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

• Published in: Nature communications Vol. 10; no. 1; pp. 2903 - 8
• Main Authors: Smith, Justin S., Nebgen, Benjamin T., Zubatyuk, Roman, Lubbers, Nicholas, Devereux, Christian, Barros, Kipton, Tretiak, Sergei, Isayev, Olexandr, Roitberg, Adrian E.
• Format: Journal Article
• Language: English
• Published: London Nature Publishing Group UK 01.07.2019; Nature Publishing Group; Nature Portfolio
• Subjects: 119/118; 639/638/440; 639/638/563/606; Accuracy; Benchmarking; Benchmarks; Computer applications; Computer Simulation; Data transfer (computers); Humanities and Social Sciences; Hydrocarbons - chemistry; INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CHEMISTRY; Isomerism; Isomerization; Learning algorithms; Machine Learning; Materials science; Mathematical models; multidisciplinary; Neural networks; Neural Networks, Computer; Organic chemistry; Science; Science (multidisciplinary); Thermochemistry; Thermodynamics; Transfer learning
• ISSN: 2041-1723; 2041-1723
• DOI: 10.1038/s41467-019-10827-4

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.