An evaluation of implicit and explicit solvent model systems for the molecular dynamics simulation of bacteriophage T4 lysozyme

In this report we examine several solvent models for use in molecular dynamics simulations of protein molecules with the Discover program from Biosym Technologies. Our goal was to find a solvent system which strikes a reasonable balance among theoretical rigor, computational efficiency, and experimental reality. We chose phage T4 lysozyme as our model protein and analyzed 14 simulations using different solvent models. We tested both implicit and explicit solvent models using either a linear distance-dependent dielectric or a constant dielectric. Use of a linear distance-dependent dielectric with implicit solvent significantly diminished atomic fluctuations in the protein and kept the protein close to the starting crystal structure. In systems using a constant dielectric and explicit solvent, atomic fluctuations were much greater and the protein was able to sample a larger portion of conformational space. A series of nonbonded cutoff distances (9.0, 11.5, 15.0, 20.0 Å) using both abrupt and smooth truncation of the nonbonded cutoff distances were tested. The method of dual cutoffs was also tested. We found that a minimum nonbonded cutoff distance of 15.0 Å was needed in order to properly couple solvent and solute. Distances shorter than 15.0 Å resulted in a significant temperature gradient between the solvent and solute. In all trajectories using the proprietary Discover switching function, we found significant denaturation in the protein backbone; we were able to run successful trajectories only in those simulations that used no switching function. We were able to significantly reduce the computational burden by using dual cutoffs and still calculate a quality trajectory. In this method, we found that an outer cutoff distance of 15.0 Å and an inner cutoff distance of 11.5 worked well. While a 10 Å shell of explicit water yielded the best results, a 6 A shell of water yielded satisfactory results with nearly a 40% reduction in computational cost. © 1994 John Wiley & Sons, Inc.     

Distinguishing native conformations of proteins from decoys with an effective free energy estimator based on the OPLS all-atom force field and the Surface Generalized Born solvent model

Protein decoy data sets provide a benchmark for testing scoring functions designed for fold recognition and protein homology modeling problems. It is commonly believed that statistical potentials based on reduced atomic models are better able to discriminate native-like from misfolded decoys than scoring functions based on more detailed molecular mechanics models. Recent benchmark tests on small data sets, however, suggest otherwise. In this work, we report the results of extensive decoy detection tests using an effective free energy function based on the OPLS all-atom (OPLS-AA) force field and the Surface Generalized Born (SGB) model for the solvent electrostatic effects. The OPLS-AA/SGB effective free energy is used as a scoring function to detect native protein folds among a total of 48,832 decoys for 32 different proteins from Park and Levitt's 4-state-reduced, Levitt's local-minima, Baker's ROSETTA all-atom, and Skolnick's decoy sets. Solvent electrostatic effects are included through the Surface Generalized Born (SGB) model. All structures are locally minimized without restraints. From an analysis of the individual energy components of the OPLS-AA/SGB energy function for the native and the best-ranked decoy, it is determined that a balance of the terms of the potential is responsible for the minimized energies that most successfully distinguish the native from the misfolded conformations. Different combinations of individual energy terms provide less discrimination than the total energy. The results are consistent with observations that all-atom molecular potentials coupled with intermediate level solvent dielectric models are competitive with knowledge-based potentials for decoy detection and protein modeling problems such as fold recognition and homology modeling.     

Protein under pressure: molecular dynamics simulation of the arc repressor

Experimental nuclear magnetic resonance results for the Arc Repressor have shown that this dimeric protein dissociates into a molten globule at high pressure. This structural change is accompanied by a modification of the hydrogen-bonding pattern of the intermolecular beta-sheet: it changes its character from intermolecular to intramolecular with respect to the two monomers. Molecular dynamics simulations of the Arc Repressor, as a monomer and a dimer, at elevated pressure have been performed with the aim to study this hypothesis and to identify the major structural and dynamical changes of the protein under such conditions. The monomer appears less stable than the dimer. However, the complete dissociation has not been seen because of the long timescale needed to observe this phenomenon. In fact, the protein structure altered very little when increasing the pressure. It became slightly compressed and the dynamics of the side-chains and the unfolding process slowed down. Increasing both, temperature and pressure, a tendency of conversion of intermolecular into intramolecular hydrogen bonds in the beta-sheet region has been detected, supporting the mentioned hypothesis. Also, the onset of denaturation of the separated chains was observed.     

Comparison of protein solution structures refined by molecular dynamics simulation in vacuum,with a generalized Born model,and with explicit water

The inclusion of explicit solvent water in molecular dynamics refinement of NMR structures ought to provide the most physically meaningful accounting for the effects of solvent on structure, but is computationally expensive. In order to evaluate the validity of commonly used vacuum refinements and of recently developed continuum solvent model methods, we have used three different methods to refine a set of NMR solution structures of a medium sized protein, Escherichia coliglutaredoxin 2, from starting structures calculated using the program DYANA. The three different refinement protocols used molecular dynamics simulated annealing with the program AMBER in vacuum (VAC), including a generalized Born (GB) solvent model, and a full calculation including explicit solvent water (WAT). The structures obtained using the three methods of refinements were very similar, a reflection of their generally well-determined nature. However, the structures refined with the generalized Born model were more similar to those from explicit water refinement than those refined in vacuum. Significant improvement was seen in the percentage of backbone dihedral angles in the most favored regions of , space and in hydrogen bond pattern for structures refined with the GB and WAT models, compared with the structures refined in vacuum. The explicit water calculation took an average of 200 h of CPU time per structure on an SGI cluster, compared to 15–90 h for the GB calculation (depending on the parameters used) and 2 h for the vacuum calculation. The generalized Born solvent model proved to be an excellent compromise between the vacuum and explicit water refinements, giving results comparable to those of the explicit water calculation. Some improvement for and angle distribution and hydrogen bond pattern can also be achieved by energy minimizing the vacuum structures with the GB model, which takes a much shorter time than MD simulations with the GB model.     

alpha-helix formation: discontinuous molecular dynamics on an intermediate-resolution protein model.

An intermediate-resolution model of small, homogeneous peptides is introduced, and discontinuous molecular dynamics simulation is applied to study secondary structure formation. Physically, each model residue consists of a detailed three-bead backbone and a simplified single-bead side-chain. Excluded volume and hydrogen bond interactions are constructed with discontinuous (i.e., hard-sphere and square-well) potentials. Simulation results show that the backbone motion of the model is limited to realistic regions of Phi-Psi conformational space. Model polyalanine chains undergo a locally cooperative transition to form alpha-helices that are stabilized by backbone hydrogen bonding, while model polyglycine chains tend to adopt nonhelical structures. When side-chain size is increased beyond a critical diameter, steric interactions prevent formation of long alpha-helices. These trends in helicity as a function of residue type have been well documented by experimental, theoretical, and simulation studies and demonstrate the ability of the intermediate-resolution model developed in this work to accurately mimic realistic peptide behavior. The efficient algorithm used permits observation of the complete helix-coil transition within 15 min on a single-processor workstation, suggesting that simulations of very long times are possible with this model.     

Implicit solvent models for flexible protein-protein docking by molecular dynamics simulation

The suitability of three implicit solvent models for flexible protein-protein docking by procedures using molecular dynamics simulation is investigated. The three models are (i) the generalized Born (GB) model implemented in the program AMBER6.0; (ii) a distance-dependent dielectric (DDD) model; and (iii) a surface area-dependent model that we have parameterized and call the NPSA model. This is a distance-dependent dielectric model modified by neutralizing the ionizable side-chains and adding a surface area-dependent solvation term. These solvent models were first tested in molecular dynamics simulations at 300 K of the native structures of barnase, barstar, segment B1 of protein G, and three WW domains. These protein structures display a range of secondary structure contents and stabilities. Then, to investigate the performance of the implicit solvent models in protein docking, molecular dynamics simulations of barnase/barstar complexation, as well as PIN1 WW domain/peptide complexation, were conducted, starting from separated unbound structures. The simulations show that the NPSA model has significant advantages over the DDD and GB models in maintaining the native structures of the proteins and providing more accurate docked complexes.     

Computational protein design with a generalized Born solvent model: application to Asparaginyl-tRNA synthetase

Computational Protein Design (CPD) is a promising method for high throughput protein and ligand mutagenesis. Recently, we developed a CPD method that used a polar-hydrogen energy function for protein interactions and a Coulomb/Accessible Surface Area (CASA) model for solvent effects. We applied this method to engineer aspartyl-adenylate (AspAMP) specificity into Asparaginyl-tRNA synthetase (AsnRS), whose substrate is asparaginyl-adenylate (AsnAMP). Here, we implement a more accurate function, with an all-atom energy for protein interactions and a residue-pairwise generalized Born model for solvent effects. As a first test, we compute aminoacid affinities for several point mutants of Aspartyl-tRNA synthetase (AspRS) and Tyrosyl-tRNA synthetase and stability changes for three helical peptides and compare with experiment. As a second test, we readdress the problem of AsnRS aminoacid engineering. We compare three design criteria, which optimize the folding free-energy, the absolute AspAMP affinity, and the relative (AspAMP-AsnAMP) affinity. The sequences and conformations are improved with respect to our previous, polar-hydrogen/CASA study: For several designed complexes, the AspAMP carboxylate forms three interactions with a conserved arginine and a designed lysine, as in the active site of the AspRS:AspAMP complex. The conformations and interactions are well maintained in molecular dynamics simulations and the sequences have an inverted specificity, favoring AspAMP over AsnAMP. The method is not fully successful, since experimental measurements with the seven most promising sequences show that they do not catalyze at a detectable level the adenylation of Asp (or Asn) with ATP. This may be due to weak AspAMP binding and/or disruption of transition-state stabilization.     

Distinguish protein decoys by using a scoring function based on a new AMBER force field, short molecular dynamics simulations, and the generalized born solvent model

Recent works have shown the ability of physics-based potentials (e.g., CHARMM and OPLS-AA) and energy minimization to differentiate the native protein structures from large ensemble of non-native structures. In this study, we extended previous work by other authors and developed an energy scoring function using a new set of AMBER parameters (also recently developed in our laboratory) in conjunction with molecular dynamics and the Generalized Born solvent model. We evaluated the performance of our new scoring function by examining its ability to distinguish between the native and decoy protein structures. Here we present a systematic comparison of our results with those obtained with use of other physics-based potentials by previous authors. A total of 7 decoy sets, 117 protein sequences, and more than 41,000 structures were evaluated. The results of our study showed that our new scoring function represents a significant improvement over previously published physics-based scoring functions.     

Evaluation of a fast implicit solvent model for molecular dynamics simulations.

A solvation term based on the solvent accessible surface area (SASA) is combined with the CHARMM polar hydrogen force field for the efficient simulation of peptides and small proteins in aqueous solution. Only two atomic solvation parameters are used: one is negative for favoring the direct solvation of polar groups and the other positive for taking into account the hydrophobic effect on apolar groups. To approximate the water screening effects on the intrasolute electrostatic interactions, a distance-dependent dielectric function is used and ionic side chains are neutralized. The use of an analytical approximation of the SASA renders the model extremely efficient (i.e., only about 50% slower than in vacuo simulations). The limitations and range of applicability of the SASA model are assessed by simulations of proteins and structured peptides. For the latter, the present study and results reported elsewhere show that with the SASA model it is possible to sample a significant amount of folding/unfolding transitions, which permit the study of the thermodynamics and kinetics of folding at an atomic level of detail.     

Protein molecular dynamics with electrostatic force entirely determined by a single Poisson-Boltzmann calculation

The electrostatic force including the intramolecular Coulombic interactions and the electrostatic contribution of solvation effect were entirely calculated by using the finite difference Poisson-Boltzmann method (FDPB), which was incorporated into the GROMOS96 force field to complete a new finite difference stochastic dynamics procedure (FDSD). Simulations were performed on an insulin dimer. Different relative dielectric constants were successively assigned to the protein interior; a value of 17 was selected as optimal for our system. The simulation data were analyzed and compared with those obtained from 500-ps molecular dynamics (MD) simulation with explicit water and a 500-ps conventional stochastic dynamics (SD) simulation without the mean solvent force. The results indicate that the FDSD method with GROMOS96 force field is suitable to study the dynamics and structure of proteins in solution if used with the optimal protein dielectric constant.     

Folding simulations of Trp‐cage mini protein in explicit solvent using biasing potential replica‐exchange molecular dynamics simulations

Replica exchange molecular dynamics (RexMD) simulations are frequently used for studying structure formation and dynamics of peptides and proteins. A significant drawback of standard temperature RexMD is, however, the rapid increase of the replica number with increasing system size to cover a desired temperature range. A recently developed Hamiltonian RexMD method has been used to study folding of the Trp‐cage protein. It employs a biasing potential that lowers the backbone dihedral barriers and promotes peptide backbone transitions along the replica coordinate. In two independent applications of the biasing potential RexMD method including explicit solvent and starting from a completely unfolded structure the formation of near‐native conformations was observed after 30–40 ns simulation time. The conformation representing the most populated cluster at the final simulation stage had a backbone root mean square deviation of ～1.3 Å from the experimental structure. This was achieved with a very modest number of five replicas making it well suited for peptide and protein folding and refinement studies including explicit solvent. In contrast, during five independent continuous 70 ns molecular dynamics simulations formation of collapsed states but no near native structure formation was observed. The simulations predict a largely collapsed state with a significant helical propensity for the helical domain of the Trp‐cage protein already in the unfolded state. Hydrogen bonded bridging water molecules were identified that could play an active role by stabilizing the arrangement of the helical domain with respect to the rest of the chain already in intermediate states of the protein. Proteins 2009. © 2008 Wiley‐Liss, Inc.     

Protein Structure Prediction with a Combined Solvation Free Energy-Molecular Mechanics Force Field

Abstract

Models of protein structure are frequently used to determine the physical characteristics of a protein when the crystal structure is not available. We developed a procedure to optimize such models, by use of a combined solvation free energy and molecular mechanics force field. Appropriately chosen atomic solvation parameters were defined using the criterion that the resulting protein model should deviate least from the crystal structure upon a forty picosecond molecular dynamics simulation carried out using the combined force field. Several tests were performed to refine the set of atomic solvation parameters which best complement the molecular mechanics forces. Four sets of parameters from the literature were tested and an empirically optimized set is proposed. The parameters are defined on a well characterized small molecule (alanyl dipeptide) and on the highly refined crystal structure of rat trypsin, and then tested on a second highly refined crystal structure of α-lytic protease. The new set of atomic solvation parameters provides a significant improvement over molecular mechanics alone in energy minimization of protein structures. This combined force field also has advantages over the use of explicit solvent as it is possible to take solvent effects into account during energetic conformational searching when modeling a homologous protein structure from a known crystal structure.     

Prediction of protein solvent accessibility using support vector machines

A Support Vector Machine learning system has been trained to predict protein solvent accessibility from the primary structure. Different kernel functions and sliding window sizes have been explored to find how they affect the prediction performance. Using a cut-off threshold of 15% that splits the dataset evenly (an equal number of exposed and buried residues), this method was able to achieve a prediction accuracy of 70.1% for single sequence input and 73.9% for multiple alignment sequence input, respectively. The prediction of three and more states of solvent accessibility was also studied and compared with other methods. The prediction accuracies are better than, or comparable to, those obtained by other methods such as neural networks, Bayesian classification, multiple linear regression, and information theory. In addition, our results further suggest that this system may be combined with other prediction methods to achieve more reliable results, and that the Support Vector Machine method is a very useful tool for biological sequence analysis.     

An analysis of the molecular origin of osmolyte-dependent protein stability

Protein solvation is the key determinant for isothermal, concentration-dependent effects on protein equilibria, such as folding. The required solvation information can be extracted from experimental thermodynamic data using Kirkwood-Buff theory. Here we derive and discuss general properties of proteins and osmolytes that are pertinent to their biochemical behavior. We find that hydration depends very little on osmolyte concentration and type. Strong dependencies on both osmolyte concentration and type are found for osmolyte self-solvation and protein-osmolyte solvation changes upon unfolding. However, solvation in osmolyte solutions does not involve complex concentration dependencies as found in organic molecules that are not used as osmolytes in nature. It is argued that the simple solvation behavior of naturally occurring osmolytes is a prerequisite for their usefulness in osmotic regulation in vivo.     

A molecular dynamics study of acylphosphatase in aggregation-promoting conditions: the influence of trifluoroethanol/water solvent

The 98-residue protein acylphosphatase exhibits a high propensity for aggregation under certain conditions. Aggregates formed from wild-type acylphosphatase in the presence of 2,2,2-trifluoroethanol and from highly destabilized mutants are essentially identical in structure. Furthermore, it has been shown by mutational studies that different regions of the protein are important for aggregation and folding. In the present molecular dynamics study, we compare the behavior of the protein in aqueous solution and in a 25% (v/v) 2,2,2-trifluoroethanol/water environment mimicking the experimental conditions. The 2,2,2-trifluoroethanol surrounding affects the structure of the protein mostly in the regions important for aggregation, in good agreement with experimental data. This suggests that the early step of (partly) unfolding, which precedes the aggregation process, has been observed.     

Quantum simulation for muoniated and deuterated methyl radicals in implicit water solvent: combined ab initio path integral molecular dynamics and the polarizable continuum model simulation study

We have combined ab initio path integral molecular dynamics (PIMD) simulation and the polarizable continuum model (PCM) method to efficiently incorporate solvent effects into nuclear quantum fluctuation of molecular systems. Our combined ab initio PIMD–PCM simulation was applied to muoniated and deuterated methyl radical immersed in implicit water solvent to gain information on solvent and isotope effects from one simulation run. We found that solvent effects lead to the bond elongation and a decrease in the magnitude of isotropic hyperfine coupling constants. These are consistent with the trends in conventional static calculations and experiments. In addition, the performance of cavity models (universal force field, united atom specified for Kohn–Sham and these hybrid models) and the conservation of the PIMD–PCM Hamiltonian were accessed. We confirmed that solvent effects on nuclear quantum fluctuation are efficiently computed using our combined simulation of quantum solute in implicit solvent.     

Molecular dynamics studies of a DNA-binding protein: 2. An evaluation of implicit and explicit solvent models for the molecular dynamics simulation of the Escherichia coli trp repressor.

Although aqueous simulations with periodic boundary conditions more accurately describe protein dynamics than in vacuo simulations, these are computationally intensive for most proteins. Trp repressor dynamic simulations with a small water shell surrounding the starting model yield protein trajectories that are markedly improved over gas phase, yet computationally efficient. Explicit water in molecular dynamics simulations maintains surface exposure of protein hydrophilic atoms and burial of hydrophobic atoms by opposing the otherwise asymmetric protein-protein forces. This properly orients protein surface side chains, reduces protein fluctuations, and lowers the overall root mean square deviation from the crystal structure. For simulations with crystallographic waters only, a linear or sigmoidal distance-dependent dielectric yields a much better trajectory than does a constant dielectric model. As more water is added to the starting model, the differences between using distance-dependent and constant dielectric models becomes smaller, although the linear distance-dependent dielectric yields an average structure closer to the crystal structure than does a constant dielectric model. Multiplicative constants greater than one, for the linear distance-dependent dielectric simulations, produced trajectories that are progressively worse in describing trp repressor dynamics. Simulations of bovine pancreatic trypsin were used to ensure that the trp repressor results were not protein dependent and to explore the effect of the nonbonded cutoff on the distance-dependent and constant dielectric simulation models. The nonbonded cutoff markedly affected the constant but not distance-dependent dielectric bovine pancreatic trypsin inhibitor simulations. As with trp repressor, the distance-dependent dielectric model with a shell of water surrounding the protein produced a trajectory in better agreement with the crystal structure than a constant dielectric model, and the physical properties of the trajectory average structure, both with and without a nonbonded cutoff, were comparable.     

On using time-averaging restraints in molecular dynamics simulation

Introducing experimental values as restraints into molecular dynamics (MD) simulations to bias the values of particular molecular properties, such as nuclear Overhauser effect intensities or distances, 3J coupling constants, chemical shifts or crystallographic structure factors, towards experimental values is a widely used structure refinement method. To account for the averaging of experimentally derived quantities inherent in the experimental techniques, time-averaging restraining methods may be used. In the case of structure refinement using 3J coupling constants from NMR experiments, time-averaging methods previously proposed can suffer from large artificially induced structural fluctuations. A modified time-averaged restraining potential energy function is proposed which overcomes this problem. The different possible approaches are compared using stochastic dynamics simulations of antamanide, a cyclic peptide of ten residues.     

Computational sidechain placement and protein mutagenesis with implicit solvent models

Structure prediction and computational protein design should benefit from accurate solvent models. We have applied implicit solvent models to two problems that are central to this area. First, we performed sidechain placement for 29 proteins, using a solvent model that combines a screened Coulomb term with an Accessible Surface Area term (CASA model). With optimized parameters, the prediction quality is comparable with earlier work that omitted electrostatics and solvation altogether. Second, we computed the stability changes associated with point mutations involving ionized sidechains. For over 1000 mutations, including many fully or partly buried positions, we compared CASA and two generalized Born models (GB) with a more accurate model, which solves the Poisson equation of continuum electrostatics numerically. CASA predicts the correct sign and order of magnitude of the stability change for 81% of the mutations, compared to 97% with the best GB. We also considered 140 mutations for which experimental data are available. Comparing to experiment requires additional assumptions about the unfolded protein structure, protein relaxation in response to the mutations, and contributions from the hydrophobic effect. With a simple, commonly-used unfolded state model, the mean unsigned error is 2.1 kcal/mol with both CASA and the best GB. Overall, the electrostatic model is not important for sidechain placement; CASA and GB are equivalent for surface mutations, while GB is far superior for fully or partly buried positions. Thus, for problems like protein design that involve all these aspects, the most recent GB models represent an important step forward. Along with the recent discovery of efficient, pairwise implementations of GB, this will open new possibilities for the computational engineering of proteins.