New York University
One of the computational grand challenge problems is the development of methodology capable of sampling conformational equilibria in systems characterized by rough energy landscapes. If met, many important problems, most notably biomolecular structure prediction and the discovery of the polymorphism in organic molecular crystals could be significantly impacted. In this talk, I will discuss new approaches for enhancing sampling and mapping out the free energy landscape of systems described by rough potential energy surfaces. The new techniques are based on a molecular dynamics calculation in which a set of pre-selected collective variables (CVs) is singled out for enhanced sampling. The enhancement is achieved by applying a high temperature to the CVs and decoupling them adiabatically from the remaining degrees of freedom. This framework also allows incorporation of bias potentials for increased efficiency and will be shown to outperform popular approaches such as metadynamics. Application of an isobaric version of the method to organic molecular crystals such as benzene and naphthalene allows all of the known polymorphs to be identified from a short simulation. I will demonstrate the performance of a biased version of the enhanced sampling approach by generating the conformational equilibria of a number of small polypeptides. Finally, I will discuss a new multi-scale approach, employing gentlest ascent dynamics and stochastic relaxation, for rapidly locating minima and saddle points on the free energy landscape. Such an approach leads directly to a graph-based representation of high-dimensional free energy surfaces.