have a calculation that takes a week for you to complete, you've
got to have a lot of different things going on at once while you're waiting
for the calculations to be done," Sherrill explains. "But if you could get the
results in a couple of hours, you could immediately see what didn't work and
how to change it. You could be a lot more interactive and recover more quickly
from mistakes."
Based on IBM's new copper chip technology, the machine boasts 47 gigabytes
of memory and 764 gigabytes of disk storage. Despite its power, Hernandez
isn't advising researchers to discard their test tubes just yet.
"In order to do everything on a computer, we would have to devise a
theoretical model able to mimic nature completely," he says. "I believe that
we are always going to have surprises from nature."
In common chemical processes, the reactants
make up a small part of the overall environment, which does not change
substantially as the reaction proceeds. In such systems, the environment is
considered to be at equilibrium.
Understanding complex reactions is also important
to Sherrill, an assistant professor whose focus is electronic structure theory
and its application to photochemistry and highly reactive systems. His work
has implications for improving anti-cancer drugs, understanding the process of
vision and tracing the role of copper in the body.
The computational chemistry work of Hernandez and
Sherrill advanced in October 2000 with the installation of a 72-processor IBM
SP supercomputer that forms the core of the new Center for Computational
Molecular Science and Technology. The machine, shared with other researchers
in the College of Sciences and elsewhere at Georgia Tech, is one of the most
powerful academic supercomputers in the Southeast.
For more information, contact Rigoberto
Hernandez, School of Chemistry and Biochemistry, Georgia Tech, Atlanta, GA
30332-0400. (Telephone: 404-894-0594) (E-mail: hernandez@chemistry.gatech.edu);
or David Sherrill, same address. (Telephone: 404-894-4037) (E-mail: david.sherrill@chemistry.gatech.edu)
Contents
Research
Horizons GT
Research News GTRI Georgia Tech
Send questions and comments regarding
these pages to Webmaster@gtri.gatech.edu
Last updated: Feb. 16, 2001