In this review, we have discussed a range of applications of molecular simulation methods to problems of peptide conformational dynamics and thermodynamics, along with a brief overview of the methods themselves. It is reasonable to ask what general conclusions can be drawn from the applications discussed in this review. Many interesting ideas are contained in the original papers that have not been discussed here, but the following observations consider some of the more general points we have found in preparing this article. (1) Quantum chemistry calculations that use large basis sets and include some description of electron correlation give a reasonably consistent picture of the nature of the (gas-phase) potential-energy surface for small dipeptide models, and recent empirical force fields reproduce this behavior with good fidelity. Uncertainties in relative conformational energies are probably still larger than room temperature thermal energies, so that calculations that involve Boltzmann averages over various conformations must still be interpreted with caution. There is also a large literature, not discussed here, on empirical and quantum chemistry calibrations of force fields for amino acid side chains. The possible intrinsic errors in the resulting potentials are harder to evaluate, but may be substantial, particularly for polar and charged groups. (2) Estimates of solvation contributions to conformational energy differences are more difficult to make, both because it is difficult to find direct and detailed experimental tests and because various theoretical estimates give somewhat disparate results. The general features at the dipeptide level (stabilization of the αR conformation and broadening and merging of the C7 eq and C5 conformations) are reproduced with many models. (3) Free-energy calculations have become a powerful tool for mapping out conformational thermodynamics in a series of interesting models for helices, turns, and sheets. Although there are still some unresolved questions about the level of statisticial convergence, results using different approaches appear to be in good qualitative accord. Theoretical estimates of helical propensities for different amino acids (and different locations in the helix) can be compared to experimental estimates in model peptide and protein systems, and a combination of theory and experiment provides good insight into the origins of sequence-dependent effects in these systems. We can anticipate continuing insights from simulations in this area. (4) Free-energy calculations on forming NH⋯OC hydrogen bonds in short turns or for an isolated linear hydrogen bond show only a very small hydrogen-bond stabilization in aqueous solution. Greater stabilization is found for a β-sheet model with two adjacent hydrogen bonds, a result that is attributed to the poorer solvation of the separated strands relative to the individual amide groups. Results of this sort can provide important clues to the interpretation of the driving force for secondary structure formation in proteins. (5) Molecular dynamics simulations have also provided interesting insights into the time scales and intermediates involved in formation of helices and turns from more random peptide conformations. Bifurcated hydrogen bonds and i ← i + 3 interactions appear to be common intermediates, and local transitions (involving just a few residues in a turn, or at the end of a helix) occur with residence times in the nanosecond range. Again, these broad conclusions do not appear to depend upon details of the potential functions or simulation protocols used. It is also of interest to speculate what the future might hold. Clearly, computer hardware and software will continue to improve, making calculations like those discussed here easier to perform on a routine basis. This enhancement will certainly broaden the range of applications to consider more sequence-specific effects as well as more complicated geometrical and topological factors. This increased breadth should facilitate more direct comparisons to experiments on peptide and protein folding. We also expect to see developments in technical aspects of free-energy calculations, as in methods to reduce the spurious effects on long-range force truncation. In addition, there is an increasing focus on questions of convergence of sampling in free-energy simulations, which should lead to a better understanding of the errors to be expected. Finally, we note continuing progress on the experimental side toward greater temporal and spatial resolution in studies of peptides: NMR experiments can now probe peptide conformation on a residue-by-residue basis, and developments in time-resolved optical and IR spectroscopies allow more direct and detailed comparisons between theory and experiment. These considerations lead us to a optimistic outlook for the role of simulation methods in studies of peptide conformational dynamics and thermodynamics.
All Science Journal Classification (ASJC) codes