ForceBalance is a method and a freely available software tool for creating more accurate force fields using flexible combinations of reference data from experimental measurements and theoretical calculations.In classical molecular mechanics simulation, empirical potential energy functions (force fields) govern the molecular motions and the thermodynamic ensemble. Force fields derive their simplicity and computational tractability from making major approximations, which requires their parameters to be fitted to experimental data. Concurrently, there have also been successful efforts to derive force field parameters from theoretical calculations at higher levels of detail (e.g. ab initioquantum chemistry), using methods such as relative entropy and force matching. It is also possible to adopt a compromise approach that tries to reproduce both experimental data and theoretical calculations, but this increases the complexity of the optimization problem. In general, developing an accurate force field often requires years of hard work, involving much manual tuning of parameters and trial and error.ForceBalance addresses these challenges in force field parameterization, allowing the researcher to easily create accurate, systematically improvable, and reproducible force fields. The main features of ForceBalance which distinguish it from other methods are:(1) An optimization framework which treats experimental and theoretical reference data simultaneously and on the same footing, so that models can be parameterized using increasingly large and diverse data sets with little to no added complexity.(2) Strict regularization schemes using Gaussian or Laplacian priors in a Bayesian interpretation to prevent overfitting or divergence of model parameters during the optimization.(3) Interfaces to many molecular dynamics software packages for flexibility and distributed computing libraries to enable computationally intensive parameterizations.We have applied ForceBalance to parameterize a wide range of force fields including polarizable water models, small ligands, and proteins. We will present an overview of the method along with a small but diverse and interesting set of applications.

ForceBalance, a method for developing better force fields

246th National Meeting (2013)

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.