Transferable Dynamic Molecular Charge Assignment Using Deep Neural Networks

Journal Article · Tue Jul 31 00:00:00 EDT 2018 · Journal of Chemical Theory and Computation

DOI:https://doi.org/10.1021/acs.jctc.8b00524· OSTI ID:1467336

Nebgen, Benjamin Tyler  ^[1]; Lubbers, Nicholas Edward  ^[1]; Smith, Justin Steven  ^[2]; Sifain, Andrew E. ^[3]; Lokhov, Andrey  ^[1]; Isayev, Olexandr ^[4]; Roitberg, Adrian ^[5]; Barros, Kipton Marcos  ^[1]; Tretiak, Sergei  ^[1]

---

1. Los Alamos National Lab. (LANL), Los Alamos, NM (United States)
2. Los Alamos National Lab. (LANL), Los Alamos, NM (United States); Univ. of Florida, Gainesville, FL (United States)
3. Los Alamos National Lab. (LANL), Los Alamos, NM (United States); Univ. of Southern California, Los Angeles, CA (United States)
4. Univ. of North Carolina, Chapel Hill, NC (United States)
5. Univ. of Florida, Gainesville, FL (United States)

We report that the ability to accurately and efficiently compute quantum-mechanical partial atomistic charges has many practical applications, such as calculations of IR spectra, analysis of chemical bonding, and classical force field parametrization. Machine learning (ML) techniques provide a possible avenue for the efficient prediction of atomic partial charges. Modern ML advances in the prediction of molecular energies [i.e., the hierarchical interacting particle neural network (HIP-NN)] has provided the necessary model framework and architecture to predict transferable, extensible, and conformationally dynamic atomic partial charges based on reference density functional theory (DFT) simulations. Utilizing HIP-NN, we show that ML charge prediction can be highly accurate over a wide range of molecules (both small and large) across a variety of charge partitioning schemes such as the Hirshfeld, CM5, MSK, and NBO methods. To demonstrate transferability and size extensibility, we compare ML results with reference DFT calculations on the COMP6 benchmark, achieving errors of 0.004e^– (elementary charge). This is remarkable since this benchmark contains two proteins that are multiple times larger than the largest molecules in the training set. An application of our atomic charge predictions on nonequilibrium geometries is the generation of IR spectra for organic molecules from dynamical trajectories on a variety of organic molecules, which show good agreement with calculated IR spectra with reference method. Critically, HIP-NN charge predictions are many orders of magnitude faster than direct DFT calculations. Lastly, these combined results provide further evidence that ML (specifically HIP-NN) provides a pathway to greatly increase the range of feasible simulations while retaining quantum-level accuracy.

• Research Organization: Los Alamos National Laboratory (LANL), Los Alamos, NM (United States)
• Sponsoring Organization: USDOE National Nuclear Security Administration (NNSA); USDOE Laboratory Directed Research and Development (LDRD) Program
• Grant/Contract Number: AC52-06NA25396
• OSTI ID: 1467336
• Report Number(s): LA-UR-18-22005
• Journal Information: Journal of Chemical Theory and Computation, Vol. 14, Issue 9; ISSN 1549-9618
• Publisher: American Chemical Society
• Country of Publication: United States
• Language: English

Related Subjects

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CHEMISTRY
97 MATHEMATICS AND COMPUTING