Advanced Sampling for Molecular Simulation is Coming of Age

In the past decades, molecular simulation has progressively grown to be the third strong leg of molecular science, which can more effectively complement experiments and theories by bridging microscopic details and macroscopic observations. Generally speaking, the quality of a molecular simulation is governed by two key factors, energy function accuracy and sampling adequacy. Using conventional molecular dynamics simulation methods, any meaningful gain in either of these two aspects requires significantly increasing demand of computing resources, such as the number of processors and simulation time. Constrained by commonly tractable computing power, even with an optimal choice of the combination of model representation and simulation length, accurate prediction of thermodynamic or kinetic properties of complex chemical and biochemical events can still be challenging.

In a natural process, molecules aimlessly fluctuate; and energy transfer within the molecular system and between the molecular system and the thermal reservoir rigorously follows the physical nature, which enforces faithful molecular dynamics simulations to bear the same time-scale efforts in order to produce adequate samples for the corresponding behaviors. Facing the well-known timescale challenge, one of the most important conceptual breakthroughs has been the realization that molecular simulations do not have to loyally imitate nature. Cyber space allows one to re-stratify the propagation of molecular fluctuations, provided that target thermodynamic or kinetic information can be rigorously recovered. Through almost three decades of development, “enhanced sampling” has become a focal topic in the field of molecular simulation. It has reached a level of maturity that allows many problems to be readily tackled. More importantly, many key concepts have been better defined and more thoroughly developed; and associated techniques have been more accurately summarized and categorized. The consequent physical and mathematical rigor is destined to lead this topic to arrive at further successes in applications of simulation to decipher the intimate relationship between molecular structure and function.

To celebrate the coming of age of “advanced sampling for molecular simulation”, we organized this special issue, which collects recent works from eleven research groups that are active in this topic. These diverse works broadly represent the current trend of thinking and developments. The sampling techniques addressed in this issue include novel approaches to milestoning¹, string path sampling², replica exchange^3,4, orthogonal space sampling^5,6, accelerated molecular dynamics^7–9, self-guided Langevin dynamics¹⁰, metadynamics¹¹, and assorted variants of these algorithms. We hope that this special issue will not only effectively document this important era of molecular simulation but also motivate more creative work in the near future.