On-the-fly Probability Enhanced Sampling

Abstract

Molecular simulations have become an important scientific tool with several applications in physics, chemistry, material science and biology. Therefore, the development of more efficient and robust computational methods is of great importance, as any improvement in this respect can potentially have a positive impact on various open research fronts. One of the major challenges for molecular simulations is the sampling of so-called rare events, i.e. microscopic phenomena occurring on macroscopic time scales. Typical examples of rare events are phase transitions, chemical reactions, or protein folding. Here we propose a novel computational technique, called on-the-fly probability enhanced sampling (OPES) method, which can significantly improve molecular simulations of such phenomena. The OPES method can be described as an evolution of another popular enhanced sampling method, metadynamics, with respect to which it brings both conceptual and practical improvements. It goes a long way toward making enhanced sampling not only more efficient, but also more robust and easier to use. Furthermore, OPES unifies in the same approach two traditionally distinct sampling strategies, namely collective-variable-based methods and expanded-ensembles methods. We believe that this perspective will open interesting new possibilities in the field of enhanced sampling.