Toward accurate barriers for enzymatic reactions: QM/MM case study on p-hydroxybenzoate hydroxylase

The hydroxylation reaction catalyzed by p-hydroxybenzoate hydroxylase has been investigated by quantum mechanical/molecular mechanical (QM/MM) calculations at different levels of QM theory. The solvated enzyme was modeled (approximately 23,000 atoms in total, 49 QM atoms). The geometries of reactant and transition state were optimized for ten representative pathways using semiempirical (AM1) and density functional (B3LYP) methods as QM components. Single-point calculations at B3LYP/MM optimized geometries were performed with local correlation methods [LMP2, LCCSD(T0)] and augmented triple-zeta basis sets. A careful validation of the latter approach with regard to all computational parameters indicates convergence of the QM contribution to the computed barriers to within approximately 1 kcal mol(-1). Comparison with the available experimental data supports this assessment.     

Performance assessment of semiempirical molecular orbital methods in describing halogen bonding: quantum mechanical and quantum mechanical/molecular mechanical-molecular dynamics study

The performance of semiempirical molecular-orbital methods--MNDO, MNDO-d, AM1, RM1, PM3 and PM6--in describing halogen bonding was evaluated, and the results were compared with molecular mechanical (MM) and quantum mechanical (QM) data. Three types of performance were assessed: (1) geometrical optimizations and binding energy calculations for 27 halogen-containing molecules complexed with various Lewis bases (Two of the tested methods, AM1 and RM1, gave results that agree with the QM data.); (2) charge distribution calculations for halobenzene molecules, determined by calculating the solvation free energies of the molecules relative to benzene in explicit and implicit generalized Born (GB) solvents (None of the methods gave results that agree with the experimental data.); and (3) appropriateness of the semiempirical methods in the hybrid quantum-mechanical/molecular-mechanical (QM/MM) scheme, investigated by studying the molecular inhibition of CK2 protein by eight halobenzimidazole and -benzotriazole derivatives using hybrid QM/MM molecular-dynamics (MD) simulations with the inhibitor described at the QM level by the AM1 method and the rest of the system described at the MM level. The pure MM approach with inclusion of an extra point of positive charge on the halogen atom approach gave better results than the hybrid QM/MM approach involving the AM1 method. Also, in comparison with the pure MM-GBSA (generalized Born surface area) binding energies and experimental data, the calculated QM/MM-GBSA binding energies of the inhibitors were improved by replacing the G(GB,QM/MM) solvation term with the corresponding G(GB,MM) term.     

Application of a semi-empirical dispersion correction for modeling water clusters

The Grimme-D3 semi-empirical dispersion energy correction has been implemented for the original effective fragment potential for water (EFP1), and for systems that contain water molecules described by both correlated ab initio quantum mechanical (QM) molecules and EFP1. Binding energies obtained with these EFP1-D and QM/EFP1-D methods were tested using 27 benchmark species, including neutral, protonated, deprotonated, and auto-ionized water clusters and nine solute–water binary complexes. The EFP1-D and QM/EFP1-D binding energies are compared with those obtained using fully QM methods: second-order perturbation theory, and coupled cluster theory, CCSD(T), at the complete basis set (CBS) limit. The results show that the EFP1-D and QM/EFP1-D binding energies are in good agreement with CCSD(T)/CBS binding energies with a mean absolute error of 5.9 kcal/mol for water clusters and 0.8 kcal/mol for solute–water binary complexes. © 2018 Wiley Periodicals, Inc.     

Molecular dynamics and free energy perturbation study of hydride-ion transfer step in dihydrofolate reductase using combined quantum and molecular mechanical model

We used molecular dynamics simulation and free energy perturbation (FEP) methods to investigate the hydride-ion transfer step in the mechanism for the nicotinamide adenine dinucleotide phosphate (NADPH)-dependent reduction of a novel substrate by the enzyme dihydrofolate reductase (DHFR). The system is represented by a coupled quantum mechanical and molecular mechanical (QM/MM) model based on the AM1 semiempirical molecular orbital method for the reacting substrate and NADPH cofactor fragments, the AMBER force field for DHFR, and the TIP3P model for solvent water. The FEP calculations were performed for a number of choices for the QM system. The substrate, 8-methylpterin, was treated quantum mechanically in all the calculations, while the larger cofactor molecule was partitioned into various QM and MM regions with the addition of “link” atoms (F, CH₃, and H). Calculations were also carried out with the entire NADPH molecule treated by QM. The free energies of reaction and the net charges on the NADPH fragments were used to determine the most appropriate QM/MM model. The hydride-ion transfer was also carried out over several FEP pathways, and the QM and QM/MM component free energies thus calculated were found to be state functions (i.e., independent of pathway). A ca. 10 kcal/mol increase in free energy for the hydride-ion transfer with an activation barrier of ca. 30 kcal/mol was calculated. The increase in free energy on the hydride-ion transfer arose largely from the QM/MM component. Analysis of the QM/MM energy components suggests that, although a number of charged residues may contribute to the free energy change through long-range electrostatic interactions, the only interaction that can account for the 10 kcal/mol increase in free energy is the hydrogen bond between the carboxylate side chain of Glu30 (avian DHFR) and the activated (protonated) substrate. © 1998 John Wiley & Sons, Inc. J Comput Chem 19: 977–988, 1998     

Quantum mechanical/molecular mechanical study on the mechanism of the enzymatic Baeyer-Villiger reaction

We report a combined quantum mechanical/molecular mechanical (QM/MM) study on the mechanism of the enzymatic Baeyer-Villiger reaction catalyzed by cyclohexanone monooxygenase (CHMO). In QM/MM geometry optimizations and reaction path calculations, density functional theory (B3LYP/TZVP) is used to describe the QM region consisting of the substrate (cyclohexanone), the isoalloxazine ring of C4a-peroxyflavin, the side chain of Arg-329, and the nicotinamide ring and the adjacent ribose of NADP(+), while the remainder of the enzyme is represented by the CHARMM force field. QM/MM molecular dynamics simulations and free energy calculations at the semiempirical OM3/CHARMM level employ the same QM/MM partitioning. According to the QM/MM calculations, the enzyme-reactant complex contains an anionic deprotonated C4a-peroxyflavin that is stabilized by strong hydrogen bonds with the Arg-329 residue and the NADP(+) cofactor. The CHMO-catalyzed reaction proceeds via a Criegee intermediate having pronounced anionic character. The initial addition reaction has to overcome an energy barrier of about 9 kcal/mol. The formed Criegee intermediate occupies a shallow minimum on the QM/MM potential energy surface and can undergo fragmentation to the lactone product by surmounting a second energy barrier of about 7 kcal/mol. The transition state for the latter migration step is the highest point on the QM/MM energy profile. Gas-phase reoptimizations of the QM region lead to higher barriers and confirm the crucial role of the Arg-329 residue and the NADP(+) cofactor for the catalytic efficiency of CHMO. QM/MM calculations for the CHMO-catalyzed oxidation of 4-methylcyclohexanone reproduce and rationalize the experimentally observed (S)-enantioselectivity for this substrate, which is governed by the conformational preferences of the corresponding Criegee intermediate and the subsequent transition state for the migration step.     

Extension of the PDDG/PM3 and PDDG/MNDO semiempirical molecular orbital methods to the halogens

The new semiempirical methods, PDDG/PM3 and PDDG/MNDO, have been parameterized for halogens. For comparison, the original MNDO and PM3 were also reoptimized for the halogens using the same training set; these modified methods are referred to as MNDO' and PM3'. For 442 halogen-containing molecules, the smallest mean absolute error (MAE) in heats of formation is obtained with PDDG/PM3 (5.6 kcal/mol), followed by PM3' (6.1 kcal/mol), PDDG/MNDO (6.6 kcal/mol), PM3 (8.1 kcal/mol), MNDO' (8.5 kcal/mol), AM1 (11.1 kcal/mol), and MNDO (14.0 kcal/mol). For normal-valent halogen-containing molecules, the PDDG methods also provide improved heats of formation over MNDO/d. Hypervalent compounds were not included in the training set and improvements over the standard NDDO methods with sp basis sets were not obtained. For small haloalkanes, the PDDG methods yield more accurate heats of formation than are obtained from density functional theory (DFT) with the B3LYP and B3PW91 functionals using large basis sets. PDDG/PM3 and PM3' also give improved binding energies over the standard NDDO methods for complexes involving halide anions, and they are competitive with B3LYP/6-311++G(d,p) results including thermal corrections. Among the semiempirical methods studied, PDDG/PM3 also generates the best agreement with high-level ab initio G2 and CCSD(T) intrinsic activation energies for S(N)2 reactions involving methyl halides and halide anions. Finally, the MAEs in ionization potentials, dipole moments, and molecular geometries show that the parameter sets for the PDDG and reoptimized NDDO methods reduce the MAEs in heats of formation without compromising the other important QM observables.     

A combined quantum chemical/molecular mechanical study of hydrogen-bonded systems

A computational approach, which involves the combination of the OPLS force field and molecular orbital MNDO , AM 1, and PM 3 methods, has been developed to describe the effects of a large, molecular mechanically simulated environment on the Hamiltonian of a quantum chemical system. To test the validity of the combined quantum mechanical/molecular mechanical (QM /MM ) potential, a systematic study of the structures and energies of neutral and charged hydrogen-bonded complexes has been carried out, including comparisons with pure semiempirical calculations and available experimental and ab initio data. It is shown that, in many cases, the hybrid QM /MM potential behaves better than do related MNDO /M , AM 1, and PM 3 methods. As a case in point, the draw-back of AM 1 favoring bifurcated H-bonded structures over single ones is not presented in the combined AM 1/OPLS scheme. Possible ways of improvement of the combined QM /MM potential are discussed. © 1992 John Wiley & Sons, Inc.     

Comparison of linear-scaling semiempirical methods and combined quantum mechanical/molecular mechanical methods for enzymic reactions. II. An energy decomposition analysis

QM/MM methods have been developed as a computationally feasible solution to QM simulation of chemical processes, such as enzyme-catalyzed reactions, within a more approximate MM representation of the condensed-phase environment. However, there has been no independent method for checking the quality of this representation, especially for highly nonisotropic protein environments such as those surrounding enzyme active sites. Hence, the validity of QM/MM methods is largely untested. Here we use the possibility of performing all-QM calculations at the semiempirical PM3 level with a linear-scaling method (MOZYME) to assess the performance of a QM/MM method (PM3/AMBER94 force field). Using two model pathways for the hydride-ion transfer reaction of the enzyme dihydrofolate reductase studied previously (Titmuss et al., Chem Phys Lett 2000, 320, 169-176), we have analyzed the reaction energy contributions (QM, QM/MM, and MM) from the QM/MM results and compared them with analogous-region components calculated via an energy partitioning scheme implemented into MOZYME. This analysis further divided the MOZYME components into Coulomb, resonance and exchange energy terms. For the model in which the MM coordinates are kept fixed during the reaction, we find that the MOZYME and QM/MM total energy profiles agree very well, but that there are significant differences in the energy components. Most significantly there is a large change (approximately 16 kcal/mol) in the MOZYME MM component due to polarization of the MM region surrounding the active site, and which arises mostly from MM atoms close to (<10 A) the active-site QM region, which is not modelled explicitly by our QM/MM method. However, for the model where the MM coordinates are allowed to vary during the reaction, we find large differences in the MOZYME and QM/MM total energy profiles, with a discrepancy of 52 kcal/mol between the relative reaction (product-reactant) energies. This is largely due to a difference in the MM energies of 58 kcal/mol, of which we can attribute approximately 40 kcal/mol to geometry effects in the MM region and the remainder, as before, to MM region polarization. Contrary to the fixed-geometry model, there is no correlation of the MM energy changes with distance from the QM region, nor are they contributed by only a few residues. Overall, the results suggest that merely extending the size of the QM region in the QM/MM calculation is not a universal solution to the MOZYME- and QM/MM-method differences. They also suggest that attaching physical significance to MOZYME Coulomb, resonance and exchange components is problematic. Although we conclude that it would be possible to reparameterize the QM/MM force field to reproduce MOZYME energies, a better way to account for both the effects of the protein environment and known deficiencies in semiempirical methods would be to parameterize the force field based on data from DFT or ab initio QM linear-scaling calculations. Such a force field could be used efficiently in MD simulations to calculate free energies.     

Protein-ligand interaction energies with dispersion corrected density functional theory and high-level wave function based methods

With dispersion-corrected density functional theory (DFT-D3) intermolecular interaction energies for a diverse set of noncovalently bound protein-ligand complexes from the Protein Data Bank are calculated. The focus is on major contacts occurring between the drug molecule and the binding site. Generalized gradient approximation (GGA), meta-GGA, and hybrid functionals are used. DFT-D3 interaction energies are benchmarked against the best available wave function based results that are provided by the estimated complete basis set (CBS) limit of the local pair natural orbital coupled-electron pair approximation (LPNO-CEPA/1) and compared to MP2 and semiempirical data. The size of the complexes and their interaction energies (ΔE(PL)) varies between 50 and 300 atoms and from -1 to -65 kcal/mol, respectively. Basis set effects are considered by applying extended sets of triple- to quadruple-ζ quality. Computed total ΔE(PL) values show a good correlation with the dispersion contribution despite the fact that the protein-ligand complexes contain many hydrogen bonds. It is concluded that an adequate, for example, asymptotically correct, treatment of dispersion interactions is necessary for the realistic modeling of protein-ligand binding. Inclusion of the dispersion correction drastically reduces the dependence of the computed interaction energies on the density functional compared to uncorrected DFT results. DFT-D3 methods provide results that are consistent with LPNO-CEPA/1 and MP2, the differences of about 1-2 kcal/mol on average (<5% of ΔE(PL)) being on the order of their accuracy, while dispersion-corrected semiempirical AM1 and PM3 approaches show a deviating behavior. The DFT-D3 results are found to depend insignificantly on the choice of the short-range damping model. We propose to use DFT-D3 as an essential ingredient in a QM/MM approach for advanced virtual screening approaches of protein-ligand interactions to be combined with similarly "first-principle" accounts for the estimation of solvation and entropic effects.     

Benchmark calculations on models of the phosphoryl transfer reaction catalyzed by protein kinase A

Protein phosphorylation has been proved to be of great importance in many stages of cell life. In the last few years, its reaction mechanism has been extensively studied. In this work we present the analysis of the performances of several computational methods with different computational costs (from multilevel to semiempirical) to point out the best method to be used at each level in the study of phosphoryl transfer. Finally, we center on the semiempirical methods, and mainly on the AM1/d Hamiltonian with different sets of parameters, which will permit hybrid quantum mechanics/molecular mechanics (QM/MM) free energy calculations on big models at an acceptable computational cost. We have used quite a large set of molecules and model reactions to test the computational methods, reproducing all the chemical steps involved in the mainly accepted reaction pathways for the protein phosphorylation. In the end, we also present the results for an enlarged model, cut out from an entire biological model: we compare the 2-D PES at the B3LYP and AM1/d levels with the purpose of obtaining a correction for the semiempirical method. The AM1/d-PhoT semiempirical parameterization corrected using single-point energy calculations at the B3LYP/MG3S level seems to be suitable to carry out reliable QM/MM calculations of the complete biological system.     

A Mixed QM/MM Scoring Function to Predict Protein-Ligand Binding Affinity

Computational methods for predicting protein-ligand binding free energy continue to be popular as a potential cost-cutting method in the drug discovery process. However, accurate predictions are often difficult to make as estimates must be made for certain electronic and entropic terms in conventional force field based scoring functions. Mixed quantum mechanics/molecular mechanics (QM/MM) methods allow electronic effects for a small region of the protein to be calculated, treating the remaining atoms as a fixed charge background for the active site. Such a semi-empirical QM/MM scoring function has been implemented in AMBER using DivCon and tested on a set of 23 metalloprotein-ligand complexes, where QM/MM methods provide a particular advantage in the modeling of the metal ion. The binding affinity of this set of proteins can be calculated with an R(2) of 0.64 and a standard deviation of 1.88 kcal/mol without fitting and 0.71 and a standard deviation of 1.69 kcal/mol with fitted weighting of the individual scoring terms. In this study we explore using various methods to calculate terms in the binding free energy equation, including entropy estimates and minimization standards. From these studies we found that using the rotational bond estimate to ligand entropy results in a reasonable R(2) of 0.63 without fitting. We also found that using the ESCF energy of the proteins without minimization resulted in an R(2) of 0.57, when using the rotatable bond entropy estimate.     

Protein environment facilitates O2 binding in non-heme iron enzyme. An insight from ONIOM calculations on isopenicillin N synthase (IPNS)

Binding of dioxygen to a non-heme enzyme has been modeled using the ONIOM combined quantum mechanical/molecular mechanical (QM/MM) method. For the present system, isopenicillin N synthase (IPNS), binding of dioxygen is stabilized by 8-10 kcal/mol for a QM:MM (B3LYP:Amber) protein model compared to a quantum mechanical model of the active site only. In the protein system, the free energy change of O2 binding is close to zero. Two major factors consistently stabilize O2 binding. The first effect, evaluated at the QM level, originates from a change in coordination geometry of the iron center. The active-site model artificially favors the deoxy state (O2 not bound) because it allows too-large rearrangements of the five-coordinate iron site. This error is corrected when the protein is included. The corresponding effect on binding energies is 3-6 kcal/mol, depending on the coordination mode of O2 (side-on or end-on). The second major factor that stabilizes O2 binding is van der Waals interactions between dioxygen and the surrounding enzyme. These interactions, 3-4 kcal/mol at the MM level, are neglected in models that include only the active site. Polarization of the active site by surrounding amino acids does not have a significant effect on the binding energy in the present system.     

Convergence in the QM‐only and QM/MM modeling of enzymatic reactions: A case study for acetylene hydratase

We report systematic quantum mechanics‐only (QM‐only) and QM/molecular mechanics (MM) calculations on an enzyme‐catalyzed reaction to assess the convergence behavior of QM‐only and QM/MM energies with respect to the size of the chosen QM region. The QM and MM parts are described by density functional theory (typically B3LYP/def2‐SVP) and the CHARMM force field, respectively. Extending our previous work on acetylene hydratase with QM regions up to 157 atoms (Liao and Thiel, J. Chem. Theory Comput. 2012, 8, 3793), we performed QM/MM geometry optimizations with a QM region M4 composed of 408 atoms, as well as further QM/MM single‐point calculations with even larger QM regions up to 657 atoms. A charge deletion analysis was conducted for the previously used QM/MM model ( M3a , with a QM region of 157 atoms) to identify all MM residues with strong electrostatic contributions to the reaction energetics (typically more than 2 kcal/mol), which were then included in M4 . QM/MM calculations with this large QM region M4 lead to the same overall mechanism as the previous QM/MM calculations with M3a , but there are some variations in the relative energies of the stationary points, with a mean absolute deviation (MAD) of 2.7 kcal/mol. The energies of the two relevant transition states are close to each other at all levels applied (typically within 2 kcal/mol), with the first (second) one being rate‐limiting in the QM/MM calculations with M3a ( M4 ). QM‐only gas‐phase calculations give a very similar energy profile for QM region M4 (MAD of 1.7 kcal/mol), contrary to the situation for M3a where we had previously found significant discrepancies between the QM‐only and QM/MM results (MAD of 7.9 kcal/mol). Extension of the QM region beyond M4 up to M7 (657 atoms) leads to only rather small variations in the relative energies from single‐point QM‐only and QM/MM calculations (MAD typically about 1–2 kcal/mol). In the case of acetylene hydratase, a model with 408 QM atoms thus seems sufficient to achieve convergence in the computed relative energies to within 1–2 kcal/mol.Copyright © 2013 Wiley Periodicals, Inc.     

Protein/ligand binding free energies calculated with quantum mechanics/molecular mechanics

The calculation of binding affinities for flexible ligands has hitherto required the availability of reliable molecular mechanics parameters for the ligands, a restriction that can in principle be lifted by using a mixed quantum mechanics/molecular mechanics (QM/MM) representation in which the ligand is treated quantum mechanically. The feasibility of this approach is evaluated here, combining QM/MM with the Poisson-Boltzmann/surface area model of continuum solvation and testing the method on a set of 47 benzamidine derivatives binding to trypsin. The experimental range of the absolute binding energy (DeltaG = -3.9 to -7.6 kcal/mol) is reproduced well, with a root-mean-square (RMS) error of 1.2 kcal/mol. When QM/MM is applied without reoptimization to the very different ligands of FK506 binding protein the RMS error is only 0.7 kcal/mol. The results show that QM/MM is a promising new avenue for automated docking and scoring of flexible ligands. Suggestions are made for further improvements in accuracy.     

Optimal scaling factors for CM1 and CM3 atomic charges in RM1-based aqueous simulations

Scaling factors for atomic charges derived from the RM1 semiempirical quantum mechanical wavefunction in conjunction with CM1 and CM3 charge models have been optimized by minimizing errors in absolute free energies of hydration, ΔG_hyd, for a set of 40 molecules. Monte Carlo statistical mechanics simulations and free energy perturbation theory were used to annihilate the solutes in gas and in a box of TIP4P water molecules. Lennard–Jones parameters from the optimized potentials for liquid simulations‐all atom (OPLS–AA) force field were utilized for the organic compounds. Optimal charge scaling factors have been determined as 1.11 and 1.14 for the CM1R and CM3R methods, respectively, and the corresponding unsigned average errors in ΔG_hyd relative to experiment were 2.05 and 1.89 kcal/mol. Computed errors in aniline and two derivatives were particularly large for RM1 and their removal from the data set lowered the overall errors to 1.61 and 1.75 kcal/mol for CM1R and CM3R. Comparisons are made to the AM1 method which yielded total errors in ΔG_hyd of 1.50 and 1.64 kcal/mol for CM1A*1.14 and CM3A*1.15, respectively. This work is motivated by the need for a highly efficient yet accurate quantum mechanical (QM) method to study condensed‐phase and enzymatic chemical reactions via mixed QM and molecular mechanical (QM/MM) simulations. As an initial test, the Menshutkin reaction between NH₃ and CH₃Cl in water was computed using a RM1/TIP4P‐Ew/CM3R procedure and the resultant ΔG^?, ΔG_rxn, and geometries were in reasonable accord with other computational methods; however, some potentially serious shortcomings in RM1 are discussed. © 2011 Wiley Periodicals, Inc. J Comput Chem, 2011     

Force matching as a stepping stone to QM/MM CB[8] host/guest binding free energies: a SAMPL6 cautionary tale

Use of quantum mechanical/molecular mechanical (QM/MM) methods in binding free energy calculations, particularly in the SAMPL challenge, often fail to achieve improvement over standard additive (MM) force fields. Frequently, the implementation is through use of reference potentials, or the so-called “indirect approach”, and inherently relies on sufficient overlap existing between MM and QM/MM configurational spaces. This overlap is generally poor, particularly for the use of free energy perturbation to perform the MM to QM/MM free energy correction at the end states of interest (e.g., bound and unbound states). However, by utilizing MM parameters that best reproduce forces obtained at the desired QM level of theory, it is possible to lessen the configurational disparity between MM and QM/MM. To this end, we sought to use force matching to generate MM parameters for the SAMPL6 CB[8] host–guest binding challenge, classically compute binding free energies, and apply energetic end state corrections to obtain QM/MM binding free energy differences. For the standard set of 11 molecules and the bonus set (including three additional challenge molecules), error statistics, such as the root mean square deviation (RMSE) were moderately poor (5.5 and 5.4 kcal/mol). Correlation statistics, however, were in the top two for both standard and bonus set submissions (\(R^{2}\) of 0.42 and 0.26, \(\tau\) of 0.64 and 0.47 respectively). High RMSE and moderate correlation strongly indicated the presence of systematic error. Identifiable issues were ameliorated for two of the guest molecules, resulting in a reduction of error and pointing to strong prospects for the future use of this methodology.