Molecular Modeling | Research & Encyclopedia Articles

Molecular Modeling

For much of the history of chemistry, the proof of any chemical reaction was in the doing. The bench top was the proving ground. The blackboard or the back of the envelope might serve as the site for working out a reaction but actually making the compound was the true test. This is still true today but in the past fifty years, the availability of computers and computation software, along with a deeper understanding of the theoretical underpinnings of chemistry, have made the job of the synthetic chemistry much easier. Testing out products, reagents, and reactions using minutes or hours of computer time instead of days or months at the bench has greatly increased productivity and design in research.

Computers have enabled the facile storage and retrieval of structures and other chemical information. The interesting mathematical problems posed by the new quantum mechanics required massive amounts of computational power. Simple calculations could require years when done with pad and pencil. Indeed, many dissertations arose from solving just one aspect of a much larger problem.

Solving quantum mechanical calculations wasn't the only use for computers.Crystallography, with its massive matrices of interdependent variables and large data sets of observations, was also begging for some form of automation of calculations. In essence, the computer freed the theoretical chemist and crystallographer to think about the process instead of doing the arithmatic. Of course, this is a sweeping statement that neglects the work required to set up and debug the computer programs but the chemistry, like most of the physical sciences, was definitely ready for computers to come along. To this day, some of the largest users of computational time and power on any campus in the country are chemists.

As computers have evolved, so to have the uses to which they have been applied. With the relatively slow and simple machines of the early 1960s, the level of sophistication for any program was severely limited. A measure of the effect of computers on chemistry can be observed in the number of crystal structures accomplished and submitted to the Cambridge Crystallographic database. In the early 1960s, this number was in the hundreds. By the 1980s, in the thousands. By the mid-1990s, the number of structures is in the hundreds of thousands and beyond the capacity of the database to track. That this growth in structures parallels the growth in computer power is an indication of the effect that computers have had on chemistry. And it is not just in shear numbers but the level of complexity of the structures that has increased. Where in the mid-1980s, protein crystal structures were still a rarity--the 1988 Nobel Prize in Chemistry was awarded to Johann Deisenhofer, Robert Huber, and Hartmut Michel for the structural elucidation of a bacterial photosynthetic reaction center. By the mid-1990s, several hundred structures were registered with and submitted to the Brookhaven Protein Data Bank.

Crystallographic structures are, in one sense, a model of a chemical compound. They are electron density maps that provide a minimum in the least squares fit of the theoretically calculated structure factor with the experimentally observed data. That is, they are a representation of the molecule based on X-ray or neutron diffraction data. However, they are only one type of model. Expanding the number and accuracy of the models in chemistry has been an underlying driving force for theoretical chemistry over the last two decades of the 20th century.

In exploring the chemical world, chemists have always come up against a barrier. Atoms are tiny. They are far too tiny to see by ordinary methods (hence, the need for X-ray Diffraction Crystallography). This presents a number of problems. For example, how does one visualize whether or not a particular neurotransmitter will fit into a particular receptor? How does one know which conformation of a simple alkane will be the most thermodynamically stable in a non-polar solvent? How about in a polar solvent? The best solution is difficult to come by through intuition or guess work. Computational chemistry has eliminated some of the guesswork.

In dealing with molecules, two basic approaches have been used. From a historical perspective, the first approach was to model compounds using pure quantum mechanical calculations. That is, the orbitals of the molecules were composed by a combination of atomic orbitals with increasingly sophisticated levels of complexity and mathematical accuracy. The size of the basis set often determined the accuracy of the results and the agreement with experimental values. The quantum or ab initio approach to calculating molecular properties and features is the "best." Unfortunately, it is also the most time consuming. Historically, even simple molecules, with only a few atoms, would take days to calculate. The advent of high speed super computers and parallel processing has decreased the length of time involved in a calculation but this approach is still not "fast." Certainly, computer programs running these algorithms are not interactive.

The difficulties with the mathematically rigorous approach of ab initio spawned a number of simpler solutions. In essence, these approaches all have in common the fact that they do not try to calculate the real parameters for the electrons in each atom of a molecule but use observable variables to model molecular behavior. Molecular mechanics, for example, models bond lengths and angles based upon a minimization of molecular energy as determined using the stretching force constants for bonds (available from infrared spectroscopy), torsional energies (which are also obtained from spectroscopy), and idealized bond lengths (from crystallographically-determined structures). The arrangement of atoms in such a molecule is the best arrangement for the parameters available but ultimately, the quality of the structure depends on the quality of the parameters. In the early history of molecular mechanics, there were some difficulties but as the techniques have been tried, tested, and refined, they now provide answers with a greater certainty or that have a higher degree of confidence. More importantly, they do this in an interactive fashion which allows the researcher to visualize the results and refine the model.

This ability to "visualize" chemical compounds and their interactions has been critical to the development of a number of areas of chemistry and biochemistry but probably the most important is in the area of drug design. The ability to "dock" molecules into active sites for enzymes allows researchers to, in the first instance, detect the active site by using the naturally active compound and, secondly, to design synthetic molecules with the same docking properties but that are not chemically active. That is, one approach to turning off an enzyme is to block its active site with a non-reactive compound. This is a bit like sticking the wrong key in a car's ignition and then getting the key stuck. It prevents the car from working.

Interactive docking programs are one of the ways that molecular modeling is helping to shape industrial processes. Other approaches, though, have included the modeling of molecules on surfaces. For example, a metal catalyst needs to bind a molecule prior to reaction. The mode of binding, the orientation with respect to the surface, and the energies involved all affect the catalytic properties of the metal and the rate of conversion. Modeling this form of interaction using a computer simulation is one of the ways that computational chemistry is bridging the gap with industry. Matching computer generated predictions with industrial results is also providing some assurance for the accuracy of the models which, in turn, allows for more confidence in their predictions. But it also facilitates an understanding of what the chemistry means. That is, why certain reactions proceed and others don't or why some molecules never achieve a 1:1 coverage of a surface.

An example is the model of oxygen dissociation on a rhodium surface developed by Andrew Rappe of the University of Pennsylvania. Intuitively, complete coverage might be expected but experimentally only a 50% coverage is observed. Modeling the process indicates that repulsive force between the dissociating molecule and neighbouring oxygen increases the energy of the transition state limiting the possible coverage.

Molecular modeling has advanced and continues to advance to a level where extremely large systems can now be studied. By adopting the appropriate model, even systems with millions of atoms can be rendered computationally tractable. For example, investigating pore creation in the surface of silicates or testing the lubricating properties of wear inhibitors are expanding both an understanding of molecules and the computational programs required. It is the synergy between the computer programs and the type of problems which need to be addressed that drives forward much of the work in computational chemistry and molecular modeling.

Also of interest is the synergy between the problems addressed by chemists and those in other disciplines. For example, the same algorithms that allow protein chemists to model complex protein structures provide the three dimensional computer simulated characters used in movies. The force field interactions provide the basis for "bumping" into objects in computer games. A visual form of hypertext mark-up language (HTML) is being developed for chemists but will see applications in home shopping. The progress and level of computer programming exceeds just the needs of computational chemistry. Arguably, this is one area of chemistry that is leaving its mark as it develops.

However, molecular modeling is developing and keeping pace with the latest developments in computers. Chemists have no end of problems that certainly can be visualized on a computer screen. The number of algorithms and the variety of programs available for solving molecular mechanics or ab initio calculations is growing. Sorting out drug interactions, catalyst design, thermodynamic properties, or biochemical pathways are all within the power of molecular modeling techniques. And, although they are computer generated and not real atoms, it does give the chemist a chance to examine chemistry at the level of the atom. After all, we will never be able to see it for ourselves.

This is the complete article, containing 1,631 words (approx. 5 pages at 300 words per page).