Building Force Fields: An Automatic, Systematic, and Reproducible Approach

Wang, Lee‐Ping; Martı́nez, Todd J.; Pande, Vijay S.

doi:10.1021/jz500737m

Cited by 548 publications

(802 citation statements)

References 33 publications

Supporting

Mentioning

784

Contrasting

Unclassified

Order By: Relevance

“…The first vector s⃗1, along which the data does not vary, numerically corresponds to the normalized vector u⃗1 that defines the total charge and is determined by the stoichiometry of the molecule: (20)where u⃗1 = (nX nH) and q⃗ = (qX qH). The second vector s⃗2, that is, the vector along which the data show a significant variation, numerically corresponds to a normalized vector u⃗2 = (nH −nX), also determined by the stoichiometry.…”

Section: Covariance Matrix Analysis Of Ga Resultsmentioning

confidence: 99%

“…However, simultaneous fitting of several parameters describing intermolecular interactions (point charges, Lennard-Jones parameters, and in the case of polarizable force fields, atomic polarizabilities) may significantly improve the accuracy of force field description. 20,21 These simultaneous optimizations of different force field terms can take advantage of extensive training sets that can be easily generated using electronic structure calculations and may include data on the intermolecular interaction energies. [22][23][24][25][26] Moreover, in this approach the fitted interaction energy would implicitly include the polarization effects, even staying within the fixed point-charge force field framework.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

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

Section: Covariance Matrix Analysis Of Ga Resultsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

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

2015

View full text Add to dashboard Cite

show abstract

“…14 The bilayers were simulated under NPT conditions (i.e., no applied surface tension) using a Langevin thermostat with a friction coefficient of 1 parameters for DPPC and POPC, while the water molecules were represented using the TIP3P, 10 TIP3P-FB, and TIP4P-FB models. 9 Deuterium order parameters were calculated using the MEMPLUGIN 15 extension of VMD 1.9.2. 16 Electron density and neutron scattering profiles were calculated using the Density Profile extension 17 of VMD and transformed into reciprocal space using the theory described by Benz et al citeBenz2005 and atomic parameters of the SimToExp code.…”

Section: Lipid Bilayer Simulationsmentioning

confidence: 99%

The CHARMM36 Force Field for Lipids Can Be Used With More Accurate Water Models

Sajadi¹,

Rowley²

2018

Preprint

View full text Add to dashboard Cite

The CHARMM36 force field for lipids is widely used in simulations of lipid bilayers. The CHARMM family of force fields were developed for use with the TIP3P water model. This water model has an anomalously high dielectric constant and low viscosity, which limits its accuracy in the calculation of quantities like permeability coefficients. The TIP3P-FB and TIP4P-FB water models are more accurate in terms of the dielectric constant and transport properties, which could allow more accurate simulations of systems containing water and lipids. To test whether the CHARMM36 lipid force field is compatible with the TIP3P-FB and TIP4P-FB water models, we have performed simulations of DPPC and POPC bilayers. The calculated headgroup area, compressibility, order parameters, and X-ray form factors are in good agreement with the experimental values, indicating that these improved water models can be used with the CHARMM36 lipid force field without modification. The water permeability predicted by these models is significantly different; the TIP3P-model diffusion in solution and at the lipid-water interface is anomalously fast due to the spuriously low viscosity of TIP3P-model water, but the PMF of permeation is higher for the TIP3P-FB and a TIP4P-FB models due to their high excess chemical potentials.

show abstract

“…can also be used. [22][23] There has also been a growing trend to parameterise force fields against global properties such as solvation and transfer free energies. 16-18, 21, 24 The difficulty of parameterizing against time dependent or fluctuation based properties is that they require the sampling of multiple configurations (or states of the system) as well as an analysis of convergence.…”

Section: Force Fieldsmentioning

confidence: 99%

“…Specifically, if the KSSE value is greater than the target error for texcl = 0, and the KSSE value is less than the target error for some value of texcl > 0, teq is taken to be the first value of texcl such that KSSE < Etarget i.e. : (22) Note that since the KSSE values have the same units as the time series property being averages, Etarget represents the error tolerance relevant to a given application. Data points within the equilibration region of a time series (t < teq) can then be excluded from all ensemble average calculations to avoid bias from the initial conditions.…”

Section: Estimation Of Equilibration Timementioning

confidence: 99%

Improving the accuracy of molecular dynamics simulations: parameterisation of interaction potentials for small molecules

Stroet¹

View full text Add to dashboard Cite

The ability to accurately describe molecular systems with classical molecular dynamics (MD) is critically dependent on an understanding of the sources of error in the design, evaluation and analysis of the underlying model. One significant source of error is in the degree to which the equations and parameters used to represent atomic interaction can reproduce the experimental properties relevant to a particular application. While highly optimised and well-validated interaction parameters are available for common biomolecules (amino acids, sugars, lipids etc.), this is not the case for heterogeneous small molecules such as co-factors, substrates and drugs. The enormity of the chemical space covered by small molecules necessitates the development of automated approaches to parameterising and validating interaction parameters. Several tools are available that can be used to generate interaction parameters for small molecules compatible with a particular existing (bio)molecular force field. Despite their popularity, many are poorly validated, while some have been shown to be inappropriate for many applications. In other cases, validation studies have demonstrated reasonable performance, however, they also suggest significant improvements can be made. One such tool is the Automated Topology Builder (ATB, http://atb.uq.edu.au/) which provides interaction parameters for small molecules compatible with the GROMOS biomolecular force field.As part of this work, a fully automated protocol for the calculation of solvation energy by thermodynamic integration has been developed and applied to the large-scale validation of the ATB against experimental solvation data in water and hexane. It is shown that the largest errors are due to the Lennard-Jones parameters used for functional groups not present in common biomolecules. Other sources of error included the atomic charges assigned by the ATB for the united atom carbons and incompatibilities between GROMOS Lennard-Jones parameters and the ATB charge model. Significant sources of error were also identified in the evaluation of MD models, both for the calculation of solvation energy as well as for MD simulations of lipid membranes. In particular, the use of multiple-time-step algorithms as a time-saving technique. For the calculation of solvation energy by TI, the twin-range cutoff scheme-commonly used in simulations with the GROMOS force field-was found to introduce a systematic error of about 1 kJ/mol. While for lipid systems, various III changes made to the integration algorithm of the GROMACS simulation code were found to significantly alter the properties of lipid membranes when simulated with identical force fields.Finally, a method for parameterising transferrable interaction potentials with respect to a wide range of target properties is presented. The method is based on an analysis of surfaces corresponding to the difference between calculated and target data as a function of alternative combinations of parameters.The consideration of surfaces in parameter space...

show abstract

Building Force Fields: An Automatic, Systematic, and Reproducible Approach

Cited by 548 publications

References 33 publications

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

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

The CHARMM36 Force Field for Lipids Can Be Used With More Accurate Water Models

Improving the accuracy of molecular dynamics simulations: parameterisation of interaction potentials for small molecules

Contact Info

Product

Resources

About