Essentials of Computational Chemistry

62 2 MOLECULAR MECHANICS
like X-ray crystallography obviously probes molecular structure in a condensed phase where
crystal packing and dielectric effects may have significant impact on the determined structure
(see, for example, Jacobson et al. 2002).
The above example illustrates some of the caveats in comparing theory to experiment for
a structural datum (see also Allinger, Zhou and Bergsma 1994). Care must also be taken in
assessing energetic data. Force-field calculations typically compute potential energy, whereas
equilibrium distributions of molecules are dictated by free energies (see Chapter 10). Thus,
the force-field energies of two conformers should not necessarily be expected to reproduce
an experimental equilibrium constant between them. The situation can become still more
confused for transition states, since experimental data typically are either activation free
energies or Arrhenius activation energies, neither of which corresponds directly with the
difference in potential energy between a reactant and a TS structure (see Chapter 15). Even
in those cases where the force field makes possible the computation of heats of formation
and the experimental data are available as enthalpies, it must be remembered that the effect
of zero-point vibrational energy is accounted for in an entirely average way when atom-type
reference heats of formation are parameterized, so some caution in comparison is warranted.
Finally, any experimental measurement carries with it some error, and obviously a comparison
between theory and experiment should never be expected to do better than the experimental
error. The various points discussed in this last section are all equally applicable to
comparisons between experiment and QM theories as well, and the careful practitioner would
do well always to bear them in mind.
2.6 Force Fields and Docking
Of particular interest in the field of drug design is the prediction of the strength and specificity
with which a small to medium sized molecule may bind to a biological macromolecule
(Lazaridis 2002; Shoichet et al. 2002). Many drugs function by binding to the active sites
of particular enzymes so strongly that the normal substrates of these enzymes are unable to
displace them and as a result some particular biochemical pathway is stalled.
If we consider a case where the structure of a target enzyme is known, but no structure
complexed with the drug (or the natural substrate) exists, one can imagine using computational
chemistry to evaluate the energy of interaction between the two for various positionings
of the two species. This process is known as ‘docking’. Given the size of the total system
(which includes a biopolymer) and the very large number of possible arrangements of the
drug molecule relative to the enzyme that we may wish to survey, it is clear that speedy
methods like molecular mechanics are likely to prove more useful than others. This becomes
still more true if the goal is to search a database of, say, 100 000 molecules to see if
one can find any that bind still more strongly than the current drug, so as to prospect for
pharmaceuticals of improved efficacy.
One way to make this process somewhat more efficient is to adopt rigid structures for
the various molecules. Thus, one does not attempt to perform geometry optimizations, but
simply puts the molecules in some sort of contact and evaluates their interaction energies.
To that extent, one needs only to evaluate non-bonded terms in the force field, like those