Combining Force Fields and Neural Networks for an Accurate Representation of Bonded Interactions.

We present a formalism of a neural network encoding bonded interactions in molecules. This intramolecular encoding is consistent with the models of intermolecular interactions previously designed by this group. Variants of the encoding fed into a corresponding neural network may be used to economically improve the representation of torsional degrees of freedom in any force field. We test the accuracy of the reproduction of the ab initio potential energy surface on a set of conformations of two dipeptides, methyl-capped ALA and ASP, in several scenarios. The encoding, either alone or in conjunction with an analytical potential, improves agreement with ab initio energies that are on par with those of other neural network-based potentials. Using the encoding and neural nets in tandem with an analytical model places the agreements firmly within "chemical accuracy" of ±0.5 kcal/mol.

Combining Force Fields and Neural Networks for an Accurate Representation of Chemically Diverse Molecular Interactions.

A key goal of molecular modeling is the accurate reproduction of the true quantum mechanical potential energy of arbitrary molecular ensembles with a tractable classical approximation. The challenges are that analytical expressions found in general purpose force fields struggle to faithfully represent the intermolecular quantum potential energy surface at close distances and in strong interaction regimes; that the more accurate neural network approximations do not capture crucial physics concepts, e.g., nonadditive inductive contributions and application of electric fields; and that the ultra-accurate narrowly targeted models have difficulty generalizing to the entire chemical space. We therefore designed a hybrid wide-coverage intermolecular interaction model consisting of an analytically polarizable force field combined with a short-range neural network correction for the total intermolecular interaction energy. Here, we describe the methodology and apply the model to accurately determine the properties of water, the free energy of solvation of neutral and charged molecules, and the binding free energy of ligands to proteins. The correction is subtyped for distinct chemical species to match the underlying force field, to segment and reduce the amount of quantum training data, and to increase accuracy and computational speed. For the systems considered, the hybrid ab initio parametrized Hamiltonian reproduces the two-body dimer quantum mechanics (QM) energies to within 0.03 kcal/mol and the nonadditive many-molecule contributions to within 2%. Simulations of molecular systems using this interaction model run at speeds of several nanoseconds per day.

On the importance of accounting for nuclear quantum effects in ab initio calibrated force fields in biological simulations.

In many important processes in chemistry, physics, and biology the nuclear degrees of freedom cannot be described using the laws of classical mechanics. At the same time, the vast majority of molecular simulations that employ wide-coverage force fields treat atomic motion classically. In light of the increasing desire for and accelerated development of quantum mechanics (QM)-parameterized interaction models, we reexamine whether the classical treatment is sufficient for a simple but crucial chemical species: alkanes. We show that when using an interaction model or force field in excellent agreement with the "gold standard" QM data, even very basic simulated properties of liquid alkanes, such as densities and heats of vaporization, deviate significantly from experimental values. Inclusion of nuclear quantum effects via techniques that treat nuclear degrees of freedom using the laws of classical mechanics brings the simulated properties much closer to reality.

Epoxidation Catalyzed by the Nonheme Iron(II)- and 2-Oxoglutarate-Dependent Oxygenase, AsqJ: Mechanistic Elucidation of Oxygen Atom Transfer by a Ferryl Intermediate.

Mechanisms of enzymatic epoxidation via oxygen atom transfer (OAT) to an olefin moiety is mainly derived from the studies on thiolate-heme containing epoxidases, such as cytochrome P450 epoxidases. The molecular basis of epoxidation catalyzed by nonheme-iron enzymes is much less explored. Herein, we present a detailed study on epoxidation catalyzed by the nonheme iron(II)- and 2-oxoglutarate-dependent (Fe/2OG) oxygenase, AsqJ. The native substrate and analogues with different para substituents ranging from electron-donating groups (e.g., methoxy) to electron-withdrawing groups (e.g., trifluoromethyl) were used to probe the mechanism. The results derived from transient-state enzyme kinetics, Mössbauer spectroscopy, reaction product analysis, X-ray crystallography, density functional theory calculations, and molecular dynamic simulations collectively revealed the following mechanistic insights: (1) The rapid O2 addition to the AsqJ Fe(II) center occurs with the iron-bound 2OG adopting an online-binding mode in which the C1 carboxylate group of 2OG is trans to the proximal histidine (His134) of the 2-His-1-carboxylate facial triad, instead of assuming the offline-binding mode with the C1 carboxylate group trans to the distal histidine (His211); (2) The decay rate constant of the ferryl intermediate is not strongly affected by the nature of the para substituents of the substrate during the OAT step, a reactivity behavior that is drastically different from nonheme Fe(IV)-oxo synthetic model complexes; (3) The OAT step most likely proceeds through a stepwise process with the initial formation of a C(benzylic)-O bond to generate an Fe-alkoxide species, which is observed in the AsqJ crystal structure. The subsequent C3-O bond formation completes the epoxide installation.

Crystal Structure of Cleaved Serp-1, a Myxomavirus-Derived Immune Modulating Serpin: Structural Design of Serpin Reactive Center Loop Peptides with Improved Therapeutic Function.

The Myxomavirus-derived protein Serp-1 has potent anti-inflammatory activity in models of vasculitis, lupus, viral sepsis, and transplant. Serp-1 has also been tested successfully in a Phase IIa clinical trial in unstable angina, representing a "first-in-class" therapeutic. Recently, peptides derived from the reactive center loop (RCL) have been developed as stand-alone therapeutics for reducing vasculitis and improving survival in MHV68-infected mice. However, both Serp-1 and the RCL peptides lose activity in MHV68-infected mice after antibiotic suppression of intestinal microbiota. Here, we utilize a structure-guided approach to design and test a series of next-generation RCL peptides with improved therapeutic potential that is not reduced when the peptides are combined with antibiotic treatments. The crystal structure of cleaved Serp-1 was determined to 2.5 Å resolution and reveals a classical serpin structure with potential for serpin-derived RCL peptides to bind and inhibit mammalian serpins, plasminogen activator inhibitor 1 (PAI-1), anti-thrombin III (ATIII), and α-1 antitrypsin (A1AT), and target proteases. Using in silico modeling of the Serp-1 RCL peptide, S-7, we designed several modified RCL peptides that were predicted to have stronger interactions with human serpins because of the larger number of stabilizing hydrogen bonds. Two of these peptides (MPS7-8 and -9) displayed extended activity, improving survival where activity was previously lost in antibiotic-treated MHV68-infected mice (P < 0.0001). Mass spectrometry and kinetic assays suggest interaction of the peptides with ATIII, A1AT, and target proteases in mouse and human plasma. In summary, we present the next step toward the development of a promising new class of anti-inflammatory serpin-based therapeutics.

Insights into the Desaturation of Cyclopeptin and its C3 Epimer Catalyzed by a non-Heme Iron Enzyme: Structural Characterization and Mechanism Elucidation.

AsqJ, an iron(II)- and 2-oxoglutarate-dependent enzyme found in viridicatin-type alkaloid biosynthetic pathways, catalyzes sequential desaturation and epoxidation to produce cyclopenins. Crystal structures of AsqJ bound to cyclopeptin and its C3 epimer are reported. Meanwhile, a detailed mechanistic study was carried out to decipher the desaturation mechanism. These findings suggest that a pathway involving hydrogen atom abstraction at the C10 position of the substrate by a short-lived FeIV -oxo species and the subsequent formation of a carbocation or a hydroxylated intermediate is preferred during AsqJ-catalyzed desaturation.

Computational modeling and in-vitro/in-silico correlation of phospholipid-based prodrugs for targeted drug delivery in inflammatory bowel disease.

Targeting drugs to the inflamed intestinal tissue(s) represents a major advancement in the treatment of inflammatory bowel disease (IBD). In this work we present a powerful in-silico modeling approach to guide the molecular design of novel prodrugs targeting the enzyme PLA2, which is overexpressed in the inflamed tissues of IBD patients. The prodrug consists of the drug moiety bound to the sn-2 position of phospholipid (PL) through a carbonic linker, aiming to allow PLA2 to release the free drug. The linker length dictates the affinity of the PL-drug conjugate to PLA2, and the optimal linker will enable maximal PLA2-mediated activation. Thermodynamic integration and Weighted Histogram Analysis Method (WHAM)/Umbrella Sampling method were used to compute the changes in PLA2 transition state binding free energy of the prodrug molecule (∆∆Gtr) associated with decreasing/increasing linker length. The simulations revealed that 6-carbons linker is the optimal one, whereas shorter or longer linkers resulted in decreased PLA2-mediated activation. These in-silico results were shown to be in excellent correlation with experimental in-vitro data. Overall, this modern computational approach enables optimization of the molecular design of novel prodrugs, which may allow targeting the free drug specifically to the diseased intestinal tissue of IBD patients.

Prediction of cyclohexane-water distribution coefficient for SAMPL5 drug-like compounds with the QMPFF3 and ARROW polarizable force fields.

We present the performance of blind predictions of water-cyclohexane distribution coefficients for 53 drug-like compounds in the SAMPL5 challenge by three methods currently in use within our group. Two of them utilize QMPFF3 and ARROW, polarizable force-fields of varying complexity, and the third uses the General Amber Force-Field (GAFF). The polarizable FF's are implemented in an in-house MD package, Arbalest. We find that when we had time to parametrize the functional groups with care (batch 0), the polarizable force-fields outperformed the non-polarizable one. Conversely, on the full set of 53 compounds, GAFF performed better than both QMPFF3 and ARROW. We also describe the torsion-restrain method we used to improve sampling of molecular conformational space and thus the overall accuracy of prediction. The SAMPL5 challenge highlighted several drawbacks of our force-fields, such as our significant systematic over-estimation of hydrophobic interactions, specifically for alkanes and aromatic rings.

Spectroscopic and computational studies of a bifunctional iron- and 2-oxoglutarate dependent enzyme, AsqJ.

Iron and 2-oxoglutarate dependent (Fe/2OG) enzymes exhibit an exceedingly broad reaction repertoire. The most prevalent reactivity is hydroxylation, but many other reactivities have also been discovered in recent years, including halogenation, desaturation, epoxidation, endoperoxidation, epimerization, and cyclization. To fully explore the reaction mechanisms that support such a diverse reactivities in Fe/2OG enzyme, it is necessary to utilize a multi-faceted research methodology, consisting of molecular probe design and synthesis, in vitro enzyme assay development, enzyme kinetics, spectroscopy, protein crystallography, and theoretical calculations. By using such a multi-faceted research approach, we have explored reaction mechanisms of desaturation and epoxidation catalyzed by a bi-functional Fe/2OG enzyme, AsqJ. Herein, we describe the experimental protocols and computational workflows used in our studies.

Neural Network Corrections to Intermolecular Interaction Terms of a Molecular Force Field Capture Nuclear Quantum Effects in Calculations of Liquid Thermodynamic Properties.

We incorporate nuclear quantum effects (NQE) in condensed matter simulations by introducing short-range neural network (NN) corrections to the ab initio fitted molecular force field ARROW. Force field NN corrections are fitted to average interaction energies and forces of molecular dimers, which are simulated using the Path Integral Molecular Dynamics (PIMD) technique with restrained centroid positions. The NN-corrected force field allows reproduction of the NQE for computed liquid water and methane properties such as density, radial distribution function (RDF), heat of evaporation (HVAP), and solvation free energy. Accounting for NQE through molecular force field corrections circumvents the need for explicit computationally expensive PIMD simulations in accurate calculations of the properties of chemical and biological systems. The accuracy and locality of pairwise NN NQE corrections indicate that this approach could be applicable to complex heterogeneous systems, such as proteins.

The Determination of Free Energy of Hydration of Water Ions from First Principles.

We model the autoionization of water by determining the free energy of hydration of the major intermediate species of water ions. We represent the smallest ions─the hydroxide ion OH-, the hydronium ion H3O+, and the Zundel ion H5O2+─by bonded models and the more extended ionic structures by strong nonbonded interactions (e.g., the Eigen H9O4+ = H3O+ + 3(H2O) and the Stoyanov H13O6+ = H5O2+ + 4(H2O)). Our models are faithful to the precise QM energies and their components to within 1% or less. Using the calculated free energies and atomization energies, we compute the pKa of pure water from first principles as a consistency check and arrive at a value within 1.3 log units of the experimental one. From these calculations, we conclude that the hydronium ion, and its hydrated state, the Eigen cation, are the dominant species in the water autoionization process.

Accurate determination of solvation free energies of neutral organic compounds from first principles.

The main goal of molecular simulation is to accurately predict experimental observables of molecular systems. Another long-standing goal is to devise models for arbitrary neutral organic molecules with little or no reliance on experimental data. While separately these goals have been met to various degrees, for an arbitrary system of molecules they have not been achieved simultaneously. For biophysical ensembles that exist at room temperature and pressure, and where the entropic contributions are on par with interaction strengths, it is the free energies that are both most important and most difficult to predict. We compute the free energies of solvation for a diverse set of neutral organic compounds using a polarizable force field fitted entirely to ab initio calculations. The mean absolute errors (MAE) of hydration, cyclohexane solvation, and corresponding partition coefficients are 0.2 kcal/mol, 0.3 kcal/mol and 0.22 log units, i.e. within chemical accuracy. The model (ARROW FF) is multipolar, polarizable, and its accompanying simulation stack includes nuclear quantum effects (NQE). The simulation tools' computational efficiency is on a par with current state-of-the-art packages. The construction of a wide-coverage molecular modelling toolset from first principles, together with its excellent predictive ability in the liquid phase is a major advance in biomolecular simulation.

Protein-Ligand Binding Free-Energy Calculations with ARROW─A Purely First-Principles Parameterized Polarizable Force Field.

Protein-ligand binding free-energy calculations using molecular dynamics (MD) simulations have emerged as a powerful tool for in silico drug design. Here, we present results obtained with the ARROW force field (FF)─a multipolar polarizable and physics-based model with all parameters fitted entirely to high-level ab initio quantum mechanical (QM) calculations. ARROW has already proven its ability to determine solvation free energy of arbitrary neutral compounds with unprecedented accuracy. The ARROW FF parameterization is now extended to include coverage of all amino acids including charged groups, allowing molecular simulations of a series of protein-ligand systems and prediction of their relative binding free energies. We ensure adequate sampling by applying a novel technique that is based on coupling the Hamiltonian Replica exchange (HREX) with a conformation reservoir generated via potential softening and nonequilibrium MD. ARROW provides predictions with near chemical accuracy (mean absolute error of ∼0.5 kcal/mol) for two of the three protein systems studied here (MCL1 and Thrombin). The third protein system (CDK2) reveals the difficulty in accurately describing dimer interaction energies involving polar and charged species. Overall, for all of the three protein systems studied here, ARROW FF predicts relative binding free energies of ligands with a similar accuracy level as leading nonpolarizable force fields.

Harnessing the Substrate Promiscuity of Dioxygenase AsqJ and Developing Efficient Chemoenzymatic Synthesis for Quinolones.

Nature has developed complexity-generating reactions within natural product biosynthetic pathways. However, direct utilization of these pathways to prepare compound libraries remains challenging due to limited substrate scopes, involvement of multiple-step reactions, and moderate robustness of these sophisticated enzymatic transformations. Synthetic chemistry, on the other hand, offers an alternative approach to prepare natural product analogs. However, owing to complex and diverse functional groups appended on the targeted molecules, dedicated design and development of synthetic strategies are typically required. Herein, by leveraging the power of chemo-enzymatic synthesis, we report an approach to bridge the gap between biological and synthetic strategies in the preparation of quinolone alkaloid analogs. Leading by in silico analysis, the predicted substrate analogs were chemically synthesized. The AsqJ-catalyzed asymmetric epoxidation of these substrate analogues was followed by an Lewis Acid-triggered ring contraction to complete the viridicatin formation. We evaluated the robustness of this method in gram-scale reactions. Lastly, through chemoenzymatic cascades, a library of quinolone alkaloids is effectively prepared.

Electron tunneling through sensitizer wires bound to proteins.

We report a quantitative theoretical analysis of long-range electron transfer through sensitizer wires bound in the active-site channel of cytochrome P450cam. Each sensitizer wire consists of a substrate group with high binding affinity for the enzyme active site connected to a ruthenium-diimine through a bridging aliphatic or aromatic chain. Experiments have revealed a dramatic dependence of electron transfer rates on the chemical composition of both the bridging group and the substrate. Using combined molecular dynamics simulations and electronic coupling calculations, we show that electron tunneling through perfluorinated aromatic bridges is promoted by enhanced superexchange coupling through virtual reduced states. In contrast, electron flow through aliphatic bridges occurs by hole-mediated superexchange. We have found that a small number of wire conformations with strong donor-acceptor couplings can account for the observed electron tunneling rates for sensitizer wires terminated with either ethylbenzene or adamantane. In these instances, the rate is dependent not only on electronic coupling of the donor and acceptor but also on the nuclear motion of the sensitizer wire, necessitating the calculation of average rates over the course of a molecular dynamics simulation. These calculations along with related recent findings have made it possible to analyze the results of many other sensitizer-wire experiments that in turn point to new directions in our attempts to observe reactive intermediates in the catalytic cycles of P450 and other heme enzymes.

Kinetic and theoretical comprehension of diverse rate laws and reactivity gaps in Coriolus hirsutus laccase-catalyzed oxidation of acido and cyclometalated Ru(II) complexes.

The reactivity of the acido Ru(II) complexes cis-[RuCl(2)(LL)(2)], [RuCO(3)(LL)(2)], cis-[RuCO(3)-(bquin)(2)] (LL = 2,2'-bipyridine (bpy) and 1,10-phenanthroline (phen); bquin = 2,2'-biquinoline) and cyclometalated Ru(II) derivatives of 2-phenylpyridine and 4-(2-tolyl)pyridine [Ru(o-C(6)H(4)-2-py)(phen)(2)]PF(6) (1), [Ru(o-C(6)H(3)-p-R-2-py)(bpy)(MeCN)(2)]PF(6) (2), and [Ru(o-C(6)H(3)-p-R-2-py)(phen)(MeCN)(2)]PF(6) (3) (R = H (a), Me (b)) toward laccase from Coriolus hirsutus has been investigated by conventional UV-vis spectroscopy at pH 3-7 and 25 degrees C. The acido and cyclometalated complexes are readily oxidized into the corresponding Ru(III) species, but the two types of complexes differ substantially in reactivity and obey different rate laws. The acido complexes are oxidized more slowly and the second-order kinetics, first-order in laccase and Ru(II), holds with the rate constants around 5 x 10(4) M(-1) s(-1) at pH 4.5 and 25 degrees C. The cyclometalated complexes 1-3 react much faster and the hyperbolic Michaelis-Menten kinetics holds. However, it is not due to formation of an enzyme-substrate complex but rather because of the ping-pong mechanism of catalysis, viz. E(ox) + Ru(II) --> E(red) + Ru(III) (k(1)); E(red) + 1/4O(2) --> E(ox) (k(2)), with the rate constants k(1) in the range (2-9) x 10(7) M(-1) s(-1) under the same conditions. The huge values of k(1) move the enzymatic oxidation toward a kinetic regime when the dioxygen half-reaction becomes the rate-limiting step. Cyclometalated compounds 1-3 can therefore be used for routine estimation of k(2), that is, the rate constant for reoxidation for laccases by dioxygen. The mechanism proposed was confirmed by the direct stopped-flow measurements of the k(2) rate constant (8.1 x 10(5) M(-1) s(-1) at 26 degrees C) and supported by the theoretical modeling of interaction between the bpy analogue of 1 and Coriolus hirsutes laccase using Monte Carlo simulations.

Phospholipid-Based Prodrugs for Colon-Targeted Drug Delivery: Experimental Study and In-Silico Simulations.

In ulcerative colitis (UC), the inflammation is localized in the colon, and one of the successful strategies for colon-targeting drug delivery is the prodrug approach. In this work, we present a novel phospholipid (PL)-based prodrug approach, as a tool for colonic drug targeting in UC. We aim to use the phospholipase A2 (PLA2), an enzyme that is overexpressed in the inflamed colonic tissues of UC patients, as the PL-prodrug activating enzyme, to accomplish the liberation of the parent drug from the prodrug complex at the specific diseased tissue(s). Different linker lengths between the PL and the drug moiety can dictate the rate of activation by PLA2, and subsequently determine the amount of free drugs at the site of action. The feasibility of this approach was studied with newly synthesized PL-Fmoc (fluorenylmethyloxycarbonyl) conjugates, using Fmoc as a model compound for testing our hypothesis. In vitro incubation with bee venom PLA2 demonstrated that a 7-carbon linker between the PL and Fmoc has higher activation rate than a 5-carbon linker. 4-fold higher colonic expression of PLA2 was demonstrated in colonic mucosa of colitis-induced rats when compared to healthy animals, validating our hypothesis of a colitis-targeting prodrug approach. Next, a novel molecular dynamics (MD) simulation was developed for PL-based prodrugs containing clinically relevant drugs. PL-methotrexate conjugate with 6-carbon linker showed the highest extent of PLA2-mediated activation, whereas shorter linkers were activated to a lower extent. In conclusion, this work demonstrates that for carefully designed PL-drug conjugates, PLA2 overexpression in inflamed colonic tissues can be used as prodrug-activating enzyme and drug targeting strategy, including insights into the activation mechanisms in a PLA2 binding site.

Cardiolipin deficiency leads to decreased cardiolipin peroxidation and increased resistance of cells to apoptosis.

Cardiolipin (CL), a unique mitochondrial phospholipid synthesized by CL synthase (CLS), plays important, yet not fully understood, roles in mitochondria-dependent apoptosis. We manipulated CL levels in HeLa cells by knocking down CLS using RNA interference and selected a clone of CL-deficient cells with approximately 45% of its normal content. ESI-MS analysis showed that the CL molecular species were the same in CL-deficient and CL-sufficient cells. CL deficiency did not change mitochondrial functions (membrane potential, reactive oxygen species generation, cellular ATP levels) but conferred resistance to apoptosis induced by actinomycin D (ActD), rotenone, or gamma-irradiation. During ActD-induced apoptosis, decreased CL peroxidation along with suppressed cytochrome (cyt) c release was observed in CL-deficient cells, whereas Bax translocation to mitochondria remained similar to that in CL-sufficient HeLa cells. The amounts of loosely bound cyt c (releasable under high ionic strength conditions) were the same in CL-deficient and CL-sufficient cells. Given that CL peroxidation during apoptosis is catalyzed by CL/cyt c complexes and CL oxidation products are essential for cyt c release from mitochondria, our results suggest that CL deficiency prevents adequate assembly of productive CL/cyt c complexes and CL peroxidation, resulting in increased resistance to apoptosis.

Phospholipid-Based Prodrugs for Drug Targeting in Inflammatory Bowel Disease: Computational Optimization and In-Vitro Correlation.

In inflammatory bowel disease (IBD) patients, the enzyme phospholipase A2 (PLA2) is overexpressed in the inflamed intestinal tissue, and hence may be exploited as a prodrug-activating enzyme allowing drug targeting to the site(s) of gut inflammation. The purpose of this work was to develop powerful modern computational approaches, to allow optimized a-priori design of phospholipid (PL) based prodrugs for IBD drug targeting. We performed simulations that predict the activation of PL-drug conjugates by PLA2 with both human and bee venom PLA2. The calculated results correlated well with in-vitro experimental data. In conclusion, a-priori drug design using a computational approach complements and extends experimentally derived data, and may improve resource utilization and speed drug development.