Mechanics, models and Monte Carlo

Molecular modelling promises to become the major investigative technique, or at least guidance procedure, of chemistry. The power of computational exploration is already so great that questions which only a few years ago could be tackled only by real experiments (or, more often than not, merely left unanswered) have now become open to increasingly reliable simulation. The pharmaceutical companies have been in the forefront of the development of these techniques, and in the forefront of pecuniary benefit from them, for hundreds of compounds can be screened quickly and cheaply before a commitment to synthesis and trial is made.

Molecular modelling has already saved the lives of countless rats, and presumably humans will one day benefit too. It has two major branches. One consists of the detailed investigation of the electronic structures of molecules within the framework of quantum mechanics. Although the so-called ab initio techniques, which ideally take as their starting points only the identities of the atoms in the molecule and then allow the Schrodinger equation to grind and churn, are intrinsically the most reliable, such procedures are only applicable to molecules containing a small number of atoms. For the relatively large molecules of interest to most investigators, resort has to be made to one of the increasingly sophisticated semi-empirical techniques that make use of a series of parameters selected to optimise one or other of a set of observables. Judgement is needed to select and assess the outcome of these calculations, and Andrew Leach gives particularly helpful guidance to the newcomer to this field.

The second branch of this vigorously growing tree of techniques consists of molecular mechanics. Here the starting point is an empirical parameterisation of the potential energy of a molecule in terms of its bond angles and the distances between bonded and nonbonded atoms, or even between the atoms of distinct molecules. The outcome of such calculations includes the most probable conformation of molecules, including molecules as large as proteins, the pharmacologically all-important docking procedure for a potential drug and its target protein or nucleic acid, and the energy changes associated with the interactions of a molecule with any water that surrounds it. Beyond that, molecular mechanics merges into the computation of the thermodynamic and dynamic properties of bulk solution, and draws on the statistical sampling techniques (for instance, Monte Carlo methods and their elaborations) and molecular dynamics, where Newton's laws are used to track the evolution of these awesomely complex systems.

Molecular modelling does not end with the computation of properties: it is also a search engine, an engine of elucidation, and an engine to aid imagination. In the first of these roles it enables intelligent (and increasingly artificially intelligent) inquiries to be made of databases, to look for candidates with a sequence of bases or a conformation that holds a particular promise for pharmacological activity of some kind. As an engine of elucidation, molecular modelling helps to establish correlations between structure and activity, and, through the power it gives to the investigator, to track down the contributions to a particular property. Then, as an aid to the imagination, it can identify a promising array of functional groups that should facilitate docking into a target region of a macromolecule, and suggest the actual molecular framework on which to hang them.

Because molecular mechanics and the related techniques of search comprise essentially an empirically based field that draws on parameters suggested by others, and uses statistical sampling procedures and numerical integration algorithms, it is essential to have a reliable and circumspect guide to the choices available and, not least, their limitations and dangers. Leach is surefooted and helpful in this respect and presents an impressive and helpful account of this vast and rapidly expanding field. His book is all the more remarkable in such a computationally directed field for being so readable. He guides us up gentle gradients, and allows us to leap over more precipitous material without losing a sense of direction. All I would hope for to enhance this text in a future edition is a CD providing tutorials for exploring the techniques he so lucidly describes. But Leach has already done the future of chemistry considerable service: I know of no better introduction to this subject.

Peter Atkins is professor of chemistry, University of Oxford.