The dynamical properties of complex systems are often characterized by the existence of several (meta)stable states separated by large free energy barriers. Examples prevalent throughout nature,vary from conformational changes in biomolecules to phase transitions, as well as many chemical reactions. The long time behavior of such systems is determined by transitions between the stable states, e.g. conformational or structural changes. Due to the large free energy barriers between stable states these transitions are very rare compared the shor molecular timescale (e.g. dictated by vibrations) within each of the stable states, resulting in a separation of time scales. This so-called time scale problem makes it very unpractical to study such a system over an extended time scale with regular molecular dynamics (MD) simulations. One possibility to address this problem is the use of a biasing potential along a set of order parameters. Several computational techniques, e.g umbrella sampling, blue moon sampling, hyper-dynamics, and meta-dynamics take this route. The main drawback of these techniques is the choice of order parameters. If the order parameters do not provide a good approximation to the true reaction coordinate these methods fail. Transition Path Sampling (TPS) circumvents these problems by creating an ensemble of unbiased dynamical trajectories between the reactant and products states, through performing a random walk in trajectory space via a Metropolis Monte Carlo procedure.
Further information can be found in several review papers.
While transition path sampling can be easily programmed in any MD simulation package, the handling of trajectory files is cumbersome. The OpenPathSampling package provides an easy to use implementation of several TPS algorithms. OPS also provides several tutorials to learn about the methods.
A old tutorial is given here.