Search for a command to run...
The continued development and utility of molecular dynamics simulations requires improvements in both the physical models used (force fields) and in our ability to sample the Boltzmann distribution of these models. Recent developments in both areas have made available multi-microsecond simulations of two proteins, ubiquitin and Protein G, using a number of different force fields. Although these force fields mostly share a common mathematical form, they differ in their parameters and in the philosophy by which these were derived, and previous analyses showed varying levels of agreement with experimental NMR data. To complement the comparison to experiments, we have performed a structural analysis of and comparison between these simulations, thereby providing insight into the relationship between force-field parameterization, the resulting ensemble of conformations and the agreement with experiments. In particular, our results show that, at a coarse level, many of the motional properties are preserved across several, though not all, force fields. At a finer level of detail, however, there are distinct differences in both the structure and dynamics of the two proteins, which can, together with comparison with experimental data, help to select force fields for simulations of proteins. A noteworthy observation is that force fields that have been reparameterized and improved to provide a more accurate energetic description of the balance between helical and coil structures are difficult to distinguish from their "unbalanced" counterparts in these simulations. This observation implies that simulations of stable, folded proteins, even those reaching 10 microseconds in length, may provide relatively little information that can be used to modify torsion parameters to achieve an accurate balance between different secondary structural elements.