| |
Objets d'Art
One of the most famous photos of 1953, taken by Antony Barrington Brown, shows future Nobel laureates James Watson and Francis Crick discussing some of the finer details of the antiparallel double helical structure of DNA, which they had just described in the journal Nature. Towering over the heads of the two scientists is a massive physical model of DNA that was built using rods, clamps, flat pieces of metal, and other bits of laboratory detritus. Building the monstrous model was integral to the understanding of DNA's structure—as model building often is. "Having a physical model is extremely valuable," says Professor Arthur Olson, who is a member of the Department of Molecular Biology at The Scripps Research Institute (TSRI). "A physical object in a social situation promotes interaction and discussion, and it enables one to see physically the implications of some abstract idea." From the Virtual to the TangibleA crystallographer by training himself, Olson remembers how he built his first protein structure by hand in a "Richards box"—which resembled something that you would expect to find as a professional magician's illusion rather than as a professional scientist's tool. A contraption with a physical model of a molecular structure made out of metal rods and a half silvered mirror that reflected hand-traced images of electron density, a Richards box revealed the molecular structure as a stick model (brass) of all of the protein atoms. Having such a physical model always provided easy access to the biology of molecules, but biologists didn't always have easy access to the physical models. They were hard, often impossible, to build by anyone except the most patient and steady-handed individuals. This all changed a few years later when computers became powerful enough to represent molecules graphically. "Building physical models was tedious," says Olson. "Computers allowed you to do it much better and much faster." So, for a while, physical models went the way of the dinosaur and computers reigned. Software designed to display molecules improved along with the power of the computers themselves, and other features were added, such as virtual reality gloves that allowed researchers to move the image around by "hand" and special head sets that simulated a three-dimensional feel to the flat images. But in the end the images were still flat and displayed on a computer screen. Something was lost in not being able to touch, hold, and manipulate a model physically. Today, technology is turning a corner. It is now possible to fabricate physical (or "tangible") models automatically with the click of a button on a computer. Scientists like Olson are making these tangible models and starting a renaissance that marries these models with the power of computers. Fabricating Objects AutomaticallyOlson uses a whole new class of inkjet printer to make tangible models. The printer basically lays down layer after layer of a special fine plaster powder—eventually filling a well about the size of a gallon of ice cream. The printer builds the model as the layers are laid down. On top of each layer, the printer puts down colored water in a pattern resembling the cross-section of the physical model on that plane. A fresh layer of powder goes on top of that, and the printer lays down the pattern in colored water for the next cross-section. As the thousands of layers are built up, the printer sprays different patterns on each layer, and the water acts as an adhesive and sets the plaster for that layer. In this way, the layers are built up with the colored water binding each layer together. When the process is finished, the volume of powder contains the physical three-dimensional models, which can then be lifted out, dusted off, and finished with wax. Olson uses one of the models regularly when he is discussing his research related to the human immunodefficiency virus (HIV) protease. Olson directs a large NIH-funded project that seeks to establish a drug design cycle aimed at developing, testing, and refining novel approaches to making specific inhibitors of HIV protease that are capable of limiting or eliminating drug resistance. "We're not just looking at a protease," says Olson. "We're looking at the mutational range due to drug resistance and [seeing] what we can predict about it." As part of this work, Olson meets frequently with his co-investigators at TSRI and at other institutions, and he finds that tangible models of HIV proteases help to facilitate discussion. "People see different things in three dimensions than they do on a computer screen," says Olson. "I've had people look at physical models of proteins they have studied for years and say, 'Gee, I never saw that before!'" Tangible InterfacesLast year, Olson was awarded grants from the National Science Foundation and the National Institutes of Health to pursue a project called Tangible Interfaces for Molecular Biology. This project entails building models and using them as interfaces for a computational environment. Rather than simply making such tangible interfaces as a proof-of-principle, Olson has sought collaborations with other investigators at TSRI who can apply such tools to their immediate research. "We're making models for a number of people around Scripps working on various projects," he says. "People really do appreciate these physical models, and they use them in collaborations and in the examination and exploration of structure." This is not a service but a mini-collaboration and an exchange. So far Olson has provided 10 scientists at TSRI with such models. The other scientists give him feedback, providing him with case studies of what the model is good for and what it is not good for. This feedback is valuable for Olson's goal of determining how tangible models can be used and how they can be enhanced. The point of the grant is to determine how people can use physical models to understand complex structures—especially if they are not experts in structural biology themselves. At the moment, Olson is building a "vocabulary" of molecular representations to see what is possible and what is useful to build. He is exploring possibilities such as fabricating physical models and inserting small point magnets into them as a way of simulating hydrogen bonding and other non-covalent interactions. And he is experimenting with attaching pieces of models to wire and other flexible materials as a way of representing flexibility. For instance, he models single DNA strands with tiny magnets inserted into the individual bases, which represent to hydrogen bonds. One such model, which he keeps on his desk, readily forms a double helix with the magnets clicking together as the bases pair. Olson is also trying to further enhance his physical models by using them as computer interface tools with a new technology called augmented reality. Augmented RealityWith augmented reality, it is possible to superimpose computer data and graphics onto the video of the physical model. Then researchers can hold a model in their hands and query the computer for information—asking what a particular residue type is, for instance, or looking at how that residue is conserved across a species. "The goal is to be able to superimpose any kind of annotation—any information—on top of the physical model," Olson says. In his office last week, he demonstrated one application of this, which he is developing for a high school in Seattle. He clipped a tiny video camera with a firewire connection onto his shirt and plugged it into his laptop. Then he turned on the computer and held a model of HIV protease in front of the camera. The solid model was captured by the video camera, and after he adjusted the autofocus and launched the software, the model appeared on his computer screen. This model had a little square marker attached to it that was about the size of a postage stamp and looked like a square bullseye. The bullseye is key to augmented reality because it allows the computer to interpret any given frame of the video image and tell by the shape of the tag what the transformation of the model is based upon the measured distortion of the square marker. Knowing the correct transformation in a frame-by-frame manner allows the computer to track the model in real time as it is moved in front of the camera. In his demonstration of this application, Olson first entered the coordinates for the atoms in the molecule represented by the physical model he led in his hands. Then he asked the computer to display all the histidine side chains in the molecule. There was only one. "OK," he said, "let's display the phenylalanines, too." He clicked a few keys, commenting that he is currently working on integrating voice recognition software with this so that he could give simple voice commands to the computer. When he was done, he held up the protein in front of the camera again, and on the computer screen, the video had added graphics representing the His and Phe side chains. The graphics moved as Olson tumbled the protein in his hands. Then he read in the coordinates for an inhibitor of the HIV protease and he asked the computer to display this. In a few seconds, the computer superimposed the inhibitor on the binding site of the protease. As he turned the model in his hands, the displayed inhibitor turned as well, keeping its correct orientation in the binding site. The technology is still in development, says Olson, and he is the first to admit that it is primitive. The applications are not fully automated, the video is low resolution, and the display is limited to a computer screen or, at best, a video projector. Someday he envisions integrating augmented reality with a headset display so that different people could look at the same object and each see the particular augmented features they wish to see. Still, says Olson "It works well enough so that people get a sense of what we're striving for." He feels that this technology as it exists could be a powerful tool for creating tangible interfaces for molecular biology, and he thinks that the application is only going to get better as technology improves. "You can buy a digital camera today for $100," says Olson. "In 10 years, you might be able to buy an HDTV camera for the same price." Beauty, Truth, Models, and EverythingNext month, Olson is participating in a forum discussion at the 30th International Conference on Computer Graphics and Interactive Techniques, also known as "Siggraph." This meeting brings together animators and graphics specialists who work in such diverse areas as basic science and entertainment. The panel discussion is titled, "Truth Before Beauty: Guiding Principles for Scientific and Medical Visualization," and like the rest of the convention it brings together experts from different areas to discuss a single subject. "[The panel] is dealing with the issues of visual representation in the sciences," says Olson. "They invited me to talk about my work in representing the world you can't see—the molecular world." The molecular world is one that is hard to visualize—even though we see pictures of it on a regular basis. Biology and chemistry have produced thousands and thousands of glimpses of this world in the form of structures, and many of these structures—glossy, full-color, and often quite beautiful—adorn the covers and pages of the top science journals. The point of the panel is to ask whether these representations are true to what the molecular world looks like or if they sometimes sacrifice truth for the sake of beauty. Olson sees this as a bit of a straw man. "[The molecular world] doesn't really look like anything," he says. Since proteins and the other inhabitants of the molecular world are smaller than the wavelength of visible light, they are impossible to see. Structural biologists provide models of molecular structures based on structural data they obtain, and then they usually abstract these models even more—coloring particular residues, displaying only the backbone or particular side chains. However, proteins have no color as such because color does not exist at that scale. In fact any picture of a protein is not entirely accurate because electron distributions and other dynamic features of proteins do not readily translate into static images. But this is just splitting hairs. "All representations of the molecular world are models," he says. "None of them are true per se." The measure of a molecular representation, says Olson, is how well it conveys information. As such, it is often enlightening to show a model that displays less information. If you look at a backbone model of a protein, for instance, you can easily follow the snaking amino acid chain with your eye. Backbone representations have less to do with how the proteins fill space than how they fold up. Beauty, says Olson, is another question. "Truth and beauty are not really along the same axis."
|
| |
