Molecular dynamics simulations using the drude polarizable force field on GPUs with OpenMM: Implementation,validation, and benchmarks

Presented is the implementation of the Drude force field in the open‐source OpenMM simulation package allowing for access to graphical processing unit (GPU) hardware. In the Drude model, electronic degrees of freedom are represented by negatively charged particles attached to their parent atoms via harmonic springs, such that extra computational overhead comes from these additional particles and virtual sites representing lone pairs on electronegative atoms, as well as the associated thermostat and integration algorithms. This leads to an approximately fourfold increase in computational demand over additive force fields. However, by making the Drude model accessible to consumer‐grade desktop GPU hardware it will be possible to perform simulations of one microsecond or more in less than a month, indicating that the barrier to employ polarizable models has largely been removed such that polarizable simulations with the classical Drude model are readily accessible and practical.     

Absolute free energies of biomolecules from unperturbed ensembles

Computing the absolute free energy of a macromolecule's structural state, F, is a challenging problem of high relevance. This study presents a method that computes F using only information from an unperturbed simulation of the macromolecule in the relevant conformational state, ensemble, and environment. Absolute free energies produced by this method, dubbed V aluation of L ocal C onfiguration I ntegral with D ynamics (VALOCIDY), enable comparison of alternative states. For example, comparing explicitly solvated and vaporous states of amino acid side‐chain analogs produces solvation free energies in good agreement with experiments. Also, comparisons between alternative conformational states of model heptapeptides (including the unfolded state) produce free energy differences in agreement with data from μs molecular‐dynamics simulations and experimental propensities. The potential of using VALOCIDY in computational protein design is explored via a small design problem of stabilizing a β‐turn structure. When VALOCIDY‐based estimation of folding free energy is used as the design metric, the resulting sequence folds into the desired structure within the atomistic force field used in design. The VALOCIDY‐based approach also recognizes the distinct status of the native sequence regardless of minor details of the starting template structure, in stark contrast with a traditional fixed‐backbone approach. © 2013 Wiley Periodicals, Inc.     

Linear-scaling soft-core scheme for alchemical free energy calculations

Alchemical free energy calculations involving the removal or insertion of atoms into condensed phase systems generally make use of soft-core scaling of nonbonded interactions, designed to circumvent numerical instabilities that arise from weakly interacting "hard" atoms in close proximity. Current methods model soft-core atoms by introducing a nonlinear dependence between the shape of the interaction potential and the strength of the interaction. In this article, we propose a soft-core method that avoids introducing such a nonlinear dependence, through the application of a smooth flattening of the potential energy only in a region that is energetically accessible under normal conditions. We discuss the benefits that this entails and explore a selection of applications, including enhanced methods for the estimation of free energy differences and for the automated optimization of the placement of intermediate states in multistage alchemical calculations.     

Comparison of thermodynamic integration and Bennett acceptance ratio for calculating relative protein‐ligand binding free energies

The performances of Bennett's acceptance ratio method and thermodynamic integration (TI) for the calculation of free energy differences in protein simulations are compared. For the latter, the standard trapezoidal rule, Simpson's rule, and Clenshaw‐Curtis integration are used as numerical integration methods. We evaluate the influence of the number and definition of intermediate states on the precision, accuracy, and efficiency of the free energy calculations. Our results show that non‐equidistantly spaced intermediate states are in some cases beneficial for the TI methods. Using several combinations of softness parameters and the λ power dependence, it is shown that these benefits are strongly dependent on the shape of the integrand. Although TI is more user‐friendly due to its simplicity, it was found that Bennett's acceptance ratio method is the more efficient method. It is also the least dependent on the choice of the intermediate states, making it more robust than TI. © 2013 Wiley Periodicals, Inc.     

Fast and accurate predictions of binding free energies using MM‐PBSA and MM‐GBSA

In the drug discovery process, accurate methods of computing the affinity of small molecules with a biological target are strongly needed. This is particularly true for molecular docking and virtual screening methods, which use approximated scoring functions and struggle in estimating binding energies in correlation with experimental values. Among the various methods, MM‐PBSA and MM‐GBSA are emerging as useful and effective approaches. Although these methods are typically applied to large collections of equilibrated structures of protein‐ligand complexes sampled during molecular dynamics in water, the possibility to reliably estimate ligand affinity using a single energy‐minimized structure and implicit solvation models has not been explored in sufficient detail. Herein, we thoroughly investigate this hypothesis by comparing different methods for the generation of protein‐ligand complexes and diverse methods for free energy prediction for their ability to correlate with experimental values. The methods were tested on a series of structurally diverse inhibitors of Plasmodium falciparum DHFR with known binding mode and measured affinities. The results showed that correlations between MM‐PBSA or MM‐GBSA binding free energies with experimental affinities were in most cases excellent. Importantly, we found that correlations obtained with the use of a single protein‐ligand minimized structure and with implicit solvation models were similar to those obtained after averaging over multiple MD snapshots with explicit water molecules, with consequent save of computing time without loss of accuracy. When applied to a virtual screening experiment, such an approach proved to discriminate between true binders and decoy molecules and yielded significantly better enrichment curves. © 2009 Wiley Periodicals, Inc. J Comput Chem, 2010     

Net charge changes in the calculation of relative ligand‐binding free energies via classical atomistic molecular dynamics simulation

The calculation of binding free energies of charged species to a target molecule is a frequently encountered problem in molecular dynamics studies of (bio‐)chemical thermodynamics. Many important endogenous receptor‐binding molecules, enzyme substrates, or drug molecules have a nonzero net charge. Absolute binding free energies, as well as binding free energies relative to another molecule with a different net charge will be affected by artifacts due to the used effective electrostatic interaction function and associated parameters (e.g., size of the computational box). In the present study, charging contributions to binding free energies of small oligoatomic ions to a series of model host cavities functionalized with different chemical groups are calculated with classical atomistic molecular dynamics simulation. Electrostatic interactions are treated using a lattice‐summation scheme or a cutoff‐truncation scheme with Barker–Watts reaction‐field correction, and the simulations are conducted in boxes of different edge lengths. It is illustrated that the charging free energies of the guest molecules in water and in the host strongly depend on the applied methodology and that neglect of correction terms for the artifacts introduced by the finite size of the simulated system and the use of an effective electrostatic interaction function considerably impairs the thermodynamic interpretation of guest‐host interactions. Application of correction terms for the various artifacts yields consistent results for the charging contribution to binding free energies and is thus a prerequisite for the valid interpretation or prediction of experimental data via molecular dynamics simulation. Analysis and correction of electrostatic artifacts according to the scheme proposed in the present study should therefore be considered an integral part of careful free‐energy calculation studies if changes in the net charge are involved. © 2013 The Authors Journal of Computational Chemistry Published by Wiley Periodicals, Inc.     

Calculation of the water-cyclohexane transfer free energies of neutral amino acid side-chain analogs using the OPLS all-atom force field

We calculated the free energy of solvation of the neutral analogs of 18 amino acid side-chains (not including glycine and proline) using the OPLS all-atom force field in TIP4P water, SPC water, and cyclohexane by molecular dynamics simulation and thermodynamic integration. The average unsigned errors in the free energies of solvation in TIP4P, SPC, and cyclohexane are 4.4, 4.9, and 2.1 kJ/mol respectively. Most of the calculated hydration free energies are not favorable enough compared to experiment. The largest errors are found for tryptophan, histidine, glutamic acid, and glutamine. The average unsigned errors in the free energy of transfer from TIP4P to cyclohexane and from SPC to cyclohexane are 4.0 and 4.1 kJ/mol, respectively. The largest errors, of more than 7.5 kJ/mol, are found for histidine, glutamine, and glutamatic acid.     

Minimum MD simulation length required to achieve reliable results in free energy perturbation calculations: Case study of relative binding free energies of fructose‐1,6‐bisphosphatase inhibitors

In an attempt to establish the criteria for the length of simulation to achieve the desired convergence of free energy calculations, two studies were carried out on chosen complexes of FBPase‐AMP mimics. Calculations were performed for varied length of simulations and for different starting configurations using both conventional‐ and QM/MM‐FEP methods. The results demonstrate that for small perturbations, 1248 ps simulation time could be regarded a reasonable yardstick to achieve convergence of the results. As the simulation time is extended, the errors associated with free energy calculations also gradually tapers off. Moreover, when starting the simulation from different initial configurations of the systems, the results are not changed significantly, when performed for 1248 ps. This study carried on FBPase‐AMP mimics corroborates well with our previous successful demonstration of requirement of simulation time for solvation studies, both by conventional and ab initio FEP. The establishment of aforementioned criteria of simulation length serves a useful benchmark in drug design efforts using FEP methodologies, to draw a meaningful and unequivocal conclusion. © 2011 Wiley Periodicals, Inc. J Comput Chem, 2011     

Computation of the binding affinities of catechol‐O‐methyltransferase inhibitors: Multisubstate relative free energy calculations

Alchemical free energy simulations are amongst the most accurate techniques for the computation of the free energy changes associated with noncovalent protein–ligand interactions. A procedure is presented to estimate the relative binding free energies of several ligands to the same protein target where multiple, low‐energy configurational substates might coexist, as opposed to one unique structure. The contributions of all individual substates were estimated, explicitly, with the free energy perturbation method, and combined in a rigorous fashion to compute the overall relative binding free energies and dissociation constants. It is shown that, unless the most stable bound forms are known a priori, inaccurate results may be obtained if the contributions of multiple substates are ignored. The method was applied to study the complex formed between human catechol‐O‐methyltransferase and BIA 9‐1067, a newly developed tight‐binding inhibitor that is currently under clinical evaluation for the therapy of Parkinson's disease. Our results reveal an exceptionally high‐binding affinity (K_d in subpicomolar range) and provide insightful clues on the interactions and mechanism of inhibition. The inhibitor is, itself, a slowly reacting substrate of the target enzyme and is released from the complex in the form of O‐methylated product. By comparing the experimental catalytic rate (k_cat) and the estimated dissociation rate (k_off) constants of the enzyme‐inhibitor complex, one can conclude that the observed inhibition potency (K_i) is primarily dependent on the catalytic rate constant of the inhibitor's O‐methylation, rather than the rate constant of dissociation of the complex. © 2012 Wiley Periodicals, Inc.     

On the accurate estimation of free energies using the jarzynski equality

The Jarzynski equality is one of the most widely celebrated and scrutinized nonequilibrium work theorems, relating free energy to the external work performed in nonequilibrium transitions. In practice, the required ensemble average of the Boltzmann weights of infinite nonequilibrium transitions is estimated as a finite sample average, resulting in the so-called Jarzynski estimator, . Alternatively, the second-order approximation of the Jarzynski equality, though seldom invoked, is exact for Gaussian distributions and gives rise to the Fluctuation-Dissipation estimator . Here we derive the parametric maximum-likelihood estimator (MLE) of the free energy considering unidirectional work distributions belonging to Gaussian or Gamma families, and compare this estimator to . We further consider bidirectional work distributions belonging to the same families, and compare the corresponding bidirectional to the Bennett acceptance ratio () estimator. We show that, for Gaussian unidirectional work distributions, is in fact the parametric MLE of the free energy, and as such, the most efficient estimator for this statistical family. We observe that and perform better than and , for unidirectional and bidirectional distributions, respectively. These results illustrate that the characterization of the underlying work distribution permits an optimal use of the Jarzynski equality. © 2018 Wiley Periodicals, Inc.     

Efficient calculation of relative binding free energies by umbrella sampling perturbation

An important task of biomolecular simulation is the calculation of relative binding free energies upon chemical modification of partner molecules in a biomolecular complex. The potential of mean force (PMF) along a reaction coordinate for association or dissociation of the complex can be used to estimate binding affinities. A free energy perturbation approach, termed umbrella sampling (US) perturbation, has been designed that allows an efficient calculation of the change of the PMF upon modification of a binding partner based on the trajectories obtained for the wild type reference complex. The approach was tested on the interaction of modified water molecules in aqueous solution and applied to in silico alanine scanning of a peptide‐protein complex. For the water interaction test case, excellent agreement with an explicit PMF calculation for each modification was obtained as long as no long range electrostatic perturbations were considered. For the alanine scanning, the experimentally determined ranking and binding affinity changes upon alanine substitutions could be reproduced within 0.1–2.0 kcal/mol. In addition, good agreement with explicitly calculated PMFs was obtained mostly within the sampling uncertainty. The combined US and perturbation approach yields, under the condition of sufficiently small system modifications, rigorously derived changes in free energy and is applicable to any PMF calculation. © 2014 Wiley Periodicals, Inc.     

Calculation of sequence‐dependent free energies of hydration of dipeptides formed by alanine and glycine

The relative free energies of hydration of the dipeptides glycylalanine and alanyl‐glycine in their naturally occurring form have been calculated both for the zwitterionic and protonated species. Emphasis was laid on comparisons between the conventional cutoff method and the Particle Mesh Ewald method to account for possible differences in electrostatic contributions to the free energy. Furthermore, the convergence behavior of the total free energy and its individual contributions were examined. The results, obtained by means of the thermodynamic integration technique as implemented in the free energy module of the AMBER program suite, suggest that in aqueous solution glycylalanine is more stable than alanylglycine by 2.7 kcal/mol in the zwitterionic form and by 3.5 kcal/mol in the protonated form. © 2001 John Wiley & Sons, Inc. J Comput Chem 22: 846–860, 2001     

Calculation of the relative free energy of oxidation of azurin at pH 5 and pH 9

Free energy calculations are described for the small copper‐containing redox protein Azurin from Pseudomonas aeruginosa. A thermodynamic cycle connecting the reduced and oxidized states at pH 5 and pH 9 is considered, allowing for an assessment of convergence in terms of hysteresis and cycle closure. Previously published thermodynamic integration (TI) data is compared to Hamiltonian replica exchange TI (RE‐TI) simulations using different simulation setups. The effects of varying simulation length, initial structure, position restraints on particular atoms, and the strength of temperature coupling are studied. Although the overall simulation times are too short to observe an experimentally described peptide plane rotation, it is found that RE‐TI simulations do stimulate the distribution of conformational changes over the relevant values of the TI coupling parameter λ. This results in significantly improved values for hysteresis and cycle closure when compared to regular TI. © 2012 Wiley Periodicals, Inc.     

Accuracy of free energies of hydration using CM1 and CM3 atomic charges

Absolute free energies of hydration (DeltaGhyd) have been computed for 25 diverse organic molecules using partial atomic charges derived from AM1 and PM3 wave functions via the CM1 and CM3 procedures of Cramer, Truhlar, and coworkers. Comparisons are made with results using charges fit to the electrostatic potential surface (EPS) from ab initio 6-31G* wave functions and from the OPLS-AA force field. OPLS Lennard-Jones parameters for the organic molecules were used together with the TIP4P water model in Monte Carlo simulations with free energy perturbation theory. Absolute free energies of hydration were computed for OPLS united-atom and all-atom methane by annihilating the solutes in water and in the gas phase, and absolute DeltaGhyd values for all other molecules were computed via transformation to one of these references. Optimal charge scaling factors were determined by minimizing the unsigned average error between experimental and calculated hydration free energies. The PM3-based charge models do not lead to lower average errors than obtained with the EPS charges for the subset of 13 molecules in the original study. However, improvement is obtained by scaling the CM1A partial charges by 1.14 and the CM3A charges by 1.15, which leads to average errors of 1.0 and 1.1 kcal/mol for the full set of 25 molecules. The scaled CM1A charges also yield the best results for the hydration of amides including the E/Z free-energy difference for N-methylacetamide in water.     

pmx: Automated protein structure and topology generation for alchemical perturbations

Computational protein design requires methods to accurately estimate free energy changes in protein stability or binding upon an amino acid mutation. From the different approaches available, molecular dynamics‐based alchemical free energy calculations are unique in their accuracy and solid theoretical basis. The challenge in using these methods lies in the need to generate hybrid structures and topologies representing two physical states of a system. A custom made hybrid topology may prove useful for a particular mutation of interest, however, a high throughput mutation analysis calls for a more general approach. In this work, we present an automated procedure to generate hybrid structures and topologies for the amino acid mutations in all commonly used force fields. The described software is compatible with the Gromacs simulation package. The mutation libraries are readily supported for five force fields, namely Amber99SB, Amber99SB*‐ILDN, OPLS‐AA/L, Charmm22*, and Charmm36. © 2014 The Authors Journal of Computational Chemistry Published by Wiley Periodicals, Inc.     

Algorithms for GPU‐based molecular dynamics simulations of complex fluids: Applications to water,mixtures, and liquid crystals

A custom code for molecular dynamics simulations has been designed to run on CUDA‐enabled NVIDIA graphics processing units (GPUs). The double‐precision code simulates multicomponent fluids, with intramolecular and intermolecular forces, coarse‐grained and atomistic models, holonomic constraints, Nosé–Hoover thermostats, and the generation of distribution functions. Algorithms to compute Lennard‐Jones and Gay‐Berne interactions, and the electrostatic force using Ewald summations, are discussed. A neighbor list is introduced to improve scaling with respect to system size. Three test systems are examined: SPC/E water; an n‐hexane/2‐propanol mixture; and a liquid crystal mesogen, 2‐(4‐butyloxyphenyl)‐5‐octyloxypyrimidine. Code performance is analyzed for each system. With one GPU, a 33–119 fold increase in performance is achieved compared with the serial code while the use of two GPUs leads to a 69–287 fold improvement and three GPUs yield a 101–377 fold speedup. © 2015 Wiley Periodicals, Inc.     

Improving the efficiency and reliability of free energy perturbation calculations using overlap sampling methods

A challenge in free energy calculation for complex molecular systems by computer simulation is to obtain a reliable estimate within feasible computational time. In this study, we suggest an answer to this challenge by exploring a simple method, overlap sampling (OS), for producing reliable free-energy results in an efficient way. The formalism of the OS method is based on ensuring sampling of important overlapping phase space during perturbation calculations. This technique samples both forward and reverse free energy perturbation (FEP) to improve the free-energy calculation. It considers the asymmetry of the FEP calculation and features an ability to optimize both the precision and the accuracy of the measurement without affecting the simulation process itself. The OS method is tested at two optimization levels: no optimization (simple OS), and full optimization (equivalent to Bennett's method), and compared to conventional FEP techniques, including the widely used direct FEP averaging method, on three alchemical mutation systems: (a) an anion transformation in water solution, (b) mutation between methanol and ethane, and (c) alchemical change of an adenosine molecule. It is consistently shown that the reliability of free-energy estimates can be greatly improved using the OS techniques at both optimization levels, while the performance of Bennett's method is particularly striking. In addition, the efficiency of a calculation can be significantly improved because the method is able to (a) converge to the right answer quickly, and (b) work for large perturbations. The basic two-stage OS method can be extended to admit additional stages, if needed. We suggest that the OS method can be used as a general perturbation technique for computing free energy differences in molecular simulations.     

Specific interactions and binding free energies between thermolysin and dipeptides: molecular simulations combined with ab initio molecular orbital and classical vibrational analysis

Thermolysin (TLN) is a metalloprotease widely used as a nonspecific protease for sequencing peptide and synthesizing many useful chemical compounds by the chemical industry. It was experimentally shown that the activity and functions of TLN are inhibited by the binding of many types of amino acid dipeptides. However, the binding mechanisms between TLN and dipeptides have not been clarified at the atomic and electronic levels. In this study, we investigated the binding mechanisms between TLN and four dipeptides. Specific interactions and binding free energies (BFEs) between TLN and the dipeptides were calculated using molecular simulations based on classical molecular dynamics and ab initio fragment molecular orbital (FMO) methods. The molecular systems were embedded in solvating water molecules during calculations. The calculated BFEs were qualitatively consistent with the trend of the experimentally observed inhibition of TLN activity by binding of the dipeptides. In addition, the specific interactions between the dipeptides and each amino acid residue of TLN or solvating water molecules were elucidated by the FMO calculations.     

Evaluating nonpolarizable nucleic acid force fields: A systematic comparison of the nucleobases hydration free energies and chloroform‐to‐water partition coefficients

Nucleic acid force fields have been shown to reproduce structural properties of DNA and RNA very well, but comparative studies with respect to thermodynamic properties are rare. As a test for thermodynamic properties, we have computed hydration free energies and chloroform‐to‐water partition coefficients of nucleobases using the AMBER‐99, AMBER‐gaff, CHARMM‐27, GROMOS‐45a4/53a6 and OPLS‐AA force fields. A mutual force field comparison showed a very large spread in the calculated thermodynamic properties, demonstrating that some of the parameter sets require further optimization. The choice of solvent model used in the simulation does not have a significant effect on the results. Comparing the hydration free energies obtained by the various force fields to the adenine and thymine experimental values showed a very large deviation for the GROMOS and AMBER parameter sets. Validation against experimental partition coefficients showed good agreement for the CHARMM‐27 parameter set. In view of mutation studies, differences in partition coefficient between two bases were also compared, and good agreement between experiments and calculations was found for the AMBER‐99 parameter set. Overall, the CHARMM‐27 parameter set performs best with respect to the thermodynamic properties tested here. © 2012 Wiley Periodicals, Inc.     

Influence of drawing and heat treatment on surface free energies of polyethylene terephthalate fibers

The surface free energies of polyethylene terepthalate fibers with different draw ratios were experimentally determined by contact angle measurements inn-alkane/water systems. The dispersive component of the surface free energy increased with increasing draw ratio, whereas the nondispersive one remained almost constant. After heat treatment, the dispersive surface free energy increased, but was reduced above 140°C. The nondispersive component increased by heat treatment at 190°C. The increases in the density and birefringence of the fibres due to the drawing and heat treatment suggested that the increase in the dispersive surface free energy was caused by the increase in the atomic density at the fiber surface due to drawing and heat treatment. ESCA results indicated that the increment in the nondispersive surface free energy due to heat treatment was caused by the addition of functional groups to the fiber surface due to heat treatment.