Dataset Information

Force Field for Water Based on Neural Network.

ABSTRACT: We developed a novel neural network-based force field for water based on training with high-level ab initio theory. The force field was built based on an electrostatically embedded many-body expansion method truncated at binary interactions. The many-body expansion method is a common strategy to partition the total Hamiltonian of large systems into a hierarchy of few-body terms. Neural networks were trained to represent electrostatically embedded one-body and two-body interactions, which require as input only one and two water molecule calculations at the level of ab initio electronic structure method CCSD/aug-cc-pVDZ embedded in the molecular mechanics water environment, making it efficient as a general force field construction approach. Structural and dynamic properties of liquid water calculated with our force field show good agreement with experimental results. We constructed two sets of neural network based force fields: nonpolarizable and polarizable force fields. Simulation results show that the nonpolarizable force field using fixed TIP3P charges has already behaved well, since polarization effects and many-body effects are implicitly included due to the electrostatic embedding scheme. Our results demonstrate that the electrostatically embedded many-body expansion combined with neural network provides a promising and systematic way to build next-generation force fields at high accuracy and low computational costs, especially for large systems.

SUBMITTER: Wang H

PROVIDER: S-EPMC6241215 | biostudies-literature | 2018 Jun

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Force Field for Water Based on Neural Network.

Wang Hao H Yang Weitao W

The journal of physical chemistry letters 20180604 12

We developed a novel neural network-based force field for water based on training with high-level ab initio theory. The force field was built based on an electrostatically embedded many-body expansion method truncated at binary interactions. The many-body expansion method is a common strategy to partition the total Hamiltonian of large systems into a hierarchy of few-body terms. Neural networks were trained to represent electrostatically embedded one-body and two-body interactions, which require ...[more]

PMID: 29775313

Similar Datasets

Project description:A quantum mechanical force field (QMFF) for water is described. Unlike traditional approaches that use quantum mechanical results and experimental data to parameterize empirical potential energy functions, the present QMFF uses a quantum mechanical framework to represent intramolecular and intermolecular interactions in an entire condensed-phase system. In particular, the internal energy terms used in molecular mechanics are replaced by a quantum mechanical formalism that naturally includes electronic polarization due to intermolecular interactions and its effects on the force constants of the intramolecular force field. As a quantum mechanical force field, both intermolecular interactions and the Hamiltonian describing the individual molecular fragments can be parameterized to strive for accuracy and computational efficiency. In this work, we introduce a polarizable molecular orbital model Hamiltonian for water and for oxygen- and hydrogen-containing compounds, whereas the electrostatic potential responsible for intermolecular interactions in the liquid and in solution is modeled by a three-point charge representation that realistically reproduces the total molecular dipole moment and the local hybridization contributions. The present QMFF for water, which is called the XP3P (explicit polarization with three-point-charge potential) model, is suitable for modeling both gas-phase clusters and liquid water. The paper demonstrates the performance of the XP3P model for water and proton clusters and the properties of the pure liquid from about 900 × 10(6) self-consistent-field calculations on a periodic system consisting of 267 water molecules. The unusual dipole derivative behavior of water, which is incorrectly modeled in molecular mechanics, is naturally reproduced as a result of an electronic structural treatment of chemical bonding by XP3P. We anticipate that the XP3P model will be useful for studying proton transport in solution and solid phases as well as across biological ion channels through membranes.

Project description:Computational prediction of noncovalent binding free energies with methods based on molecular mechanical force fields has become increasingly routine in drug discovery projects, where they promise to speed the discovery of small molecule ligands to bind targeted proteins with high affinity. Because the reliability of free energy methods still has significant room for improvement, new force fields, or modifications of existing ones, are regularly introduced with the aim of improving the accuracy of molecular simulations. However, comparatively little work has been done to systematically assess how well force fields perform, particularly in relation to the calculation of binding affinities. Hardware advances have made these calculations feasible, but comprehensive force field assessments for protein-ligand sized systems still remain costly. Here, we turn to cyclodextrin host-guest systems, which feature many hallmarks of protein-ligand binding interactions but are generally much more tractable due to their small size. We present absolute binding free energy and enthalpy calculations, using the attach-pull-release (APR) approach, on a set of 43 cyclodextrin-guest pairs for which experimental ITC data are available. The test set comprises both α- and β-cyclodextrin hosts binding a series of small organic guests, each with one of three functional groups: ammonium, alcohol, or carboxylate. Four water models are considered (TIP3P, TIP4Pew, SPC/E, and OPC), along with two partial charge assignment procedures (RESP and AM1-BCC) and two cyclodextrin host force fields. The results suggest a complex set of considerations when choosing a force field for biomolecular simulations. For example, some force field combinations clearly outperform others at the binding enthalpy calculations but not for the binding free energy. Additionally, a force field combination which we expected to be the worst performer gave the most accurate binding free energies - but the least accurate binding enthalpies. The results have implications for the development of improved force fields, and we propose this test set, and potential future elaborations of it, as a powerful validation suite to evaluate new force fields and help guide future force field development.

Project description:The use of neural networks to predict molecular properties calculated from high level quantum mechanical calculations has made significant advances in recent years, but most models need input geometries from DFT optimizations which limit their applicability in practice. In this work, we explored how machine learning can be used to predict molecular atomization energies and conformation stability using optimized geometries from Merck Molecular Force Field (MMFF). On the basis of the recently introduced deep tensor neural network (DTNN) approach, we first improved its training efficiency and performed an extensive search of its hyperparameters, and developed a DTNN_7ib model which has a test accuracy of 0.34 kcal/mol mean absolute error (MAE) on QM9 data set. Then using atomic vector representations in the DTNN_7ib model, we employed transfer learning (TL) strategy to train readout layers on the QM9M data set, in which QM properties are the same as in QM9 [calculated at the B3LYP/6-31G(2df,p) level] while molecular geometries are corresponding local minima optimized with MMFF94 force field. The developed TL_QM9M model can achieve an MAE of 0.79 kcal/mol using MMFF optimized geometries. Furthermore, we demonstrated that the same transfer learning strategy with the same atomic vector representation can be used to develop a machine learning model that can achieve an MAE of 0.51 kcal/mol in molecular energy prediction using MMFF geometries for an eMol9_CM conformation data set, which consists of 9959 molecules and 88?234 conformations with energies calculated at the B3LYP/6-31G* level. Our results indicate that DFT-level accuracy of molecular energy prediction can be achieved using force-field optimized geometries and atomic vector representations learned from deep tensor neural network, and integrated molecular modeling and machine learning would be a promising approach to develop more powerful computational tools for molecular conformation analysis.

Dataset Information

Force Field for Water Based on Neural Network.

Publications

Force Field for Water Based on Neural Network.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets