Interactive Protein Structure Visualization

Apple: What are we looking at here, and why was this image created?

Dr. Ferrin: It's a three-dimensional model of a virus structure, more specifically mammalian orthoreovirus viron, shown at 7 angstrom resolution. The image was created using Chimera, a scientific visualization tool, and the data was obtained using cryo-electron microscopy.

In this image, we're looking at the structure of the virus so we can see how it's assembled. An animation like this is used for illustrative purposes and reovirus is a useful model for studying virus-cell and virus-host interations. The color-coding is based on the six different types of proteins that make up the virus structure. This animation reveals the structure of the virus layer by layer, how these different proteins are assembled relative to one another, what proteins are in contact with other proteins, and how they make up the layers in the capsid.

Apple: Why is this kind of image valuable to scientists?

Dr. Ferrin: One of the things about viruses is that they are typically comprised of a relatively few number of different proteins, but with multiple copies of these proteins organized with a high degree of symmetry. The reovirus virion shown here is greater than 850 angstroms in diameter and comprises two concentric protein capsids, each having icosahedral symmetry. The protein capsids surround the double-stranded RNA-based genome of the virus. One of the interesting questions about viruses is how they penetrate a cell membrane to begin infection, and in this case the outermost protein on the reovirus virion has been implicated in cell membrane perforation. There are 600 copies of this one protein in each reovirus virion.

Basically, you've got multiple proteins somehow interacting with each other and forming a complex, and you'd like to understand the details of the interactions and how that complex functions within the virus.

Apple: How do scientists create images like this?

Dr. Ferrin: By using computer software to analyze and create models based on experimental data collected with an electron microscope. By collecting the image data associated with several identical virus particles, you can construct a three-dimensional density map representing the proteins comprising the virus.

Normally, you aren't able to see the detailed conformation of the individual proteins and boundaries between them because of the low image resolution. So you'd use computers to do segmentation filtering to find the differences in the density map that represent the boundaries between parts of the structure. You're trying to isolate a particular protein from the background noise data, for example, and you can use various kinds of computational edge-detection algorithms to do that.

Apple: Are there other ways scientists can communicate this same kind of data?

Dr. Ferrin: What's shown here is an animation. But the real strength of Chimera is as an interactive visualization tool. In interactive mode, you use your mouse or some other input device, such as a 3-D joystick, to manipulate the model. It's akin to holding the model in your hand, manipulating it, and viewing it from any desired angle. But more than this, you can control what parts of the model you see, how you see it (that is, how the model is rendered graphically), and much more. By using special stereo glasses, you can even view it all in 3D. And with a stereoscopic video projector and stereo glasses such as we have in my lab at UCSF, a group of people can visualize models like this and together comprehend and discuss the complex spatial relationships that lie within. It's a fantastic way for scientists and students to learn about complex biological structures.

Of course, it goes without saying that the underlying data is numerical, so another obvious way to disseminate the data is via a large spreadsheet or XML-formatted data file. But this method is totally ineffective for anything other than simple data transfer. The old adage about a picture — in this case, an interactive 3-D picture — being worth 100,000 words couldn't be more true in structural biology.

Apple: How do scientists decide how to best represent an image in order to communicate its importance most effectively?

Dr. Ferrin: What's best for answering one scientific question is not necessarily best for another question. For example, if you're interested in the overall organization of proteins within a virus, the animation shown here is quite effective. But if you're interested in why a particular protein within the virus forms trimeric units, then you'll want to focus in on the protein-protein interface, fit the crystal structure of the protein, if available, to the electron density map, and try to determine just what amino acid residues within the protein are responsible for the underlying hydrogen binding. So it's scientific curiosity and inquiry that drives the visualization of the data.

The beauty of an interactive and feature-rich application like Chimera is that you can quickly try different approaches to visualizing data — in this case complex biological structure data.

Apple: Tell us a little bit about Chimera.

Dr. Ferrin: Chimera is an application used to visualize, interactively manipulate, and analyze molecular models — either at the atomic level, or for whole assemblies of proteins — such as you see with this particular reovirus virion.

It can display three-dimensional electron microscope density maps and then interactively adjust the contour levels. So you can filter the data and then color code different parts of it based on different criteria. For instance, you can choose to color-code based on a spherical radial function. That is, you indicate on the image where the center of the sphere is, and then color everything within, say, 50 angstroms of that middle point. Or, as was done in this reovirus animation, you could color a thin layer within the density map that is, say, between 45 and 55 angstroms from that center spot. This approach works particularly well with viruses because of their high degree of symmetry.

Apple: How is Chimera different from earlier popular applications such as UCSF Midas and Midas-Plus?

Dr. Ferrin: Those were earlier-generation applications that were focused on atomic level models built from X-ray crystallography data. So you were looking at individual proteins and small molecules, and their relative orientation to one another in the crystal structure. Used with Chimera, we can now do everything from individual atoms up to super-molecular assemblies of atoms.

Chimera can be used to create these types of models, but also models of much larger molecular assemblies, such as viruses, ribosomes, and chromsomes. Molecular assemblies are of keen scientific interest in biology today, and hence the ability to visualize these complex structures is critical our understanding of their structure and function.

In addition, Chimera was designed to be extensible. That is, it is relatively straightforward, without the need to modify Chimera source code, to add customized functionality. This is important to many cutting-edge research projects, since it’s virtually impossible for an application such as Chimera to provide all the visualization an analysis functions that every research scientist needs. By making Chimera extensible, researchers can add novel functionality specifically based on their project needs.

Lastly, Midas and MidasPlus were originally designed to run on relatively expensive dedicated computer graphics workstations. In contrast, Chimera takes advantage of the powerful yet inexpensive computer graphics hardware now included in virtually every personal computer, including laptops, thus making interactive molecular visualization and analysis accessible to every scientist, educator, and student.

Apple: What do you envision for scientists in Chimera's next iteration?

Dr. Ferrin: More analytical capabilities, meaning additional ways of looking at, filtering, and analyzing data, together with better ways of creating illustrative animations, such as the one we're showing here.

Virus models, such as orthoreovirus, are relatively simple structures because they have a high degree of symmetry, but there are other very important biological complexes that have much less symmetry. We'd like to understand these better, too, and that may very well require an application capable of understanding dynamic processes.

For instance, one of the most important yet least understood is the ribosome. It's composed of a large number of proteins, there's not a lot of symmetry associated with it, and it works dynamically, like a small engine running at an extremely small scale. Without the dynamic understanding, you'd miss the critical molecular interactions. And it's the detail of these interactions that require analysis if you're going to understand how the ribosome takes RNA and builds proteins based on the genetic code found in the RNA.

Apple: What other scientific modeling advances do you see happening in the near future?

Dr. Ferrin: We'll soon be able to get higher-resolution electron microscopic data, data from single-particle reconstruction and electron tomography, and data from high-resolution light microscopy. So that instead of just being able to resolve proteins as simple blobs in a large density map, we'll have a greater ability to see some features on the surface of these proteins and be able to recognize some secondary structural elements such as helices. In general, the resolution of these microscopic approaches is getting better and better.