Comparison of methods for deriving atomic charges from the electrostatic potential and moments

Sigfridsson, Emma; Ryde, Ulf

doi:10.1002/(sici)1096-987x(199803)19:4<377::aid-jcc1>3.0.co;2-p

Cited by 337 publications

(117 citation statements)

References 34 publications

Supporting

Mentioning

117

Contrasting

Order By: Relevance

“…These calculations were performed with default values for all parameters (implying a water-like probe molecule) and a dielectric constant of 4. For the generation of the cavity, we used the optimised COSMO radii in Turbomole (1.30, 2.00, 1.83, 1.72, and 2.16 Å for H, C, N, O, and S, respectively [47] and 2.0 Å for the metals [48]). …”

Section: Qm Calculationsmentioning

confidence: 99%

Effect of Geometry Optimizations on QM-Cluster and QM/MM Studies of Reaction Energies in Proteins

Sumner

Söderhjelm

Ryde

2013

J. Chem. Theory Comput.

Self Cite

120

View full text Add to dashboard Cite

We have examined the effect of geometry optimisation on energies calculated with the quantum-mechanical (QM) cluster, the combined QM and molecular-mechanics (QM/MM), the big-QM approaches (very large single-point QM calculations taken from QM/MMoptimised structures, including all atoms within 4.5 Å of the minimal active site, all buried charged groups in the protein, and truncations moved at least three residues away from the active site). We study a simple proton-transfer reaction between His-79 and Cys-546 in the active site of [Ni,Fe] hydrogenase and optimise QM systems of 50 different sizes (56-362 atoms). Geometries optimised with QM/MM are stable and reliable, whereas QM-cluster optimisations give larger changes in the structures and sometimes lead to large distortions in the active site if some hydrogen-bond partners to the metal ligands are omitted. Keeping 2-3 atoms for each truncated residue (rather than one) fixed in the optimisation improves the results, but does not solve all problems for the QM-cluster optimisations. QM-cluster energies in vacuum and a continuum solvent are insensitive to the geometry optimisations with a mean absolute change upon the optimisations of only 4-7 kJ/mol. This shows that geometry optimisations do not improve the convergence of QM-cluster calculations -there is still a ~60 kJ/mol difference between calculations in which groups have been added to the QM system according to their distance to the active site or based on QM/MM free-energy components. QM/MM energies do not show such a difference, but they converge rather slowly with respect to the size of the QM system, although the convergence is improved by moving away truncations from the active site. The big-QM energies are stable over the 50 different optimised structures, 57±1 kJ/mol, although some smaller trends can be discerned. This shows that both QM-cluster geometries and energies should be interpreted with caution. Instead, we recommend QM/MM for geometry optimisations and energies calculated by the big-QM approach.Key Words: Quantum mechanical cluster calculations, QM/MM, geometry optimisation, density-functional theory, [Ni,Fe] hydrogenase. 2 IntroductionDuring the latest two decades, quantum mechanical (QM) calculations have been established as an important and powerful complement to experiments for the study of biochemical reactions [1,2,3,4,5,6,7,8]. In particular, it has been shown that QM calculations can give structures of active sites with an accuracy that is better than what is obtained in lowand medium-resolution crystal structures [9], and they can provide structures of transition states of putative enzyme reactions, which are hard to study experimentally. The calculations also give energies of the transition states and intermediates, which can be used to compare different putative reaction mechanisms.Two approaches are used for such calculations. In the QM cluster approach, the most important residues of the active site (typically 50-200 atoms) are cut out of the enzyme [1,2,3,4]. The geom...

show abstract

Section: Qm Calculationsmentioning

confidence: 99%

Effect of Geometry Optimizations on QM-Cluster and QM/MM Studies of Reaction Energies in Proteins

Sumner

Söderhjelm

Ryde

2013

J. Chem. Theory Comput.

Self Cite

120

View full text Add to dashboard Cite

show abstract

“…49 Although the atom-centered MEP-derived point charges provide a clear interpretation of the electrostatic properties and are computationally inexpensive, they can poorly reproduce the anisotropic electronic features (e.g., lone pairs, π-systems), 50,51 and also suffer from several technical difficulties. The optimized values of the point charges not only depend on the grid density and size, or the spatial orientation of the molecule relative to the Cartesian axes, 48,[52][53][54][55][56] they also can be inconsistent even across very similar molecules, at odds with the fundamental chemical concept of the transferability of atomic properties. Not only the MEP-fitted charges for atoms of a common functional group in chemically similar molecules may be very different, the charges obtained for the conformers of the same molecule often vary by more than one electron unit.…”

Section: Introductionmentioning

confidence: 98%

Genetic Algorithm Optimization of Point Charges in Force Field Development: Challenges and Insights

2015

View full text Add to dashboard Cite

Evolutionary methods, such as genetic algorithms (GAs), provide powerful tools for optimization of the force field parameters, especially in the case of simultaneous fitting of the force field terms against extensive reference data. However, GA fitting of the nonbonded interaction parameters that includes point charges has not been explored in the literature, likely due to numerous difficulties with even a simpler problem of the least-squares fitting of the atomic point charges against a reference molecular electrostatic potential (MEP), which often demonstrates an unusually high variation of the fitted charges on buried atoms. Here, we examine the performance of the GA approach for the least-squares MEP point charge fitting, and show that the GA optimizations suffer from a magnified version of the classical buried atom effect, producing highly scattered yet correlated solutions. This effect can be understood in terms of the linearly independent, natural coordinates of the MEP fitting problem defined by the eigenvectors of the least-squares sum Hessian matrix, which are also equivalent to the eigenvectors of the covariance matrix evaluated for the scattered GA solutions. GAs quickly converge with respect to the high-curvature coordinates defined by the eigenvectors related to the leading terms of the multipole expansion, but have difficulty converging with respect to the low-curvature coordinates that mostly depend on the buried atom charges. The performance of the evolutionary techniques dramatically improves when the point charge optimization is performed using the Hessian or covariance matrix eigenvectors, an approach with a significant potential for the evolutionary optimization of the fixed-charge biomolecular force fields.

show abstract

“…[7,9] CM5 was designed to address such drawbacks. [8] Like CM5, our ACA charge model is local-by-design, thus averting the problem of artificial long-range effects.…”

Section: Mskmentioning

confidence: 99%

“…However, MSK suffers from basis set sensitivity, particularly for "buried atoms" located inside large molecules. [7][8][9] Charge model 5 (CM5) [8] is an extension of Hirshfeld analysis, [10] with additional parametrization in order to approximately reproduce ab initio and experimental dipoles of 614 gas-phase dipoles. Unlike MSK, Hirshfeld and CM5 are nearly independent of basis set.…”

mentioning

confidence: 99%

“…However, the far-field potential is relatively insensitive to the partial charge assignments of internal atoms. [7][8][9] Because MSK performs its charge assignments according to global (rather than local) criteria, the assigned charges can deviate significantly from the local charge density field.Another related difficulty of MSK is that it exhibits a noticeable basis set dependence. [7,9] CM5 was designed to address such drawbacks.…”

mentioning

confidence: 99%

See 1 more Smart Citation

Discovering a Transferable Charge Assignment Model Using Machine Learning

Sifain¹,

Lubbers²,

Nebgen³

et al. 2018

Preprint

View full text Add to dashboard Cite

Partial atomic charge assignment is of immense practical value to force field parametrization, molecular docking, and cheminformatics. Machine learning has emerged as a powerful tool for modeling chemistry at unprecedented computational speeds given ground-truth values, but for the task of charge assignment, the choice of ground-truth may not be obvious. In this letter, we use machine learning to discover a charge model by training a neural network to molecular dipole moments using a large, diverse set of CHNO molecular conformations. The new model, called Affordable Charge Assignment (ACA), is computationally inexpensive and predicts dipoles of out-of-sample molecules accurately. Furthermore, dipole-inferred ACA charges are transferable to dipole and even quadrupole moments of much larger molecules than those used for training. We apply ACA to long dynamical trajectories of biomolecules and successfully produce their infrared spectra. Additionally, we compare ACA with existing charge models and find that ACA assigns similar charges to Charge Model 5, but with a greatly reduced computational cost. Graphical TOC Entry Extensibility Tests Molecular Size Training DatasetKeywords machine learning, neural networks, quantum chemisty 2 Electrostatic interactions contribute strongly to the forces within and between molecules.These interactions depend on the charge density field ⇢(r), which is computationally demanding to compute. Simplified models of the charge density, such as atom-centered monopoles, are commonly employed. These partial atomic charges result in faster computation as well as provide a qualitative understanding of the underlying chemistry. [1][2][3][4] However, the decomposition of charge density into atomic charges is, by itself, an ambiguous task. Additional principles are necessary to make the charge assignment task well-defined. Here we show that a Machine Learning model, trained only on the dipole moments of small molecules, discovers a charge model that is transferable to quadrupole predictions and extensible to much larger molecules.Existing popular charge models have also been designed to reproduce observables of the electrostatic potential. The Merz-Singh-Kollman (MSK) [5,6] charge model exactly replicates the dipole moment and approximates the electrostatic potential on many points surrounding the molecule, resulting in high-quality electrostatic properties exterior to the molecule. However, MSK suffers from basis set sensitivity, particularly for "buried atoms" located inside large molecules. [7][8][9] Charge model 5 (CM5) [8] is an extension of Hirshfeld analysis, [10] with additional parametrization in order to approximately reproduce ab initio and experimental dipoles of 614 gas-phase dipoles. Unlike MSK, Hirshfeld and CM5 are nearly independent of basis set. [9] This insensitivity allows CM5 to use a single set of model parameters. The corresponding tradeoff is that its charges do not reproduce electrostatic fields as well as MSK.A limitation of these conventional charge models is...

show abstract

Comparison of methods for deriving atomic charges from the electrostatic potential and moments

Cited by 337 publications

References 34 publications

Effect of Geometry Optimizations on QM-Cluster and QM/MM Studies of Reaction Energies in Proteins

Effect of Geometry Optimizations on QM-Cluster and QM/MM Studies of Reaction Energies in Proteins

Genetic Algorithm Optimization of Point Charges in Force Field Development: Challenges and Insights

Discovering a Transferable Charge Assignment Model Using Machine Learning

Contact Info

Product

Resources

About