A simulating read

This text straddles the boundary between theory and experiment. Often simulations are described as computer experiments with large software packages being akin to complex pieces of apparatus that are fed with input data by the user rather than with real molecules. The analogy runs deep.

Just as experimentalists will use their machines better if they possess an understanding of their internal workings, so in the same way simulators need an appreciation of the algorithms in the package and of the underlying theory and its approximations.

Martin Field''s book is aimed at the novice user who is likely to be a graduate student or researcher in computational chemistry or biophysics.

The provision of example programs ensures that readers should achieve a reasonable understanding of how simulations are performed and how the programs work.

All the basics are covered, starting with the information that any molecular simulation requires, that of the coordinates of the atoms in the system of interest. Just how these data can be manipulated is fully explained.

The meat, and in many ways the Achilles'' heel, of molecular simulation is the set of potential- energy functions that lead to a calculation of the energy of the system of atoms at a given set of coordinates. Here the book is less critical than it might be, as the approximations are horrendous and the number of disposable parameters to be fitted by experiment is very large.

Nonetheless, defenders of current simulation packages can point to significant successes in refining protein structures and computing energy profiles for enzyme reactions.

Both the main styles of proceeding with simulations are covered. Once one has the energy of a system of atoms, gradients in the energy permit the following of the path on the energy hypersurface as a function of time, so-called molecular dynamics. Alternatively, the atom positions may be altered, energy recalculated and the properties derived from averaging over random positions following the Metropolis Monte Carlo algorithm. In both variants it is possible to compute free energies that are directly related to the quantities measured by experiment.

It is a pity that the latest fashion of combining quantum mechanics with the crude but effective molecular mechanics was deemed too advanced for this introduction, as it is likely to be the chosen method for many applications.

Simulation has been given wider prominence since the announcement by IBM of its $100-million Bluegene project to build a supercomputer to attempt to model the detailed folding dynamics of proteins: predicting structure from sequence. That move of simulation into the big science league will surely attract newcomers, and here they have the ideal starting primer. This book is likely to spend much time sitting next to the mouse by a terminal.

W. Graham Richards is chairman of chemistry, University of Oxford.