Imagine this scenario. You sit at your computer to study the structure of a crucial protein--one that you've painstakingly cloned, produced in a protein-expression system, and purified after many hours in the laboratory. The crystallography laboratory that you collaborate with produced crystals of your protein weeks ago. They collected X-ray diffraction data and plugged it into their workstation-based molecular-modeling system, coming up with a three-dimensional structure of your protein showing where the component amino acids were located and how they associated with a metallic cofactor at the protein's center.
On your computer screen, you rotate the structure that they sent to you on a disk, viewing it from several angles. You consider how it might interact with other molecules. After you have viewed the model in several different ways, a potential binding site suddenly becomes apparent. A possible mechanism for the protein's function pops into your head, and you can't wait to get back to the lab to start the experiments that will test your hypothesis.
Sound familiar? The scientist in this scenario could be a molecular biologist trying to identify the active site of an enzyme, a cell biologist studying how a particular gene product influences development, or a biomedical researcher designing a drug that inhibits a disease mechanism. But what makes the situation unusual is that this researcher is working poolside, sipping an icy drink, with a molecular model of the protein displayed on the screen of a laptop computer.
It has been less than a decade since the first researchers in the fledgling field of structural biology created three- dimensional images of macromolecules on a computer screen. Molecular-modeling systems have gone from cumbersome programs requiring the computing power of a mainframe, to the client/server environment of workstations, to the user- friendly interfaces currently available for desktop and even laptop computers. These systems allow bench-level scientists the opportunity to observe interesting molecules in three dimensions.
As is true in most areas of computing, manufacturers of molecular-modeling systems are striving to reach a common goal. "Everyone is looking at ways to bring more functionality to the bench scientist," explains Michael Sullivan, marketing communications specialist at Tripos Associates Inc. of St. Louis. "With today's microcomputer-based systems, we're finding that we can do more of that. By the same token, we now have more data than ever before, and so we need the higher-end systems to sift through all that data before we can bring it to the benchtop."
"The trend right now in computing is toward smaller and faster," contends Jim Grasso, technical specialist for CambridgeSoft Corp. (formerly Cambridge Scientific Computing) in Cambridge, Mass. "Researchers want to get their modeling data processed fast, and on a desktop system, some applications that require complex calculations--like a protein/ligand interaction, for example--might seem sluggish. On a workstation, which has multi-megabytes of RAM, the application will run much faster."
However, the gap that separates the desktop modeling systems and the workstations in terms of speed and number-crunching capabilities seems to be narrowing. "In reality, the difference between a workstation and the desktop machines is quite blurred," says Herman Zinnen, senior applications scientist with Oxford Molecular Group's CAChe Scientific of Beaverton, Ore. "The newer machines are sufficiently powerful to do visualizations, molecular modeling, and molecular orbital calculations, as well as QSAR [quantitative structure activity relationships] and QSPR [quantitative structure property relationships]."
Computer-aided molecular-design (CAMD) systems range in complexity from those that can draw two- and three-dimensional stick figures of molecules--useful for enhancing grant proposals and scientific publications--to higher-end molecular-simulation packages that run on the more powerful workstations. Desktop drawing packages, such as ChemDraw from Cambridge Scientific and ISIS/Draw, part of the multicomponent ISIS package from MDL Information Systems Inc. in San Leandro, Calif., run on both Macintosh and PC platforms.
These and several similar packages from companies such as SoftShell International Ltd. of Grand Junction, Colo. (ChemIntosh and ChemWindow) and Tripos Associates (ChemPrint), are considered the standards in the industry--those in- corporating the most necessary features. Authors trying to explain complicated chemical mechanisms at the molecular level--such as how a protein binds to DNA, or how enzymes catalyze a particular reaction--can draw a structural model and import it into their manuscripts, to better illustrate how the molecules react. In journal articles, "it is helpful to see a visual representation of a structure," states Richard P. Cunningham, a biochemist at the State University of New York, Albany's Center for Biochemistry and Biophysics, who performs structure/function studies on DNA repair enzymes. "For example, graphic representations of the results of DNA footprinting experiments make it much easier to understand how the protein actually binds to the DNA."
Many of the drawing packages are being upgraded this year to include analytical tools in addition to structure-drawing capabilities. The newest version of ChemDraw (ChemDraw 3.5), expected to go on the market in May or June, will include features such as calculations for molecular formula and molecular weight, Lewis structures, and a chemical syntax checker. Like the spelling check in a word-processing package, a chemical syntax check will determine if the structure is chemically correct. "If you mistakenly draw a structure with a five-bonded carbon, for example, the program will give you an error signal," Grasso of CambridgeSoft notes.
At the heart of molecular modeling are systems that allow researchers to re-create proteins and inorganic chemical structures from data derived from X-ray crystallography or from nuclear magnetic resonance (NMR) spectroscopy. X-ray crystallography is a technique dating to 1913, when Lawrence Bragg of Cambridge, England, first realized that X-rays beamed at crystals cause resident electrons to oscillate and give off secondary X-rays. These waves travel in all directions, producing a distinct diffraction pattern on photographic film placed a fixed distance behind the crystal. To a casual viewer, this pattern appears as a number of dots situated in concentric rings, hardly indicative of a molecular structure.
Crystallographers versed in interpreting these dots use the X-ray diffraction pattern to map areas of electron density, creating a three-dimensional picture of the molecules in the crystal. In 1953, long before computerized molecular modeling was even conceived, James D. Watson and Francis Crick used X- ray diffraction data supplied by fellow researchers Rosalyn Franklin and Maurice Wilkins to decipher the three-dimensional structure of DNA. In the 1970s, crystallographer T. Alwyn Jones at the Max Planck Institute in Germany hooked a graphics terminal to a Siemens computer mainframe and created a program called FRODO, which enabled him to construct three-dimensional re-creations of molecules from electron-density data on a computer screen (T.A. Jones, Methods in Enzymology, 115:157- 71, 1985).
After FRODO, programs began appearing all over the world, each one doing the job a little differently. These programs were exclusively for mainframes at first, and later downsized to smaller machines as the computational power increased.
Molecular-modeling programs, initially shared freely throughout the scientific community, are now commercially available at both the workstation and desktop levels.
Workstation-based systems are heralded for their speed, but carry a hefty price tag; they can run in the tens of thousands of dollars. The less costly desktop systems, with prices ranging in the thousands, provide users with convenience and flexibility. "Many people prefer a desktop machine simply for its ease of use," observes Zinnen of CAChe Scientific. "An additional advantage is that, while these machines are now powerful enough to do all of the chemistry calculations, one can just as easily do spreadsheets and word processing on the same machine." Desktop packages come with user-friendly interfaces that allow mainstream researchers to perform sophisticated molecular-modeling tasks without having to learn programming or graphics first.
Desk- or benchtop systems for molecular modeling and design include Alchemy III for MacIntosh and PC platforms from Tripos Associates, which also markets SYBYL, a popular workstation system; Chem3D from CambridgeSoft, currently limited to Macintosh platforms (a Windows version is forthcoming); and HyperChem, developed by Hypercube of Waterloo, Ontario, and sold through distributor reseller channels. Using these packages, scientists can reconstruct even complex biological molecules, such as protein or DNA, from atomic coordinates derived from crystallography data. They also enable one to perform complicated modeling tasks, including energy minimizations and molecular dynamics, electrostatic potentials, and other chemical and physical calculations based on the structure of the molecule.
"The energy-minimization calculations arrange the structure in the lowest energy configuration according to either parameters set by the computer or your own experimental parameters, while a molecular-dynamics run creates a series of different energy conformations," explains CambridgeSoft's Grasso. "These features provide a global minimization, which is key in molecular modeling."
In addition to the analytical features, desktop systems allow researchers to view molecular structures in 3-D and manipulate them--rotate them, or change the bonding information in a molecule, for instance. And they can produce stunning pictures of molecules in all of their three-dimensional glory.
"Viewing structures in three dimensions makes it easier to identify potential binding sites on the enzyme," comments SUNY-Albany's Cunningham.
"In three dimensions, it is easier to identify [amino acid] residues involved in the enzyme's catalytic mechanism. These residues then become targets for site-directed mutagenesis, where we change specific amino acids and look at how the change affects function."
Software links between various packages keep files generated in one program compatible with other programs. Molecular structures can be passed, for example, between Tripos' ChemPrint and Alchemy, so the molecules created with the drawing software can be modeled and analyzed. In many cases, file formats are compatible with popular applications like MOPAC and ZINDO, which are widely used quantum-mechanics programs. Software links are also common with databases--such as Brookhaven National Laboratory's Brookhaven Protein Database or the Cambridge Structural Database, a reference set of crystal structures from the Cambridge Crystallographic Data Center in Cambridge, England--along with numerous other computerized chemical storehouses.
"File compatibility means that you can perform energy minimizations with Chem3D on a structure retrieved from a data bank, and then transfer the file to MOPAC and perform other calculations," says CambridgeSoft's Grasso.
File-conversion software, such as Mol2Mol from Cherwell Scientific Publishing Inc. of Oxford, England, and Palo Alto, Calif., converts files between file formats used in several popular PC modeling systems. With this type of software, researchers can convert files from one format into their preferred package. Mol2Mol additionally features graphic display capabilities, allowing users to inspect molecular structures created in other software applications, manipulate the molecules if desired, and calculate geometric data.
Currently the limitations of working with desktop systems are in the area of large molecules and the speed of the analytical functions. "PC and Mac systems will perform energy minimizations and molecular dynamics, and that's a good start," comments Sullivan of Tripos Associates. "However, when we talk about extensive protein work or de novo design, the higher-end packages are more applicable, simply because there is more computer power on the higher end." Tripos' UNIX-based SYBYL package consists of several modules that are designed to address these areas of CAMD.
The ISIS package from MDL Information Systems is a multi- component system that also offers applications like structure drawing and three-dimensional modeling as part of a single interface. Through this interface, users can store, manage, and communicate scientific data, including databases.
Computer-aided drug design (CADD) is one area of structural biology requiring all of the molecular-design industry's acquired power and knowledge. There has been an explosion of interest in the field as pharmaceutical companies reach for the elusive brass ring--a chemical compound effective against some aspect of a disease state.
One approach to CADD is to sift through data banks filled with chemical structures accumulated over years of research, looking for likely drug candidates. The goal of many of these studies is to identify a key structure called a pharmacophore hypothesis, or biophore, which may be a functional group or a set of atoms. The biophore is the biological target, usually present in active molecules but absent in inactive ones. Once identified, the biophore can be modeled and imported into a three-dimensional database-searching program such as MDL's MACCS-II, which runs on a mainframe. The program scans the chemical inventory for a pharmacophore, a structure that matches and perhaps might inhibit the biophore. "Looking at the problem in 3-D is definitely the best way to determine what type of chemical structure would be the best," maintains Grasso of CambridgeSoft.
"Three-dimensional database searching is an integral part of computer-aided drug design efforts," says Osman Gner, senior scientist with MDL. In a classic example, 3-D database searching and molecular-modeling tools were used together to identify an inhibitor of HIV protease, an enzyme crucial to the replication of the HIV virus. Using a combination of molecular modeling, 3-D database searching, X-ray crystallography, and synthetic chemistry, a research team headed by Patrick Lam from DuPont Merck Pharmaceutical Co. in Wilmington, Del., designed a chemical that fit into a receptor site on the HIV protease molecule, which, in vitro, effectively inhibited its binding.
"This is an excellent example of how the collaborative use of computer-aided drug design by scientists in different disciplines--crystallographers, modelers, and medicinal chemists--resulted in the development of an inhibitory chemical," states Guener. He adds that these types of studies are finding their way more frequently into the scientific literature. Collaborative efforts and communication seem to be key in this rapidly expanding field.
Holly Ahern is a science writer and an assistant professor of biology at Adirondack Community College in Queensbury, N.Y.