Artificial Intelligence Enhanced Molecular Simulations

Jun Zhang

doi:10.1021/acs.jctc.3c00214

Artificial Intelligence Enhanced Molecular Simulations

2023-06-26

43

PRE

AI

DOI：10.1021/acs.jctc.3c00214

Original

Original request for help

Save

Abstract

En 中文

Molecular simulations, which simulate the motions ofparticlesaccording to fundamental laws of physics, have been applied to a widerange of fields from physics and materials science to biochemistryand drug discovery. Developed for computationally intensive applications,most molecular simulation software involves significant use of hard-codedderivatives and code reuse across various programming languages. Inthis Review, we first align the relationship between molecular simulationsand artificial intelligence (AI) and reveal the coherence betweenthe two. We then discuss how the AI platform can create new possibilitiesand deliver new solutions to molecular simulations, from the perspectiveof algorithms, programming paradigms, and even hardware. Rather thanfocusing solely on increasingly complex neural network models, weintroduce various concepts and techniques brought about by modernAI and explore how they can be transacted to molecular simulations.To this end, we summarized several representative applications ofmolecular simulations enhanced by AI, including from differentiableprogramming and high-throughput simulations. Finally, we look aheadto promising directions that may help address existing issues in thecurrent framework of AI-enhanced molecular simulations.

Keywords:

FREE-ENERGY LANDSCAPES

DYNAMICS SIMULATIONS

NEURAL-NETWORK

TRANSITION

ALGORITHMS

PARALLEL

COMPLEX

AMBER

FRAMEWORK

Artificial Intelligence Enhanced Molecular Simulations

Abstract

Journal

Organization