Date=24.03.1991; Publication=Columbia_Missourian; Page=00; Book=zz;

IDEAS

Computational Chemistry

by David Holman

SYBYL is bright, sexy, sophisticated and fascinating. She joined the M.U. chemistry department last fall, and the professors and graduate students are already competing for space in her appointment book. SYBYL is a remarkable computer program that runs on the department's new Silicon Graphics IRIS-4D computers. The combination of hardware and software makes a powerful, state-of-the-art molecular modeling system. Such a system enables researchers to construct three-dimensional models of very complex molecules in the computer and to look at them on the monitor screen from any desired angle. This ability is especially important in biological chemistry because the biological activity of very large molecules, such as enzymes and proteins, depends upon both their chemical composition and their shape. The Silicon Graphics system looks about like any personal computer that you might see in any of thousands of business offices. There is a keyboard, a mouse, a large-screen monitor. But instead of a single central processing unit, the M.U. setup uses two computers, which the chemists have named Castor and Pollux.

On a rainy day in September, graduate student Chris Horan has a date with SYBYL. Horan is just learning the program. He uses the mouse to draw a molecule on the screen. He doesn't have to draw each individual atom because SYBYL knows the structure of common atom groups that can be assembled into more complex molecules. A benzene ring or an amino acid can be added with a click of the mouse button. Horan's first drawing shows dots for atoms and lines for the bonds between them, just like the drawings in your basic chemistry textbook except the computer shows each element in a different color. He selects the "optimize" function from SYBYL's menu, and his molecule suddenly changes shape. SYBYL has computed the most stable conformation for the chain of atoms that Horan constructed. Another click of the mouse button and the drawing turns into a 3-D stick-and-ball model. Click again and the actual surfaces appear. The model now looks like a twisted mass of brightly colored fish eggs, stuck together and floating in space. Horan moves the mouse across his drawing pad and the molecule on the screen begins to rotate so he can see it from a different angle.

It is also possible to put two molecules at a time on the screen to see how they will "dock" or fit together during a chemical reaction. This is one of SYBYL's most attractive features, but Horan hasn't figured it out yet. He spends about half an hour experimenting with the program, demonstrating different display modes, watching the basic components of living things, a million times larger than life, twisting and turning in computer space. He leaves the room smiling.

"It's hard to believe you're actually working when you're having this much fun," Horan says. But molecular modeling is more than a game of electronic Tinker Toys. It is only one aspect of a growing area of study known as computational chemistry, and it's serious business today.

"It's not only theoretically interesting now. It is commercially of use," says assistant professor Rainer Glaser, a computational chemist and Horan's faculty mentor. "You will rarely find any of the major companies in America that do not have a theoretical chemistry group with molecular modeling."

The main applications for this technology are in pharmaceuticals, agricultural chemistry and diet foods. Research in these areas involves creating thousands of compounds and testing them for their biological activity to see if they do produce the desired effect in an organism. It would be nice if a chemist could predict a compound's biological activity before synthesizing it, rather than testing for it after the fact, but this is hard to do.

Life is a continual remodeling process controlled by programs encoded in an organism's genes. All living things are complicated chemical factories that are constantly building large molecules in some locations and taking them apart in others--digesting food, building new tissue, fighting infection, removing waste products and sending messages back and forth to keep the factory running smoothly. Many of the chemical reactions that occur in living cells are turned on and off by a sort of lock and key mechanism. Large protein molecules called receptors, found on the surface of special cells, act as the lock. Each receptor is twisted into a unique shape, a sort of keyhole that protects a reactive site within the receptor. The lock can only be opened, and the reaction turned on, by a messenger molecule with the correct shape to fit into the protein keyhole and thus reach the reactive site in the receptor.

The search for new drugs, pesticides, and other biologically active compounds is often a game of chemical lock picking. Researchers look for compounds that will either mimic an organism's natural keys and turn on a reaction or block the key hole to prevent the organism's own keys from working. A successful compound should have the desired biological activity without affecting other reactions and causing undesirable side effects. Trial and error, the traditional research method, is extremely expensive and time-consuming. It's like shooting craps, but the odds are worse. Molecular modeling can improve the odds by letting the researcher examine the shape of the key before it is actually made. A theoretical chemist can test an idea for a new compound with a computer modeling program and, in a day or two, have a fair idea of the compound's biological activity. To synthesize and field test the same compound might take a team of chemists a year or more.

Molecular modeling is a powerful tool, and several programs are now available, but it does have its limitations.

"If it did exactly what it purports to do, we could just have one chemist synthesizing only those compounds that are suggested by the program," says Grant Dubois, director of a new products research team at NutraSweet Co. "But in reality, things don't work that way."

Dubois says his company employs two full-time computational chemists and a molecular design team. They have been using molecular modeling programs for several years, but Dubois is not putting all his eggs in the modeling basket. "We use a variety of methodologies," he says. He apparently hasn't forgotten that aspartame, the sweetener that made NutraSweet Co. possible, was discovered by pharmaceutical researchers who were looking for something entirely different at the time. A slavish adherence to models, or any other regimen, doesn't leave much room for serendipity. And sometimes the models are wrong, especially in taste perception, where the nature of the receptors is not yet well understood.

There is nothing inherently wrong with molecular modeling. As with any computer model of any complex system, its weakness comes from the same source as its strength--the assumptions on which the model is based.

Seated in a squeaky wooden swivel chair, which seems oddly anachronistic in his computer-filled office, Rainer Glaser gets to the root of the problem. Getting down to basics is his specialty. He computes the energies and structures of molecules ab initio, from the beginning, using the laws of quantum mechanics.

"If you do calculations ab initio, you can be very certain that you have a model that's very close to the real thing, whether you are doing a known molecule or an unknown molecule," he says. But very large molecules, such as proteins, which may contain hundreds of atoms, cannot be computed using quantum chemistry. The calculations are simply too overwhelming, even for a big computer.

"In the stuff we are doing here, sometimes we have little molecules that have two nitrogens, an oxygen, a carbon and a couple of hydrogens," Glaser continues, "and it runs for six days on our computer. Just for one computation. You cannot compute these very large molecules. Even if the computers grow in their speed, as they have in the past few years, we will not be able to do that. Ever."

But the rigorous methods of quantum chemistry can be used to attack the large molecules one small piece at a time. For example, a hydroxyl group (one hydrogen and one oxygen atom) bonds to a larger group of atoms in some characteristic fashion. The computational chemist can calculate a standard length for that bond and a characteristic energy curve. The modeler now has a set of parameters which help to predict how the hydroxyl group will behave in other molecules.

"With molecular modeling we are not doing real quantum chemistry any more," Glaser says. "We think of each bond between atoms as a little spring with a certain force constant. We think of each angle in the formation as having another force constant, just like a little spring. These force constants we take from the ab initio computations and from empirical observations, and we just fit them in the molecular model so they reproduce what we already know. Then we assume that we can use that to predict things that we don't know yet." But parameters derived from known molecules may not always apply precisely to unknown ones. So there is a cycle of constant improvement that includes the theoretical chemist, the computer program writer and the synthetic chemist. Each provides grist for another's mill, and gradually the parameters are refined and the models become more accurate.

The field of computational chemistry is very young. It is concerned with calculating the properties of molecules from principles of quantum mechanics and then applying that knowledge to problems that arise in other areas of chemistry. The basic theories of quantum chemistry were worked out in the 1930s, but they could not be applied to practical problems until recently, when computers became much more powerful and much less expensive. Without computers the chemist could not do the millions of computations required, and even if the calculations could be done, nobody else could understand the results.

"All our results come in numbers," Glaser says, pulling a thick loose-leaf binder from his desk. "This is what we get. Twenty-five, thirty pages of numbers. Nothing else. This doesn't tell anybody anything unless he is an expert. So we have to go into producing graphics--surface plots, line plots, contour plots and the likes. This is where the scientific work really starts, if you look at these calculated results in the form of pictures."

Glaser leads a gallery tour of the graphic art that covers his laboratory walls, talking about electron density and energy and gradient vector field lines. A chemist may see these things. Others will see strange spider webs, old lace, exotic flowers or butterflies, landscapes of alien planets.

Computers and graphics programs like SYBYL have literally opened new horizons for chemists--and scientists in every field. They are bringing scientists exciting snapshots from unseen worlds, and travel pictures seem to stimulate imagination and the desire for further travel.

While Chris Horan looks for more examples of molecular art, Glaser stands with one hand on his new VAX 3100 computer--1.2 gigabytes of memory in a box the size of a carry-on suitcase, computing power that would have filled his entire lab ten years ago. He's standing there and he's bitching, in a cheerful sort of way.

"Greed comes to every computational chemist," he says. "We never have enough computer time. We always want more. We are still doing relatively small molecular systems. There are so many areas of chemistry that we can hardly touch because the computers just aren't fast enough yet."