scholarly journals Linear Atomic Cluster Expansion Force Fields for Organic Molecules: beyond RMSE

Author(s):  
David Peter Kovacs ◽  
Cas van der Oord ◽  
Jiri Kucera ◽  
Alice Allen ◽  
Daniel Cole ◽  
...  
2021 ◽  
Author(s):  
David Peter Kovacs ◽  
Cas van der Oord ◽  
Jiri Kucera ◽  
Alice Allen ◽  
Daniel Cole ◽  
...  

2021 ◽  
Author(s):  
David Peter Kovacs ◽  
Cas van der Oord ◽  
Jiri Kucera ◽  
Alice Allen ◽  
Daniel Cole ◽  
...  

2021 ◽  
Author(s):  
David Peter Kovacs ◽  
Cas van der Oord ◽  
Jiri Kucera ◽  
Alice Allen ◽  
Daniel Cole ◽  
...  

We demonstrate that fast and accurate linear force fields can be built for molecules using the Atomic Cluster Expansion (ACE) framework. The ACE models parametrize the Potential Energy Surface in terms of body ordered symmetric polynomials making the functional form reminiscent of traditional molecular mechanics force fields. We show that the 4 or 5-body ACE force fields improve on the accuracy of the empirical force fields by up to a factor of 10, reaching the accuracy typical of recently proposed machine learning based approaches. We not only show state of the art accuracy and speed on the widely used MD17 and ISO17 benchmark datasets, but also go beyond RMSE by comparing a number of ML and empirical force fields to ACE on more important tasks such as normal mode prediction, high temperature molecular dynamics, dihedral torsional profile prediction and even bond breaking. We also demonstrate the smoothness, transferability and extrapolation capabilities of ACE on a new challenging benchmark dataset comprising a potential energy surface of a flexible drug-like molecule.


2018 ◽  
Author(s):  
Maximiliano Riquelme ◽  
Alejandro Lara ◽  
David L. Mobley ◽  
Toon Vestraelen ◽  
Adelio R Matamala ◽  
...  

<div>Computer simulations of bio-molecular systems often use force fields, which are combinations of simple empirical atom-based functions to describe the molecular interactions. Even though polarizable force fields give a more detailed description of intermolecular interactions, nonpolarizable force fields, developed several decades ago, are often still preferred because of their reduced computation cost. Electrostatic interactions play a major role in bio-molecular systems and are therein described by atomic point charges.</div><div>In this work, we address the performance of different atomic charges to reproduce experimental hydration free energies in the FreeSolv database in combination with the GAFF force field. Atomic charges were calculated by two atoms-in-molecules approaches, Hirshfeld-I and Minimal Basis Iterative Stockholder (MBIS). To account for polarization effects, the charges were derived from the solute's electron density computed with an implicit solvent model and the energy required to polarize the solute was added to the free energy cycle. The calculated hydration free energies were analyzed with an error model, revealing systematic errors associated with specific functional groups or chemical elements. The best agreement with the experimental data is observed for the MBIS atomic charge method, including the solvent polarization, with a root mean square error of 2.0 kcal mol<sup>-1</sup> for the 613 organic molecules studied. The largest deviation was observed for phosphor-containing molecules and the molecules with amide, ester and amine functional groups.</div>


Author(s):  
Joshua Horton ◽  
Alice Allen ◽  
Leela Dodda ◽  
Daniel Cole

<div><div><div><p>Modern molecular mechanics force fields are widely used for modelling the dynamics and interactions of small organic molecules using libraries of transferable force field parameters. For molecules outside the training set, parameters may be missing or inaccurate, and in these cases, it may be preferable to derive molecule-specific parameters. Here we present an intuitive parameter derivation toolkit, QUBEKit (QUantum mechanical BEspoke Kit), which enables the automated generation of system-specific small molecule force field parameters directly from quantum mechanics. QUBEKit is written in python and combines the latest QM parameter derivation methodologies with a novel method for deriving the positions and charges of off-center virtual sites. As a proof of concept, we have re-derived a complete set of parameters for 109 small organic molecules, and assessed the accuracy by comparing computed liquid properties with experiment. QUBEKit gives highly competitive results when compared to standard transferable force fields, with mean unsigned errors of 0.024 g/cm3, 0.79 kcal/mol and 1.17 kcal/mol for the liquid density, heat of vaporization and free energy of hydration respectively. This indicates that the derived parameters are suitable for molecular modelling applications, including computer-aided drug design.</p></div></div></div>


Author(s):  
Joshua Horton ◽  
Alice Allen ◽  
Leela Dodda ◽  
Daniel Cole

<div><div><div><p>Modern molecular mechanics force fields are widely used for modelling the dynamics and interactions of small organic molecules using libraries of transferable force field parameters. For molecules outside the training set, parameters may be missing or inaccurate, and in these cases, it may be preferable to derive molecule-specific parameters. Here we present an intuitive parameter derivation toolkit, QUBEKit (QUantum mechanical BEspoke Kit), which enables the automated generation of system-specific small molecule force field parameters directly from quantum mechanics. QUBEKit is written in python and combines the latest QM parameter derivation methodologies with a novel method for deriving the positions and charges of off-center virtual sites. As a proof of concept, we have re-derived a complete set of parameters for 109 small organic molecules, and assessed the accuracy by comparing computed liquid properties with experiment. QUBEKit gives highly competitive results when compared to standard transferable force fields, with mean unsigned errors of 0.024 g/cm3, 0.79 kcal/mol and 1.17 kcal/mol for the liquid density, heat of vaporization and free energy of hydration respectively. This indicates that the derived parameters are suitable for molecular modelling applications, including computer-aided drug design.</p></div></div></div>


Author(s):  
Joaquin Miranda ◽  
Thomas Gruhn

Using a cluster expansion scheme, we evaluate the structural stability of (V, Nb)CoSb half-Heusler alloys over a wide range of V and Nb concentrations when the alloy is simultaneously subject...


Tetrahedron ◽  
2021 ◽  
Vol 79 ◽  
pp. 131865
Author(s):  
Toby Lewis-Atwell ◽  
Piers A. Townsend ◽  
Matthew N. Grayson

2019 ◽  
Vol 5 (5) ◽  
pp. eaaw2210 ◽  
Author(s):  
Alessandro Lunghi ◽  
Stefano Sanvito

Computational studies of chemical processes taking place over extended size and time scales are inaccessible by electronic structure theories and can be tackled only by atomistic models such as force fields. These have evolved over the years to describe the most diverse systems. However, as we improve the performance of a force field for a particular physical/chemical situation, we are also moving away from a unified description. Here, we demonstrate that a unified picture of the covalent bond is achievable within the framework of machine learning–based force fields. Ridge regression, together with a representation of the atomic environment in terms of bispectrum components, can be used to map a general potential energy surface for molecular systems at chemical accuracy. This protocol sets the ground for the generation of an accurate and universal class of potentials for both organic and organometallic compounds with no specific assumptions on the chemistry involved.


Sign in / Sign up

Export Citation Format

Share Document