Printer Friendly

Shaking and baking to atomic positions.

An expert crystallographer can take literally years to work out, largely by trial and error, the positions of atoms in a given molecule. An innovative, computer-intensive method of extracting information directly from X-ray diffraction data now offers the possibility of cutting that time to hours.

In a recent test of this new method, researchers needed only a few hours of computer time to provide the information needed to find the positions of 104 atoms in pairs of molecules of a compound related to the immune-suppressing drug cyclosporin. Using more conventional methods, Russian scientists had spent nearly a decade trying to determine its structure - without any success.

The researchers who developed the technique, led by Herbert A. Hauptman of the Medical Foundation of Buffalo and the State University of New York at Buffalo, announced their achievement at last week's American Crystallographic Association meeting, held in Pittsburgh. This success marked the first time the researchers had applied their technique to a large molecule with a previously unknown structure.

The technique has the potential to dramatically accelerate the designing of drugs for specific purposes, a process that relies on knowledge of chemical structures of molecules, Hauptman says.

In X-ray crystallography researchers bombard a single crystal of a given substance with X-rays of a certain wavelength. The orderly rows of atoms within the crystal deflect these X-rays in particular directions to produce a distinctive pattern of spots on a photographic plate.

The positions and intensities of these spots, which could number in the thousands, provide information about the locations of atoms within molecules of the given substance. But that isn't enough to draw a complete three-dimensional portrait of the unknown molecule. Because they don't know the times when X-rays arrived at each spot, crystallographers generally lack information about the so-called phase relationships of the diffracted X-rays. And because they need both intensity and phase data to find a molecular structure, experts must often rely on informed intuition to aid in unraveling a molecule's structure.

Several years ago, Hauptman proposed a mathematical formula that he claimed could be used for zeroing in on the missing phase information. This complicated formula exploits subtle relationships between phases and measured diffraction intensities. If one could minimize the value of this complicated expression containing thousands of variables, one could solve the phase problem.

To make the scheme workable, computer scientist Russ Miller of SUNY-Buffalo and his co-workers developed algorithms specially designed to run efficiently on computers consisting of large numbers of parallel processors, which operate simultaneously while working on different pieces of a problem.

The result was a "shake-and-bake" strategy, which starts with "trial" molecules consisting of random arrangements of atoms, subject only to chemical constraints. "You essentially toss the atoms in a shoebox with enough chemical knowledge to make sure the arrangement makes chemical sense," Miller says.

The computer calculates the phases associated with these atomic positions. Then it randomly perturbs the resulting phases in such a way as to provide a slightly lower value for Hauptman's formula, and computes the new positions of the atoms. The atoms, in turn, move slightly, leading to another calculation of the phases, and so on. The computer may go through as many as 200 such phase-position cycles for each initial arrangement of atoms.

In April, the researchers used this technique to solve the structure of the antibiotic gramicidin-A, which contains 317 atoms, excluding hydrogen. They accomplished in a matter of weeks what had originally taken David Langs, a senior research scientist at the Medical Foundation of Buffalo, 10 years to work out using other methods.

Last month, having a few days of free time between projects, Miller came to Langs asking for crystallographic data on which to test his newly refined version of the structure-determination algorithm. Langs, highly skeptical that anything would come of it, suggested finding the structure of a molecule that had long stymied Russian scientists. Over two years, Langs himself had tried half a million possibilities with no success.

"I put the data into the program before I went to bed at night, and the next morning I looked through the results," Miller says. One of the 64 random arrangements of atoms with which he had started gave a value for Hauptman's formula that was significantly lower than the others, indicating the right answer.

This particular trial result provided sufficient information for Langs to work out the details of the molecule's structure in only a few hours . Out of the millions of possible atomic arrangements, Miller's algorithm had converged on the correct structure with remarkable efficiency It was like looking for a needle in a haystack with the advantage of having a magnet on hand.

Subsequent tests involving 640 random atomic arrangements produced two additional results that converged to the same, correct molecular structure. "I had been lucky" Miller says. "The first one happened to come out in the first batch of 64." Each set of 64 trials took about 1.5 hours on a multiprocessor computer known as the CM-5 Connection Machine.

After this initial success, Langs provided Miller with a second unknown structure related to the first but deemed more difficult to solve. "We gave it a shot, and the answer emerged just as rapidly," Miller says.

"It was quite remarkable to us. It made believers out of a lot of people," says Jane Griffin, head of molecular biophysics at the Medical Foundation of Buffalo. "But we still have a lot of research to do in extending [the method], seeing how high the resolution of the data has to be, and solving a variety of other problems."
COPYRIGHT 1992 Science Service, Inc.
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 1992, Gale Group. All rights reserved. Gale Group is a Thomson Corporation Company.

Article Details
Printer friendly Cite/link Email Feedback
Title Annotation:data research time lessened on position of atoms in molecules
Author:Peterson, Ivars
Publication:Science News
Date:Aug 22, 1992
Previous Article:Turtle recovery could take many decades.
Next Article:Laser process shapes microscopic parts.

Related Articles
A periodic table for molecules.
Making 'movies' of biological molecules.
Getting a feel for atoms: 'magic wrist' takes scientists into a new sensory realm.
Scaling chemistry's peaks.
Pinpoint splitting of molecules.
Boron molecules that look like fullerenes.
Atomic rebound makes breaking up hard.
Ultracold molecules form inside superatom.
Writing with warm atoms.
Molecular anatomy revealed: CT scans for molecules show the gauzy nature of matter.

Terms of use | Copyright © 2017 Farlex, Inc. | Feedback | For webmasters