Open Science for Molecular Modelling with the Open Force Field Initiative

Abstract

Drawing on computational methods that are based around training to extensive condensed phase physical property and quantum mechanical datasets, I will describe some of our efforts to design accurate and transferable inter- and intra-molecular potentials, with a view to applications in condensed phase atomistic modelling and computer-aided drug design. I will explain how recent collaborations with the Open Force Field Initiative (https://openforcefield.org) enable the development of fast, accurate alternatives to traditional non-bonded functional forms. I will describe the development of a graph neural network based charge model targeting accurate electrostatic properties of organic molecules, and the use of Open Force Field infrastructure to train alternatives to the Lennard-Jones functional form. Finally, I will describe progress towards fast GPU-based optimisation of valence parameters and automated bond/angle/torsion typing schemes for accurate simulation of conformational dynamics.