Biomolecular dynamics with machine-learned quantum-mechanical force fields trained on diverse chemical fragments

Sci Adv. 2024 Apr 5;10(14):eadn4397. doi: 10.1126/sciadv.adn4397. Epub 2024 Apr 5.

Oliver T Unke^{1

2

3}, Martin Stöhr⁴, Stefan Ganscha¹, Thomas Unterthiner¹, Hartmut Maennel¹, Sergii Kashubin¹, Daniel Ahlin¹, Michael Gastegger^{2

3

5}, Leonardo Medrano Sandonas⁴, Joshua T Berryman⁴, Alexandre Tkatchenko⁴, Klaus-Robert Müller^{1

2

6

7

8}

¹ Google DeepMind, Tucholskystraße 2, 10117 Berlin, Germany and Brandschenkestrasse 110, 8002 Zürich, Switzerland.
² Machine Learning Group, Technische Universität Berlin, 10587 Berlin, Germany.
³ DFG Cluster of Excellence "Unifying Systems in Catalysis" (UniSysCat), Technische Universität Berlin, 10623 Berlin, Germany.
⁴ Department of Physics and Materials Science, University of Luxembourg, L-1511 Luxembourg City, Luxembourg.
⁵ BASLEARN - TU Berlin/BASF Joint Lab for Machine Learning, Technische Universität Berlin, 10587 Berlin, Germany.
⁶ Department of Artificial Intelligence, Korea University, Anam-dong, Seongbuk-gu, Seoul 02841, Korea.
⁷ Max Planck Institute for Informatics, Stuhlsatzenhausweg, 66123 Saarbrücken, Germany.
⁸ BIFOLD - Berlin Institute for the Foundations of Learning and Data, Berlin, Germany.

Abstract

The GEMS method enables molecular dynamics simulations of large heterogeneous systems at ab initio quality.