Fusing heterogeneous data for the calibration of molecular dynamics force fields using hierarchical Bayesian models.

Area of Science:

• Computational chemistry
• Statistical mechanics
• Data science

Background:

• Molecular Dynamics (MD) simulations require accurate force fields for reliable predictions.
• Calibrating these force fields often relies on limited or heterogeneous experimental data.
• Integrating diverse datasets systematically remains a challenge in force field development.

Purpose of the Study:

• To develop a hierarchical Bayesian framework for integrating heterogeneous experimental data.
• To enable robust calibration of Molecular Dynamics (MD) force fields.
• To overcome limitations of subjective data weighting in parameterization.

Main Methods:

• Hierarchical Bayesian modeling to fuse diverse experimental data.
• Application to water using diffusivity, radial distribution function, and density data.
• Development of a surrogate model using the empirical interpolation method.
• Implementation of a parallel transitional Markov chain Monte Carlo technique for computational efficiency.

Main Results:

• Demonstrated successful robust calibration of MD force fields for water.
• Successfully integrated heterogeneous experimental data under various thermodynamic conditions.
• The framework bypasses subjective weighting of experimental data for parameter identification.

Conclusions:

• The proposed hierarchical Bayesian framework offers a systematic approach to force field calibration.
• The integration of diverse experimental data leads to more robust and reliable force fields.
• Computational efficiency is achieved through surrogate modeling and parallel computing techniques.