Automation of <scp>AMOEBA</scp> polarizable force field for small molecules: Poltype 2

Walker, Brandon; Liu, Chengwen; Wait, E.; Ren, Pengyu

doi:10.1002/jcc.26954

Cited by 35 publications

(51 citation statements)

References 46 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…More details about the functional form and parametrization of AMOEBA can be found in reference. 33 Electrostatics and many-body polarization long-range interactions are fully included through the use of the Smooth Particle Mesh Ewald approach 34,35 that allows for periodic boundary conditions simulations. Besides water 28 , AMOEBA is a general force field available for many solvent 36 , ions, 37,38 proteins 39 and nucleic acids 40 biomolecular simulations.…”

Section: The Amoeba Polarizable Force Fieldmentioning

confidence: 99%

“…We withdrew molecules from the datasets that have chemical elements not available in ANI-2x. This led us to a total of 39 molecules solvated in water from reference 33 , 20 molecules solvated in toluene, 6 in acetonitrile and 6 in DMSO from J. Essex et al 58 . All the systems were prepared following standard equilibration protocol: after a geometry optimization, they were progressively heated up to 300 K in NVT and then equilibrated for 1 ns in the NPT ensemble at the same temperature and 1 atmosphere.…”

Section: Computational Detailsmentioning

confidence: 99%

See 1 more Smart Citation

Scalable Hybrid Deep Neural Networks/Polarizable Potentials Biomolecular Simulations including long-range effects

Inizan¹,

Plé²,

Adjoua³

et al. 2022

Preprint

Self Cite

View full text Add to dashboard Cite

We present Deep-HP, a scalable multi-GPUs simulation platform dedicated to machine learning potentials. As part of the Tinker-HP molecular dynamics (MD) package, it allows users to efficiently use their Pytorch/TensorFlow Deep Neural Networks (DNNs) models (ANI, DeePMD ...) to perform MD simulations up to million-atoms systems. Thanks to its MPI multi-GPUs setup inherited from Tinker-HP, Deep-HP extends DNNs simulation timescales while offering the possibility of coupling them to any classical (FFs) and many-body polarizable (PFFs) force fields. Towards biophysical applications, we present a novel hybrid molecular dynamics simulation protocol coupling the AMOEBA PFF to the ANI-2x DNN where solvent-solvent and solvent-solute interactions are computed using AMOEBA while the solute-solute interactions are computed by ANI-2x. The strategy allows for the explicit inclusion of physical long-range effects such as multipolar electrostatics and many-body polarization via efficient periodic boundary conditions. The DNNs/PFFs partition can be user-defined allowing for hybrid simulations to include biosimulation key ingredients such as polarizable solvents, and polarizable counter-ions. To reduce the DNN computational performance gap compared to FFs, we rely on extensive multiple time stepping and focus on the models contributions to low frequency modes of nuclear forces. Therefore, the approach primarily evaluates the AMOEBA forces while including the ANI-2x ones only via a correction step resulting in an overall 20-fold acceleration over standard Velocity Verlet integration. Thanks to this approach, accessing µs simulations with hybrid DNN/PFF models becomes possible. This allowed us to evaluate the accuracy of the ANI-2x/AMOEBA model by calculating the solvation free energies of various ligands in four different solvents, and the binding free energies of 13 challenging ligand-protein complexes. The hybrid approach is shown to be accurate even for charged ligands. Overall, the hybrid approach can outperform AMOEBA for solvation free energies in non-aqueous solvent and ligand binding free energies. This opens further perspectives for large scale hybrid DNN/PFF simulations towards biophysical applications. The Deep-HP platform will also help to foster the development of next generation of DNN models able to capture both short-range quantum accuracy and long-range physical/many-body effects.

show abstract

Section: The Amoeba Polarizable Force Fieldmentioning

confidence: 99%

Section: Computational Detailsmentioning

confidence: 99%

Scalable Hybrid Deep Neural Networks/Polarizable Potentials Biomolecular Simulations including long-range effects

Inizan¹,

Plé²,

Adjoua³

et al. 2022

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

“…AMOEBA (Ponder et al, 2010;Ren et al, 2011) parameters were generated using the PolType2 (Wu et al, 2012;Walker et al, 2022) automatic parameterization program on SDF files obtained from PubChem (Kim et al, 2021). Local optimization of coordinates and lattice parameters of each experimental structure to an energetic convergence criterion of 0.1 kcal mol À1 A ˚À1 (1 kcal mol À1 = 4.184 kJ mol À1 ) was performed according to AMOEBA using Force Field X.…”

Section: Data For Evaluating the Pac Algorithmmentioning

confidence: 99%

Progressive alignment of crystals: reproducible and efficient assessment of crystal structure similarity

Nessler¹,

Okada²,

Hermon³

et al. 2022

J Appl Crystallogr

View full text Add to dashboard Cite

During in silico crystal structure prediction of organic molecules, millions of candidate structures are often generated. These candidates must be compared to remove duplicates prior to further analysis (e.g. optimization with electronic structure methods) and ultimately compared with structures determined experimentally. The agreement of predicted and experimental structures forms the basis of evaluating the results from the Cambridge Crystallographic Data Centre (CCDC) blind assessment of crystal structure prediction, which further motivates the pursuit of rigorous alignments. Evaluating crystal structure packings using coordinate root-mean-square deviation (RMSD) for N molecules (or N asymmetric units) in a reproducible manner requires metrics to describe the shape of the compared molecular clusters to account for alternative approaches used to prioritize selection of molecules. Described here is a flexible algorithm called Progressive Alignment of Crystals (PAC) to evaluate crystal packing similarity using coordinate RMSD and introducing the radius of gyration (R g) as a metric to quantify the shape of the superimposed clusters. It is shown that the absence of metrics to describe cluster shape adds ambiguity to the results of the CCDC blind assessments because it is not possible to determine whether the superposition algorithm has prioritized tightly packed molecular clusters (i.e. to minimize R g) or prioritized reduced RMSD (i.e. via possibly elongated clusters with relatively larger R g). For example, it is shown that when the PAC algorithm described here uses single linkage to prioritize molecules for inclusion in the superimposed clusters, the results are nearly identical to those calculated by the widely used program COMPACK. However, the lower R g values obtained by the use of average linkage are favored for molecule prioritization because the resulting RMSDs more equally reflect the importance of packing along each dimension. It is shown that the PAC algorithm is faster than COMPACK when using a single process and its utility for biomolecular crystals is demonstrated. Finally, parallel scaling up to 64 processes in the open-source code Force Field X is presented.

show abstract

“…AMOEBA-IL follows the AMOEBA procedure to parameterize the bonded terms [29]. The non-bonded terms are initially parameterized by comparing individual inter-molecular interaction contributions of gas phase dimers to QM energy decomposition analysis (EDA) reference values as described in [14].…”

Section: Parameterization Of Imidazolium-based Cations For Amoeba-ilmentioning

confidence: 99%

Development of Imidazolium-Based Parameters for AMOEBA–IL

CERVANTES

Cisneros

2022

Preprint

View full text Add to dashboard Cite

A new approach for the efficient parametrization of the polarizable ionic liquid potential AMOEBA–IL, and its application to develop parameters for imidazolium– based cations is presented. The new approach relies on the development of parameters for fragments that can be transferred to generate new molecules. The parametrization uses the original AMOEBA–IL parametrization approach, in- cluding the use of Gaussian Electrostatic Model-Distributed Multipoles (GEM-DM) for the permanent multipoles and approximation of the Van der Waals parameters using quantum mechanics energy decomposition analysis (QM-EDA) data. Based on this, functional groups of the selected initial structures are employed as building blocks to develop parameters for new imidazolium-based cations (symmetric or asymmetric) with longer alkyl chains. The parameters obtained with this proposed method were compared with inter-molecular interactions from QM references via energy decomposition analysis using Symmetry Adapted Perturbation Theory (SAPT) and counterpoise corrected total inter-molecular interactions. The validation of the new parametrized cations was carried out by running molecular dynamics simulations on a series of imidazolium-based IL’s with different anions to compare selected thermodynamic and transport properties, including density ρ, enthalpy of vaporization ∆H vap , radial distribution function g(r), and diffusion coefficients D±, with experimental data. Overall, the calculated gas–phase and bulk properties show agreement with the reference data. The new procedure provides a straightforward approach to generate the required AMOEBA–IL parameters for any imidazolium-based cation.

show abstract

Automation of AMOEBA polarizable force field for small molecules: Poltype 2

Cited by 35 publications

References 46 publications

Scalable Hybrid Deep Neural Networks/Polarizable Potentials Biomolecular Simulations including long-range effects

Scalable Hybrid Deep Neural Networks/Polarizable Potentials Biomolecular Simulations including long-range effects

Progressive alignment of crystals: reproducible and efficient assessment of crystal structure similarity

Development of Imidazolium-Based Parameters for AMOEBA–IL

Contact Info

Product

Resources

About