Twirling ribbons, billowing bubbles: computer visualization brings complex aspects of life into view.

It's no big deal to sketch a water molecule, with its two hydrogen atoms linked to an oxygen, or to picture how those atoms interact with another substance. But when molecules contain hundreds or thousands of atoms, even a brilliant chemist has trouble keeping them all straight, never mind tracking how each atom moves or changes when confronted with other molecules.

Nor does the challenge end with molecules. Researchers, teachers, and physicians want to compile the massive amounts of data acquired through various imaging and analytical technologies in order to make and manipulate clear pictures of cells, organs, even entire organisms.

That's where computers, particularly computer graphics, come in. By incorporating computational and visualization techniques into their experimental repertoire, researchers can make sense of ever more complex data and substances. X-ray crystallographers, for example, have demonstrated that amino acids, the building blocks of proteins, form chains that twist into incredibly convoluted configurations, or motifs. Sometimes these structures are shaped by other, helper molecules. "But from my point of view, that's just the beginning," says Arthur J. Olson of the Scripps Research Institute in La Jolla, Calif.

Like other computational chemists and molecular biologists, Olson has built on those data and, with certain mathematical procedures, has reenacted molecular minglings. "That's really the crucial aspect of biology," adds Michael Colvin of Sandia National Laboratories in Livermore, Calif.

As part of these efforts, researchers have become artists, sketching with a keyboard instead of pastels. Each new line of computer code, each faster chip has brought a little more of the world beyond our vision into view. Sometimes, the scientists ask the computer simply to compile all the data into a comprehensible picture, one based totally, or as much as possible, on existing physical and chemical laws. Other times, researchers tap the computer program's intuition to filter out unimportant data and fill in missing details. "[Visualization] lets you see things that you might not have been able to see before," says Helen M. Berman, an X-ray crystallographer at Rutgers University in New Brunswick, N.J.

Berman remembers the early days of computer graphics, when images came only in black, white, and shades of gray. "I thought, What do we need color for? It's just a luxury," she recalls. Now, she realizes how much better colors are at capturing the personalities of molecules. The bright blues, yellows, and purples help experts as well as novices make sense of what sometimes seems little more than a tangled mess of squiggles or -- in the case of ultrasound data -- a fuzzy image.

Already, an explosion in visualization techniques has revolutionized the depiction and analysis of molecules. But computer graphics experts and their scientist-collaborators have not stopped there. They are also portraying, in ever clearer detail, cell membranes and the development of chick embryos. They are even attempting to make picture-perfect ultrasound images of the human fetus (see sidebar).

Recently, Berman and her colleagues captured on screen the structure of collagen, the one major protein motif left for X-ray crystallographers to decipher. Collagen is a protein in bone and connective tissue that forms from three parallel chains of amino acids that twist counterclockwise into spirals. Those spirals extend down a common axis, they report in the Oct. 7 SCIENCE. The spirals hold together because of a repetitive sequence: Every third amino acid is glycine, which allows the strands to lie close together. Should another amino acid replace the glycine, the collagen might become flexible or unstable and cause disease, Berman says.

Interestingly, a single switch between a glycine and an alanine made crystallization, the first step in this kind of analysis, possible, Berman notes. To obtain a crystal, her group made and dissolved protein fragments that were 30 amino acids long, each with collagen's repetitive sequence -- and each with this one switch. The substitution leads to a slight unraveling, noticeable in an image with amino acids shown as little bubbles (A). When the fragments are portrayed as ribbons (B), the substitution site becomes less evident, but the spiraling of the three chains into a triple helix becomes much clearer, she points out. "You see different things depending on the type of representation." At Yale University, Paul B. Sigler and his colleagues looked at molecules called chaperonins, which help proteins fold into their final shapes. In the Oct. 13 NATURE, he and his colleagues describe the structure of a bacterial chaperonin, GroEL.

[CHART OMITTED]

Picture an iced doughnut divided into seven equal pieces and placed upside down on a similarly iced and divided doughnut. Each of the 14 slices consists of 547 amino acids. The amino acids occupy particular positions in the slices and arrange into three layered sections, or domains. The innermost, icing domains link the doughnuts. Pieces of the middle domains reach across and touch adjacent slices, while the outermost domains form the opening of this central cavity.

Even this description fails to tell the whole story. Imagine now that some hungry youngster had decided to taste each of the 14 bits of doughnut and left a finger hole through the center of each. These portals extend deep into the slice and probably provide the chaperonin with a place for processing energy-transfer molecules called ATP, Sigler notes. Not all the graphics generated by Sigler's group show all these details (C).

[CHART OMITTED]

Based on these structural data, Yale colleague Arthur L. Horwich made specific changes in the amino acids of the GroEL central cavity. The modified GroEL could no longer hold onto proteins, indicating that GroEL corralled proteins in its core, Horwich reported, also in the Oct. 13 NATURE.

Like Sigler and Berman, Olson was schooled as an X-ray crystallographer. Early in his career, he wanted to know the structure of the proteins that surround the genetic material in viruses. Based on the first batch of data he and his colleagues collected, he built a model out of brass parts. However, that model proved useless in his quest to understand how the proteins self-assembled to make the viral coat. "That's when I realized that we really needed more flexible tools," Olson recalls.

Now, Olson spends as much time developing computer software -- his "flexible tools" -- as analyzing molecular interactions. His programs enable him to "see" a molecule in any number of ways (D). "They don't look like anything from the traditional view, because molecules are smaller than the wavelength of [visible] light," Olson explains. "But visualization is a way of taking that structure and representing it in an interpretable way for the questions you are seeking answers to."

[CHART OMITTED]

He can look at the bonds between atoms or ask the computer to draw a space-filling version, whose soap-bubble appearance indicates the volume taken up by the various molecular components. He can highlight the backbone of a protein to get a clearer view of how it folds. Or the computer can pinpoint surfaces exposed to the outside environment or delineate the distribution of electrical charges.

Those images help Olson and others take the next step -- determining how one molecule links to another (see cover). They are also refining software that tackles the docking of one protein with another (E). This second program represents the surfaces of proteins at greater and greater resolutions as it searches for and finds the regions where proteins can bind together, Olson explains. With each iteration, the program rules out connections that prove energetically or structurally untenable and fine-tunes those that seem likely, eventually homing in on the optimal orientations.

[CHART OMITTED]

At Sandia, Colvin, too, investigates molecular connections. For one project, he portrayed a carcinogen attacking DNA (F). "The DNA has grooves, and the carcinogen just drops down into these grooves like a key into a slot," he says.

[CHART OMITTED]

Students, too, are reaping the benefits of these representations of the invisible world. At Purdue University in West Lafayette, Ind., Clark D. Gedney and his colleagues have developed several visualization procedures for teaching biology. One teaching tool incorporates pictures of cross sections of chick embryos taken at different stages of development (G). Gedney's team added colors to highlight the various tissues and had the computer reconstruct the whole embryo. Rather than wield scalpels or fuss with mounting and viewing microscope slides, students click a computer mouse to see embryos at 13 stages of development. They can look at it whole or in cross section, notes Gedney.

Finally, another simulation enables both teachers and students to build their own cell membranes and then examine the effect of adding or removing an electron from the membrane (H). The computer program makes sure the membrane works just as a real one does, Gedney says.

[CHART OMITTED]