Improved conformational sampling methods for molecular dynamics simulations

Abstract

Understanding conformational dynamics of biomolecules such as proteins is a fundamental challenge in structural biology. Native conformations of proteins can be determined experimentally via X-Ray Crystallography and NMR spectroscopy but usually such methods provide time averaged data. All-atom simulations are commonly used to supplement experimental observations where time dependent trajectories for complex systems can be obtained. However there are major challenges in computer simulations.

For successful simulations the potential function has to be accurate enough to correctly rank the local and global energy minima and the barriers in between for the simulated system. We developed an efficient method to test the accuracy of force field parameters where the energies of pre-generated conformations (decoys) were calculated for each parameter set in question and the identified energy minima were compared to experimental measurements. After generating decoy sets the evaluation of force field or other simulation parameters can be done quickly and efficiently. We used this decoy screening procedure to identify α-helical bias in existing force fields in AMBER and to develop improved force field parameters.

Another major challenge in simulations is sampling because the time scales reached with standard simulations are 3-6 orders of magnitude shorter than actual comformational transitions observed in proteins. There are several new sampling methods available where transitions between energy minima are enhanced through the use of high temperatures. Such methods are still very computationally demanding and can only be applied to small systems. We have developed two new methods to further enhance the conformational sampling to reduce computational demands and increase the convergence speed of simulations.