Sampling rare events in complex, high dimensional systems, such as crystal nucleation, protein folding and chemical reactions, remains a challenge for computational studies. Employing regular all-atom molecular dynamics (MD) simulations with a time step of a few femtoseconds becomes quickly unfeasible as the system tends to spend most of the time within a stable state hardly sampling the transition barrier regions of the phase space. Yet, these rare events often dominate the dynamical behaviour over an extended time scale. Among other approaches, transition path sampling (TPS) provides a possibility to explore transitions between stable states in rare event systems. One of the key advantages in TPS is that an a priori definition of a reaction coordinate is not required. In addition, since the dynamics used in TPS correspond to the actual underlying physical dynamics, the true kinetic mechanism is sampled. Transition interface sampling (TIS), a variant of TPS, has been developed to improve the calculation of rate constants by introducing a number of
interfaces along a certain order parameter, through which the positive effective flux can be measured. Here we introduce a reweighting scheme for the path ensembles within the TIS framework. Once the sampling has been performed within the biased TIS ensemble, the reweighting allows for the analysis of free energy landscapes and committor projections in an arbitrary order parameter space. The reweighted path ensemble can then be used to optimise non-linear reaction coordinates in any low dimensional collective variable space employing
a likelihood maximization approach.
Copyright. All Rights Reserved.