Automating chemistry of small molecules with the AMOEBA model

Small molecules are ubiquitous in protein, DNA, and RNA systems of interest. Affinity prediction and rationalization of biomolecular recognition are routinely investigated using molecular dynamics simulations for applications such as drug discovery. The Atomic Multipole Optimized Energetics for Biomolecular Applications (AMOEBA) model has been demonstrated to be superior in many cases to traditionally popular fixed charge models, due to its advanced treatment of electrostatic interactions [1-7]. Models such as AMOEBA are composed of a set of equations and parameters describing the energetics and forces in biomolecules. The possible parameter space of small molecules is, however, prohibitively large, additionally requiring very tedious manual steps and model-specific expertise to derive adequate parameters. Thus, an automated protocol for high-throughput parameterization of new molecules without the oversight of AMOEBA experts is necessary. Furthermore, no standardized programs that encapsulate the many details important for simulating systems and predicting properties with AMOEBA exist, adding to the engineering challenges for conducting scientific research with AMOEBA. This dissertation will focus on the development of software infrastructure for our automated parameterization scheme. In addition, software for setting up and running molecular dynamic simulations for predicting binding free energy for host-guest systems and hydration free energy for small molecules will also be discussed. The parameterization software will be discussed first, followed then by parameter database development and finally example systems of ATP-Mg²⁺ studied using the parameters obtained from the software. This protocol will enable non-AMOEBA experts to quickly parameterize molecules thus enabling more accurate, efficient, and reproducible molecular dynamic simulations studies with AMOEBA.