Studying rare events using forward-flux sampling: Recent breakthroughs and future outlook

Abstract

Rare events are processes that occur upon the emergence of unlikely fluctuations. Unlike what their name suggests, rare events are fairly ubiquitous in nature, as the occurrence of many structural transformations in biology and material sciences is predicated upon crossing large free energy barriers. Probing the kinetics and uncovering the molecular mechanisms of possible barrier crossings in a system is critical to predicting and controlling its structural and functional properties. Due to their activated nature, however, rare events are exceptionally difficult to study using conventional experimental and computational techniques. In recent decades, a wide variety of specialized computational techniques-known as advanced sampling techniques-have been developed to systematically capture improbable fluctuations relevant to rare events. In this perspective, we focus on a technique called forward flux sampling (Allen et al., J. Chem. Phys., 124: 024102, 2006), and overview its recent methodological variants and extensions. We also provide a detailed overview of its application to study a wide variety of rare events, and map out potential avenues for further explorations.

Parasitic crystallization of colloidal electrolytes: growing a metastable crystal from the nucleus of a stable phase

“…Since brute force results contain no approximation, and they can be regarded as a benchmark for methodologies aiming to estimate J for nucleation, these results support the validity of CNT in describing the nucleation behaviour of colloidal electrolytes (at least at T* = 1). The discrepancy between the Seeding and FFS nucleation rates is somewhat not surprising since, on the one hand, it has been previously found that FFS calculations may underestimate the nucleation rate by a few orders of magnitude when the sampling is not extremely large, 26,44,114,115 and on the other, because Seeding results may also have an uncertainty (systematic + statistical) of about 3-5 orders of magnitude in the nucleation rate (see Fig. 5).…”

Forward Flux Sampling of Polymer Desorption Paths from a Solid Surface into Dilute Solution

“…Future studies of more complex polymer desorption phenomena should take advantage of recent advances that modified FFS algorithms to account for time-dependent dynamics, “jumpy” dynamics, and multiple metastable states [ 14 ]. For example, a time-dependent flow field could be introduced, or the attraction between the wall and polymer may be reduced resulting in faster ‘jumps’ away from the wall, or the surface could be patterned resulting in multiple stable basins.…”

“…This review exposes a fundamental gap in the literature, as no published work analyzes the most basic problem of desorption of isolated chains from a simple flat surface in the absence of flow. In recent years, forward flux sampling has drastically progressed due to a repeatable workflow, innovations to reduce system constraints, and the application of the method to new fields [ 14 ]. In addition, FFS simulations use an unbiased potential and can address both equilibrium and non-equilibrium systems.…”

“…In addition, FFS simulations use an unbiased potential and can address both equilibrium and non-equilibrium systems. Many groups have used FFS for simulating nucleation, magnetic switching, gene switching, protein or DNA behavior, and membrane transport [ 14 ]. In the polymer field, FFS has been applied to polymer translocation [ 15 ], structural relaxation [ 16 , 17 ], and conformational rearrangement [ 18 , 19 , 20 ].…”

DBFOLD: An efficient algorithm for computing folding pathways of complex proteins

“…However, the implementation of replica exchange comes at the expense of realistic state transition kinetics, and replica exchange is unlikely to promote sampling saddle points in the free energy landscape, thus hindering barrier-height computation, Biasing techniques such as umbrella sampling [24,25] and Metadynamics [26] can improve sampling along saddle points, but they are only useful if proper order parameters or collective variables along which slow transitions occur are known in advance, which is not the case for most proteins. Other sampling techniques such as transition-path sampling [27,28] and forward flux sampling [29] are more tolerant of uncertainty in the order parameter(s), but these are extremely computationally expensive to implement for large proteins with multiple intermediates.…”