Approaching coupled cluster accuracy with a general-purpose neural network potential through transfer learning

By Justin S. Smith, B. Nebgen, Roman I. Zubatyuk, Nicholas Lubbers, Christian Devereux, Kipton Barros, Sergei Tretiak, Olexandr Isayev, Adrian E. Roitberg. Nature Communications, 2019

591 citations recorded by the source corpus.

Computational modeling of chemical and biological systems at atomic resolution is a crucial tool in the chemist’s toolset. The use of computer simulations requires a balance between cost and accuracy: quantum-mechanical methods provide high accuracy but are computationally expensive and scale poorly to large systems, while classical force fields are cheap and scalable, but lack transferability to new systems. Machine learning can be used to achieve the best of both approaches. Here we train a general-purpose neural network potential (ANI-1ccx) that approaches CCSD(T)/CBS accuracy on benchmarks for reaction thermochemistry, isomerization, and drug-like molecular torsions. This is achieved by training a network to DFT data then using transfer learning techniques to retrain on a dataset of gold standard QM calculations (CCSD(T)/CBS) that optimally spans chemical space. The resulting potential is broadly applicable to materials science, biology, and chemistry, and billions of times faster than CCSD(T)/CBS calculations. Computational modelling of chemical systems requires a balance between accuracy and computational cost. Here the authors use transfer learning to develop a general purpose neural network potential that approaches quantum-chemical accuracy for reaction thermochemistry, isomerization, and drug-like molecular torsions.

Subfields: Computational Theory and Mathematics, Materials Chemistry, Molecular Biology

View Paper: https://www.semanticscholar.org/paper/0010848c4044122dd79c12e3612091ac545a6e88

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computational Theory and Mathematics

In Materials Chemistry

In Molecular Biology

Citations in the corpus

Emerging materials intelligence ecosystems propelled by machine learning (2020, methodology): “Active-learning 156 and transfer-learning 157 approaches have also been used to continuously expand the domain of applicability of these potentials.”
Exploring chemical compound space with quantum-based machine learning (2019, background): “As such, these ∆learning approaches allow to invest the model complexity on the truly difficult aspects [37, 107, 108, 140–142].”
A foundation model for atomistic materials chemistry. (2023, unknown): “…reaching higher levels of electronic structure theory from a DFT baseline and beyond requires significantly less data than fitting to DFT itself ( 32, 117, 118 ) Finally, we may find that adding only a moderate number of additional configurations computed with essentially the same methodology will…”
OpenMM 8: Molecular Dynamics Simulation with Machine Learning Potentials (2023, unknown): “They support only a limited set of elements (seven for ANI-2x, four for ANI-1ccx), and do not support charged molecules.”
On-the-fly active learning of interpretable Bayesian force fields for atomistic rare events (2019, methodology): “The GP model obtained with active learning was subsequently mapped to a tabulated force ﬁ eld in order to rapidly evaluate forces on the entire "rand2 ” test set of Zhang et al. 10 , which consisted of 149 HEA structures with elements placed at random lattice sites.”
Quantum chemical accuracy from density functional approximations via machine learning (2020, background): “…has engendered many advances in the molecular sciences. atomistic These include the prediction of properties of systems across chemical space 1 – 26 , the construction of accurate force ﬁ elds 27 – 39 An ML scheme capable of realizing the aforementioned objective should satisfy several…”