Force-field parametrization, multiple birds with multiple stones – with REST2
This blog post presents an idea for a force-field parametrization strategy that leverages replica exchanges between different parametrizations of the same system.
One of the major challenges in force-field parametrization is the need to explore a vast parameter space. Typically, a series of parameter values are evaluated sequentially in an effort to match an experimental observable, but this process can be slow. Additionally, if a particular parameter choice introduces an energy barrier, it may hinder efficient sampling, making it difficult to achieve convergence on a numerical quantity that can be compared with experimental data.
To address this, temperature replica exchange can be used to enhance sampling. In this approach, replicas of the same Hamiltonian are run at different temperatures, with exchanges attempted according to the Boltzmann criterion. A key aspect of this method is comparing the energies of neighboring replicas.
A similar approach could be applied to force-field parametrization. Rather than running replicas at multiple temperatures, one could run multiple Hamiltonians, each corresponding to different parameter sets of interest. Each replica’s potential energy would then be compared to its neighbors, allowing for an exchange between them.
As a simple example, consider starting with an existing molecular model, such as the TIP3P water model, and introducing a small perturbation (λ) to generate replicas with slightly varied van der Waals (VdW) and charge parameters compared to the original model. For each replica, an observable—such as density, heat capacity, or another property—could be computed and compared to experimental results. The ability to exchange states with neighboring replicas should help accelerate convergence of the observable.