Authors
Lily Wang, Open Molecular Software Foundation
We provide an overview of the Open Force Field (OpenFF) Initiative’s progress in data-driven approaches to developing next-generation force fields. We present our work towards developing a self-consistent biopolymer force field that simultaneously models both proteins and small molecules, with co-optimisation of small molecule and protein-specific parameters. This fitting procedure includes extensive evaluation to a range of protein QM data and validation with NMR experiments for a range of peptides and small proteins. In addition, we discuss extension of OpenFF towards lipids, ions, and water models, as examples of expanding our data-driven efforts by systematically augmenting QM data and adding physical data that has typically not been used in molecular simulation. We also present the development and performance of a neural network model for assigning fast, conformer-independent partial charges for molecules of arbitrary size, removing a significant limitation for approaches such as AM1-BCC.
