What are Theory, Computation, and Modeling

Simulations act as a bridge in another sense between theory and experiment (see figure B3.3.2. We can test a theory using idealized models, conduct thought experiments , and clarify what we measure in the laboratory. We may also carry out simulations on the computer tliat are difficult or impossible in the laboratory (for example, working at extremes of temperature or pressure). 

The adsorption of protein from single component solutions is qualitatively understood, although a quantitative understanding and models or theories with predictive character are not yet available. If the structure and solution properties of the protein are known and if the solid-buffer interface properties are known, then by careful examination of the surface of the protein (ideally via molecular computer graphics), we can indeed predict what orientation of the protein is preferred on that particular surface. 

What are then the deficiencies in such a procedure Perturbation theory is not really applied as such, the model of point particles is rather poor, the 1/R-expansions contain ill-defined parameters, the energy minimization is affected (in order to reduce the computing costs) by the practical consideration of a distance cut-off threshold in the evaluation of the interaction energy, and the routine characterization of the structure in terms of its dihedral angles may be misleading. Most of these deficiencies are also present in the more sophisticated Monte Carlo and Molecular Dynamics simulations. 

There are still many unsolved problems in computational chemistry and, as discussed, many of them are not new. We still need a method for the quick and accurate calculation of free energies, and we need to understand the limits on the accuracy of any predictive methods. More broadly, what confidence limits can we put on any given prediction from any given level of theory For virtual screening, what are the most important terms left out of the standard docking functions, and indeed higher simulation methods Can we imagine a set of widely accepted well-validated model systems and benchmarks, both for experiment and theory ... 

---