Incorporating Neural Networks into the AMOEBA Polarizable Force Field

16 November 2023, Version 1
This content is a preprint and has not undergone peer review at the time of posting.

Abstract

Neural network potentials (NNPs) have great potential to bridge the gap between the accuracy of quantum mechanics and the efficiency of molecular mechanics in molecular simulation. However, most of the NNPs remain restricted by the locality assumption that ensures the model's transferability and scalability but misses out the long-range interactions. Here we present an integrated non-reactive hybrid model, AMOEBA+NN, which employs the AMOEBA potential for the short- and long-range non-bonded interactions and an NNP to capture the remaining local (covalent) contributions. A first AMOEBA+NN model was trained on the conformational energy of ANI-1x dataset and tested on several external datasets ranging from small molecules to tetrapeptides. It was encouraging to see that the hybrid model becomes significantly advantageous over the baseline models in term of accuracy as the molecules get larger, offering perspectives for the development of a generalized and improved approach.

Keywords

Neural Network Potentials
AMOEBA
Polarizable Force Fields
Machine Learning
Deep Learning

Comments

Comments are not moderated before they are posted, but they can be removed by the site moderators if they are found to be in contravention of our Commenting Policy [opens in a new tab] - please read this policy before you post. Comments should be used for scholarly discussion of the content in question. You can find more information about how to use the commenting feature here [opens in a new tab] .
This site is protected by reCAPTCHA and the Google Privacy Policy [opens in a new tab] and Terms of Service [opens in a new tab] apply.