Interpreting Experimental Data by Using Molecular Simulation Instead of Model Building

One of the most used spectroscopic techniques for resolving the structure of a biomolecule, such as a protein or peptide, is NMR spectroscopy. Because only NMR signal intensities and frequencies are measured in the experiment, a conformational interpretation of the primary data, that is, measured data, is not straightforward, especially for flexible molecules. It is hampered by the occurrence of conformational and/or time-averaging, by insufficient number of experimental data and by insufficient accuracy of experimental data. All three problematic aspects of structure refinement based on NMR nuclear Overhauser effect (NOE) intensities and (3)J coupling data are illustrated by using two beta-heptapeptides in methanol as an example. We have performed 16 molecular dynamics (MD) simulations between 20 to 100 ns in length of unrestrained and NOE distance-restrained cases (instantaneous and time-averaged) of two beta-heptapeptides with a central beta-HAla(alpha-OH) amino acid in methanol at two different temperatures using two different GROMOS force-field parameter sets, 45A3 and 53A6. The created conformational ensembles were used to interpret the primary NMR data on these molecules. They also were compared to a set of NMR model structures derived by single-structure refinement in vacuo by using standard techniques. It is shown that the conformational interpretation of measured experimental data can be significantly improved by using unrestrained, instantaneous and time-averaged restrained MD simulations of the peptides by using a thermodynamically calibrated force field and by explicitly including solvent degrees of freedom.