From Theory of Energy Decomposition Analysis to Rational Force Field Design

University Of California-Berkeley, Berkeley CA

Investigators

Abstract

Teresa Head-Gordon and Martin Head-Gordon are supported by an award from the Chemical Theory, Models and Computational Methods program in the Division of Chemistry to develop computational and theoretical models and methods that can be used to make accurate predictions for the behavior of complex molecular and bimolecular systems. This collaborative project combines the mutual development of advanced force fields and energy decomposition analysis (EDA). The field of biomolecular modeling and simulation currently relies on a simple representation of the potential energy surface of molecules based on what is known as the "pairwise additive" force field. But to design new drugs or new biomaterials the field needs new and higher accuracy molecular mechanics (MM) force fields. These new force fields introduce new terms that describe processes not considered in simple pair-wise force fields, for example, polarization and charge transfer. The introduction of such terms pose great challenges for rational force field design as well algorithmic and software challenges. The force field development is enabled in part by using information from an energy decomposition method (EDA) that is based on state-of-the art quantum mechanical (QM) calculations. This collaborative proposal is a partnership that couple's advances in EDA methods with the design and validation of next generation force fields, in a way that permits each to advance the impact of the other. These methods are used to help synthetic chemists make new drugs, to aid in the design of new functional materials, and to create new catalysts that speed up chemical reactions by many orders of magnitude. All the methods developments resulting from this proposal are being implemented in many widely used computational chemistry software packages. The proposed research is built on three main thrusts. First, to impact the efficiency of polarizable force fields, a new procedure to avoid self-consistent iterations is being developed for use in MM dynamics, and within the inner loop of QM/MM calculations. Second, the two research groups are engaged in a major effort to test and improve MM parameters in polarizable force fields for condensed phase modeling, to assess non-bonded parameters for Pauli-exclusion, permanent electrostatics, polarization, and to explore models for additional short-range contributions from charge penetration, charge transfer, and exchange repulsion. In addition, attention is paid to torsional and solute-solvent interactions in biomolecules, where there is significant scope for improvement. These efforts, informed by EDA tools, are complemented by validation studies including scalar couplings in polypeptides, Stark effects and melting curves in three proteins. Specific outcomes include an upgraded polarizable potential of the AMOEBA family, as a well as more advanced model that includes additional non-classical terms. The third thrust is the development of EDA methods. New capabilities that inform the MM developments include force decomposition analysis, partitioning of polarization into direct and indirect terms and the many-body decomposition of the latter. A soundly based EDA for coupled cluster (CC) theory is being developed. Finally, new advances in variational EDA for chemical bonds are pioneered including multiple bonds and dynamical correlation. When developed the advanced MM potential energy surfaces methodology and models will be provided to OpenMM, OMNIA, TINKER, and LibEFP, and the new proposed EDA capabilities to Q-Chem and ONETEP. The Head-Gordon groups develop training materials based on this research for use in computational chemistry community workshops and schools. Plans are to distribute these materials through interaction with the Molecular Sciences Software Institute (MolSSI) as well as through Q-Chem's instructional support web pages