Search

Force fields have been an integral part of computational chemistry for decades, providing invaluable insight and facilitating the better understanding of biomolecular system behaviour. Despite the many benefits of a force field, there continue to be deficiencies as a result of the classical architecture they are based upon. Some deficiencies, such as a point charge electrostatic description instead of a multipole moment description, have been addressed over time, permitted by the ever-increasing computational power available. However, whilst incorporating such significant improvements has improved force field accuracy, many still fail to describe several chemical effects including polarisation, non-covalent interactions and secondary/tertiary structural effects. Furthermore, force fields often fail to provide consistency when compared with other force fields. In other words, no force field is reliably performing more accurately than others, when applied to a variety of related problems. The work presented herein develops a next-generation force field entitled FFLUX, which features a novel architecture very different to any other force field. FFLUX is designed to capture the relationship between geometry and energy through a machine learning method known as kriging. Instead of a series of parameterised potentials, FFLUX uses a collection of atomic energy kriging models to make energy predictions. The energies describing atoms within FFLUX are obtained from the Interacting Quantum Atoms (IQA) energy partitioning approach, which in turn derives the energies from the electron density and nuclear charges of topological atoms described by Quantum Chemical Topology (QCT). IQA energies are shown to provide a unique insight into the relationship between geometry and energy, allowing the identification of explicit atoms and energies contributing towards torsional barriers within various systems. The IQA energies can be modelled to within 2.6% accuracy, as shown for a series of small systems including weakly bound complexes. The energies also allow an interpretation of how an atom feels its surrounding environment through intra-atomic, covalent and electrostatic energetic descriptions, which typically are seen to converge within a ~7 - 8 A horizon radius around an atom or small system. These energy convergence results are particularly relevant to tackling the transferability theme within force field development. Where energies are seen to converge, a proximity limit on the geometrical description needed for a transferable energy model is defined. Finally, the FFLUX force field is validated through successfully optimising distorted geometries of a series of small molecules, to near-ab initio accuracy.

Until such a time that computers are powerful enough to routinely perform ab initio simulation of large biomolecules, there will remain a demand for less expensive computational tools. Classical force field methods are widely used for the simulation of large molecules. However, their low computational cost comes at the price of introducing approximations to the description of the system, for example atomic point charges and Hooke type potentials. The quantum chemical topological force field, QCTFF, removes the classical approximations and uses a machine learning method, kriging, to build models that map ab initio atomic properties to changes in the internal coordinates of a chemical system. The atomic properties come from quantum chemical topology, QCT, and include atomic multipole moments and also energy terms from the interacting quantum atoms (IQA) energy decomposition scheme. By using atomic multipole moments, the electrostatic interactions between atoms is described in a more rigorous fashion than most classical force fields, and polarisation is captured through the use of kriging models. In this thesis, the QCTFF approach has been applied to a selection of test cases including small molecular dimers and amino acids. Kriging models are built using a “training set” of molecular geometries, and an investigation of different approaches for sampling amino acids is provided. The concept of the “atomic horizon sphere” is discussed, where the effect on the multipole moments of an atom in an increasingly large environment is investigated. This is an important investigation required to guide the development of future QCTFF training sets. Investigations into the effect of deprotonation of basic and acidic amino acids side chains is provided, as well as a study of the short range repulsion between atoms.

A molecule can crystallise in more than one crystal structure, a common phenomenon in organic compounds known as polymorphism. Different polymorphic forms may have significantly different physical properties, and a reliable prediction would be beneficial to the pharmaceutical industry. However, crystal structure prediction (CSP) based on the knowledge of the chemical structure had long been considered impossible. Previous failures of some CSP attempts led to speculation that the thermodynamic calculations in CSP methodologies failed to predict the kinetically favoured structures. Similarly, regarding the stabilities of co-crystals relative to their pure components, the results from lattice energy calculations and full CSP studies were inconclusive. In this thesis, these problems are addressed using the state-of-the-art CSP methodology implemented in the GRACE software. Firstly, it is shown that the low-energy predicted structures of four organic molecules, which have previously been considered difficult for CSP, correspond to their experimental structures. The possible outcomes of crystallisation can be reliably predicted by sufficiently accurate thermodynamic calculations. Then, the polymorphism of 5- chloroaspirin is investigated theoretically. The order of polymorph stability is predicted correctly and the isostructural relationships between a number of predicted structures and the experimental structures of other aspirin derivatives are established. Regarding the stabilities of co-crystals, 99 out of 102 co-crystals and salts of nicotinamide, isonicotinamide and picolinamide reported in the Cambridge Structural Database (CSD) are found to be more stable than their corresponding co-formers. Finally, full CSP studies of two co-crystal systems are conducted to explain why the co-crystals are not easily obtained experimentally.

Force field methods are used to investigate the properties of a wide variety of chemical systems on a routine basis. The expression for the electrostatic energy typically does not take into account the anisotropic nature of the atomic electron distribution or the dependence of that distribution on the system geometry. This has been suggested as a cause of the failure of force field methods to reliably predict the behaviour of chemical systems. A method for incorporation of anisotropy and polarisation is described in this work. Anisotropy is modelled by the inclusion of multipole moments centred at atoms whose values are determined by application of the methods of Quantum Chemical Topology. Polarisation, the dependence of the electron distribution on system geometry, is modelled by training machine learning models to predict atomic multipole moments from knowledge of the nuclear positions of a system. The resulting electrostatic method can be implemented for any chemical system. An application to progressively more complex systems is reported, including small organic molecules and larger molecules of biological importance. The accuracy of the method is rigorously assessed by comparison of its predictions to exact interaction energy values. A procedure for generating transferable atomic multipole moment models is defined and tested. The electrostatic method can be combined with the empirical expressions used in force field calculations to describe total system energies by fitting parameters against ab initio conformational energies. Derivatives of the energy are given and the resulting multipolar polarisable force field can be used to perform geometry optimisation calculations. Future applications to conformational searching and problems requiring dynamic descriptions of a system are feasible.

Exploratory chemistry is an important branch of computational chemistry where precursor molecules are simulated in reaction-prone conditions without pre-supposed hypotheses. These molecular dynamics simulations can yield many unexpected reaction pathways, confirming existing mechanisms or suggesting new ones. Decreasing cost of hardware and increasing savviness of software make it enticing to explore reaction spaces in this un-guided manner. However, reactions are still rare events buried in data where the molecules spend most of their time not clearing reaction barriers. Therefore, advanced data processing techniques that can parse the simulations for reaction events with high accuracy are valuable ina field with ever increasing data generation.In this work, a method is described in Chapter 2 which addresses the problem of extracting valuable reaction location data from exploratory simulations with high accuracy and low cost. The method uses first derivatives on bond order time series obtained from Ab Initio molecular dynamics data to predict temporal reaction locations with high accuracy and speed. The two tunable parameters (low pass filtering cutoff and threshold placement on the derivative) are parameterized against a “gold standard” approach that represents the ideal data processing scenario given infinite time and computing resources. This reaction event classifier is showcased on two single-molecule reactive test systems. The first, simpler test case being a heptanylium alkyl carbocation, C7H15+. The other more complicated system is Fe3(CO)9, which is a highly unsaturated iron carbonyl cluster. The effectiveness of the reaction event classifier is analyzed for both systems using heat maps and other custom plots and metrics.Deep Eutectic Solvents (DESs) are a relatively new type of mixture which exhibits similar properties to ionic liquids, but with lower cost and environmental impact. Until recently, most force fields describing DESs were not polarizable and therefore lacked certain behaviors created by their extremely partially-charged environment. This second project began with the goal of creating a polarizable force field for DESs, but eventually morphed into creating a program to automate the Tinker polarizable force field initialization process with Python. The test molecule for refining the initialization process is urea. Urea is a hydrogen bond donor which when paired with choline chloride in the correct whole-number ratio create a DES. A semi-automated Python method which creates polarizable Tinker force fields with simple setup is explained.

Pantetheine-containing ligands (PCLs) play key roles in the biosynthesis of polyketides, a large family of natural products with various bioactivities. A major hurdle that remains is our poor understanding of the protein-substrate interactions of ketoreductase (KR), a key component of polyketide synthases (PKSs). Since the poly-β-ketone intermediates are highly reactive and cannot be isolated, the first project of this dissertation employed molecular dynamics (MD) simulations to interpret the transient KR-substrate interactions. Several key factors guiding KR-substrate interactions were identified, which will help further engineering of PKSs and directing biosynthesis of novel polyketides.The reliability of MD simulations depends on the quality of the employed force field, which comprises a mathematical formula and a set of parameters to represent the potential energy of molecular systems. The parameter sets for simulating amino acids, nucleic acids, sugars, and lipids are already available. However, lack of scalable parameter sets for PCLs has hampered the computational studies of various biomolecules containing PCLs. Therefore, in the second project of this dissertation, the first Pantetheine Force Field (PFF) library containing parameter sets for various PCLs compatible with additive force fields was developed and validated.The extensively used additive force fields use fixed atom-centered partial charges to model atomic electrostatic interactions. However, additive force fields cannot accurately model atomic polarization effects, leading to unrealistic simulations in polarization-sensitive processes. The polarizable Gaussian Multipole (pGM) model with all atomic multipoles represented by Gaussian densities has been recently developed. In the third project of this dissertation, an electrostatic parameterization scheme for the pGM model was developed, and the accuracy and transferability of the pGM model were assessed. Encouragingly, the pGM model was shown to be accurate and transferable, which has the potential to serve as the template for developing the next-generation polarizable force field for modeling various biological systems.

Computer simulation methods have the power to greatly optimize drug design, reducing the time and cost that it takes to bring a new drug from lab to market. Two factors of concern in rational drug design are the binding affinity of some small molecule to a biomolecular target and the molecule's ability to permeate the cellular membrane to reach said target. In the first half of this work, I describe my efforts analyzing the protein-ligand binding affinity and membrane permeability of inhibitors of the Hv1 proton channel. Hv1 is a voltage-gated proton channel with clinical relevance, being highly expressed in highly metastatic cancer cell lines. The inhibition of Hv1 has been found to slow the growth of cancer cells. My research on Hv1 channel inhibitors lays the groundwork for future studies on adapting these molecules for more effective channel blocking.In the second half of this work, I examine the tools that are foundational to molecular simulations--force fields--as well as explore the conventional wisdom behind proton assignment in carboxylic acids. Biomolecular simulation represents a valuable computational microscope for drug discovery efforts; however, it is only through correct setup protocols and well-calibrated tools that we can obtain the most accurate and meaningful results.

Aqueous-phase heterogeneous catalysis is an important chemical process in applications such as water remediation, fuel cells, and the production of fuels and chemicals, including from biomass sources. However, designing alternative, improved catalyst materials for these applications is difficult due to fluctuations in the solvation environment surrounding the catalytic species. In order to elucidate the thermodynamics and kinetics of the relevant reactions, it is imperative to gain a better understanding of the roles of liquid water molecules in these reactions. In this work, a method combining both classical and quantum simulations was developed to generate configurations of liquid water molecules over catalytic species adsorbed on a catalyst surface, which can provide valuable insight into the roles of the liquid water reaction environment on aqueous-phase heterogeneous catalysis. The method developed in this work entails combining force field molecular dynamics (FFMD) and density functional theory (DFT) simulations. This method leverages the strengths of each type of simulation to enable the calculation of catalytic energies under “realistic” liquid water configurations. FFMD simulations are used to generate trajectories of liquid water configurations that include thermal fluctuations, while DFT simulations are used to capture the energies associated with bond breaking and forming that are required for microkinetic modeling and catalyst design studies. This FFMD-DFT method was used to calculate the interaction energies between the liquid water environment and the reaction intermediate or transition state species. The trend in the calculated interaction energies was shown to correlate with the trend in hydrogen-bond formation between liquid water molecules and the catalytic species. This work also demonstrated that entropic effects due to the thermal fluctuations in the solvation environment are a significant contribution to the free energies calculated for aqueous-phase, heterogeneously-catalyzed systems. The FFMD-DFT method was also used to calculate reaction energies, activation barriers, and pre-exponential factors to study the kinetics of example O—H and C—H cleavage reactions on a platinum catalyst surface under an aqueous reaction environment. Using this method, it was found that O—H cleavage reactions prefer H2O-mediated pathways, while C—H cleavage reactions prefer non-H2O-mediated pathways. In summary, the FFMD-DFT method developed in this work has been shown to be a robust technique for generating realistic liquid water configurations over catalytic species on a platinum catalyst surface. Those liquid water configurations can be used to calculate catalytic properties that can provide insight into the roles of water molecules in these reactions and facilitate microkinetic modeling and catalyst design studies.

The early stages of drug discovery is a long and costly process. A way to reduce the time, resources, and financial investment spent is to apply computational tools. With the tremendous progress and achievements in computational chemistry, the application of computational tools in the drug lead discovery and design has increased. Today, computational chemistry is considered a highly valuable and well established tool in drug discovery. A variety of computational chemistry methods are used for guiding molecular design and finding new potential drugs and targets. Some examples of these methods are molecular dynamics simulations, free energy calculations, virtual screening, structure-activity relationship analysis, and so on. Despite the fact that computational chemistry technics are widely used in industry and academic environment, there is still room for improvement. Here, I present several studies in which I developed and applied computational chemistry tools in drug discovery problems. The first study is the development of a tool as part of the Open Force Field consortium to learn chemical perception of force fields typing rules. Secondly, I describe my work on using a new hybrid method to calculate free energies of small molecules. Thirdly, I present a binding mode prediction study to help the understanding of structure-activity relationship in lissoclimides. Lastly, I present a study applying molecular dynamic simulation to guide the redesigning of a macrocyclic peptide.

In this thesis, I have used computational methods to the study of flexibility in metal-organic frameworks (MOFs) and the behavior of adsorbed gases and liquids in MOFs. Among nanoporous materials, MOFs are of particular interest for adsorption-based applications for reasons that include (1) their huge chemical space, which raises the possibility of tuning them for specific industrial processes and (2) the ability of the under-coordinated metals present in some MOFs to endow their respective materials with impressive adsorbate binding energies.Chapters 2 and 4 relate to analogs of a series of MOFs with under-coordinated metals (the M-MOF-74 series). In Chapter 2, I report the discovery of a novel argon adsorbate-induced deformation pattern for this framework series. This result was arrived at using a flexible framework model, and is presented as an explanation for an intriguing signal observed in experimental small-angle X-ray scattering profiles upon argon adsorption. This hypothesis is supported by a complementary investigation of an alternative explanation for the X-ray signal, in which argon atoms are proposed to adsorb at different densities in adjacent pores.In Chapter 3, I studied the dynamics of adsorbed xylene isomers in MOF-5. I compare the translational and rotational motion of the xylene isomers, and explain the differences based on molecular geometries. The most rod-like molecule, para-xylene, is more rotationally constricted in the pore due to its intermolecular interactions with the aromatic group of the MOF linker. This finding has implications for the rational design of MOFs, as a process that can take advantage of the MOF-induced variations in xylene isomer dynamics could be used for lucrative liquid-phase xylene separations.In Chapter 4, I present a method for parameterizing the type of flexible framework model used in Chapter 2 from ab initio quantum chemistry calculations. The goal of this work is to facilitate the rapid development of flexible, versatile MOF models that can capture changes in the coordination bonding of metals in MOFs. This type of model can be used to study structural transitions and may be extended to study MOF formation. We demonstrated that our approach yields models for the M-MOF-74 series that are stable and have simulated structural properties in good agreement with quantum chemistry calculations.

Molecular recognition between biomolecules and ligands is very specific in living cells. The functions of all biochemical processes and cell mechanisms are dependent upon complex but specific non-covalent intermolecular interactions. As essential building blocks in protein and nucleic acid, phosphate groups are commonly found in nucleic acids, proteins, and lipids. Nearly half of known proteins have been shown to interact with ligands containing a phosphate group. Binding of a phosphoryl group is fundamental to a range of biological processes including metabolism, biosynthesis, gene regulation, signal transduction, muscle contraction, and antibiotic resistance. Phosphorylation is one of the most common forms of reversible posttranslational modification of protein and, nearly 30% of all proteins are phosphorylated on at least one residue in cells. However, phosphate binding sites are less well defined and fundamental principles of why and how proteins recognize phosphate groups are not yet fully understood. Molecular modeling is a common tool for studying biomolecular structure, dynamics, interaction and function. Due to the complex electrostatics, high concentration of ions and intricate interactions with environment, however, the modeling and designing of highly charged drug-like molecules and nucleic acid derivatives are extremely difficult. This thesis will focus on the highly charged phosphate, including its different protonation states, and energetic and thermodynamic driving forces behind protein-phosphate recognition. This thesis work will also discuss the development of more sophisticated computational models, AMOEBA+, that are necessary for a better understanding and prediction of the physical properties of small organic molecules. Four projects will be discussed in this dissertation: two projects on force field development, and two on applying molecular dynamic simulations to understand biological processes. These projects have led to new insights into understanding of physical and chemical principles and mechanisms underlying highly protein-phosphate binding and nucleic acid stability. In addition, this thesis work will enhance the capability to develop and apply computational and theoretical frameworks to model, predict and design proteins, therapeutics, and diagnostic strategies targeting phosphates, phosphate-containing biomolecules

Enzymes in the acetate and non-ribosomal peptide pathways generate chemically diverse and complex bioactive molecules, with the intermediates being chauffeured between catalytic partners via a carrier protein. Recent efforts have been made to engineer these systems to expand their product diversity. A major stumbling block is our poor understanding of the transient protein-protein and protein-substrate interactions between the carrier protein and its many catalytic partner domains. The innate reactivity of pathway intermediates has obfuscated our mechanistic understanding of these interactions during the biosynthesis of these natural products, ultimately impeding the engineering of these systems for the generation of “unnatural” natural products.Molecular dynamics can be used to provide models of these key interactions that are difficult to capture experimentally, providing the potential to expand the diversity in these systems. Current force fields support basic biochemical building blocks, and specialized force fields support post-translational modifications and non-canonical amino acids, yet none currently exist that are capable of modeling bound intermediates from these systems.The objective of this dissertation is to address this gap in knowledge and available technology and to present the description of the development of a force field that can be used with MD and other computational techniques to provide models of these experimentally intractable transient interactions. This objective will be divided into three aims.Aim 1: The development and validation of a force field for use in investigations and engineering efforts involving these pathways. A fragmentation approach was used, providing a degree of modularity to the force field while at the same time reducing the size, complexity and degrees of freedom during the charge fitting and parameterization steps. Tutorials and a web interface were also developed to provide end users access to this tool.Aim 2: The application of the force field towards understanding dynamics and interactions in these pathways. Regulation of the transient protein-protein interactions and discrete steps in fatty acid biosynthesis remain poorly understood. MD was able to show that specific interactions with its partner are either strengthened or weakened depending on the loaded state of the carrier protein. The force field was also used to model a plant-based polyketide synthase at different pHs, resulting in the identification of an allosterically modulated pH sensor on the surface of the polyketide synthase. Aim 3: The application of the force field in engineering efforts in these systems to produce biofuels. MD simulations were employed to provide a deeper understanding of protein-substrate interactions in a reductase domain from a non-ribosomal peptide megasynthase. This led to a deeper understanding of this domain, and further identified residues critical for structure integrity and substrate binding, leading to a rationally altered variant with improved activity toward highly reduced substrates.

The first section of this work details a force field modeled on VSEPR theory. Previous studies¹ from Bartell et al. have validated the use of the following function to describe repulsion between outer atoms, X [subscript i], bonded to a shared center, A, in binary compounds of the form AX [subscript n]: V=K/r [superscript n above subscript ij]. Here, K and n are parameters and r [subscript ij] is the distance between repelled atoms. Bartell et al. fixed the bond distances A-X [subscript i] so that the atoms X [subscript i] are “points-on-a-sphere” around the central atom A. Our current work extends this POS force field to include flexibility in the bond distances A-X [subscript i]. The functional form for the energy of the bonds is that of the Morse oscillator: V [subscript m, sub-subscript i] = D [subscript e](1 -- exp[ -- α(r [subscript i] -- Re)])². Here, D [subscript e], α, and R [subscript e] are parameters of the force field and r [subscript i] is the distance between atoms A and X [subscript i]. This extended VSEPR force field is applied to PF₅. Parameters were optimized to minimize differences between the VSEPR and MP2/cc-pVTZ PF₅ diagonal quadratic force constants. Quadratic and cubic bending and stretching force constants are presented and compared between Morse-POS and MP2/cc-pVTZ methods. The second section of this dissertation focuses on the analytical transformation of force constants and molecular properties between isotopologues. Within the Born-Oppenheimer approximation, potential energy surfaces of isotopologues are identical. When beginning an exploration of a Born-Oppenheimer potential energy surface, a molecular geometry is chosen and the energy and derivatives are calculated. The most efficient choice of coordinates for these derivatives is normal coordinates, which are mass dependent. This thesis details the transformation of force constants, dipole moments, derivatives of dipole moments, rotational constants, and Coriolis constants from one isotopologue to another. Two alternative systems of coordinates for use in quantum chemical calculations are rectilinear and curvilinear internal coordinates. Rectilinear internal coordinates express the displacement of atoms in a molecule as changes in bond length, bond angle, and dihedral angle as linear combinations of Cartesian coordinates. Though other internal coordinates exist, the previously mentioned are the most commonly used. There are immediate problems with the general use of rectilinear internal coordinates. When one calculates the displacement in Cartesian coordinates needed to increase a bond angle, the bond angle does not scale linearly with the Cartesian coordinates. The following thesis provides the derivation of equations that allow for the analytical transformation from rectilinear to curvilinear coordinates of first through fourth derivatives of the energy. As shown here, these transformations may also be used in converting between the force fields of isotopologues. Additionally, the transformations used between force fields of isotopologues are generalized to the transformation of derivatives of the dipole moment vector. Because the dipole moment vector is rotationally variant, the transformation is extended to include rotational parameters that can successfully transform derivatives of vector quantities between isotopologues. Third, a benchmark comparison of atomic natural orbital (ANO) and Dunning's correlation consistent basis sets is presented²⁻⁴. ANO basis sets, made up of atomic natural orbitals of the CISD wavefunction, are used far less frequently than the correlation consistent basis sets of Dunning and coworkers. However, ANO basis sets, especially smaller truncations (like ANO0) are powerful tools in calculations that require a high-accuracy description of correlation energy as is necessary in second order vibrational perturbation theory (VPT2) calculations. A benchmark study comparing the ANOX (X=0,1,2) and cc-pVNZ (N=D,T,Q) families of basis sets is detailed here. Using VPT2, fundamental frequencies of a set of small molecules were calculated with each of these six basis sets and compared. Of the comparisons, ANO0 outperforms cc-pVDZ in this work, validating more development and interest in ANO-type basis sets. This thesis also details the construction and benchmarking of new ANO-type basis sets using atomic natural orbitals from coupled cluster wavefunctions with single and double excitations (CCSD) and single, double, and triple excitations (CCSDT). These are used to calculate the fundamental frequencies of a set of small molecules and are compared to the original ANO basis sets, which are constructed from the configuration interaction wavefunction with single and double excitations (CISD). The three ANO-type basis sets are comparable in error, though small systematic trends arise that validate the use of the original ANO-type basis sets over those constructed with CCSD or CCSDT wavefunctions. The final section discusses experimental and theoretical collaboration to characterize the rotational and vibrational spectra of isomers of dihydroxycarbene⁵⁻⁶. The first publication discussed here⁵ details measured rotational spectra of isotopomers of cis,trans-dihydroxycarbene. Using ab initio parameters and a fitting procedure, a high-accuracy semi-experimental equilibrium structure of cis,trans-dihydroxycarbene was determined. The semi-experimental equilibrium structure agrees with high-level ab initio calculations to two significant figures in the bond lengths and three significant figure in bond angles. This section also details a recent publication⁶ that uses ab initio calculations to aid in the characterization of vibrational spectra of cis,transand trans,trans-dihydroxycarbene in ⁴He nanodroplets. VPT2 fundamental frequencies are compared to experimental values, all matching to ~7 cm⁻¹. State-specific dipole moments and components of the dipole moment vector along inertial axes determined from ab initio calculations are also compared with experimental results.

Atomistic modeling and simulation methods enable a modern molecular approach to bio-medical research. Issues addressed range from structure-function relationships to structure-based drug design. The ability of these methods to address biologically relevant problems is largely determined by their accurate treatment of electrostatic interactions in the target biomolecular structure. In practical molecular simulations, the electrostatic charge density of molecules is approximated by an arrangement of fractional "point charges" throughout the molecule. While chemically intuitive and straightforward in technical implementation, models based exclusively on atom-centered charge placement, a major workhorse of the biomolecular simulations, do not necessarily provide a sufficiently detailed description of the molecular electrostatic potentials for small systems, and can become prohibitively expensive for large systems with thousands to millions of atoms. In this work, we propose a rigorous and generally applicable approach, Optimal Point Charge Approximation (OPCA), for approximating electrostatic charge distributions of biomolecules with a small number of point charges to best represent the underlying electrostatic potential, regardless of the distance to the charge distribution. OPCA places a given number of point charges so that the lowest order multipole moments of the reference charge distribution are optimally reproduced. We provide a general framework for calculating OPCAs to any order, and introduce closed-form analytical expressions for the 1-charge, 2-charge and 3-charge OPCA. We demonstrate the advantage of OPCA by applying it to a wide range of biomolecules of varied sizes. We use the concept of OPCA to develop a different, novel approach of constructing accurate and simple point charge water models. The proposed approach permits a virtually exhaustive search for optimal model parameters in the sub-space most relevant to electrostatic properties of the water molecule in liquid phase. A novel rigid 4-point Optimal Point Charge (OPC) water model constructed based on the new approach is substantially more accurate than commonly used models in terms of bulk water properties, and delivers critical accuracy improvement in practical atomistic simulations, such as RNA simulations, protein folding, protein-ligand binding and small molecule hydration. We also apply our new approach to construct a 3-point version of the Optimal Point Charge water model, referred to as OPC3. OPCA can be employed to represent large charge distributions with only a few point charges. We use this capability of OPCA to develop a multi-scale, yet fully atomistic, generalized Born approach (GB-HCPO) that can deliver up to 2 orders of magnitude speedup compared to the reference MD simulation. As a practical demonstration, we exploit the new multi-scale approach to gain insight into the structure of million-atom 30-nm chromatin fiber. Our results suggest important structural details consistent with experiment: the linker DNA fills the core region and the H3 histone tails interact with the linker DNA. OPC, OPC3 and GB-HCPO are implemented in AMBER molecular dynamics software package.

A technology for optimization of potential parameters from condensed phase simulations (POP) is discussed and illustrated. It is based on direct calculations of the derivatives of macroscopic observables with respect to the potential parameters. The derivatives are used in a local minimization scheme, comparing simulated and experimental data. In particular, we show that the Newton Trust-Region protocol allows for accurate and robust optimization. POP is illustrated for a toy problem of alanine dipeptide and is applied to folding of the peptide WAAAH. The helix fraction is highly sensitive to the potential parameters while the slope of the melting curve is not. The sensitivity variations make it difficult to satisfy both observations simultaneously. We conjecture that there is no set of parameters that reproduces experimental melting curves of short peptides that are modeled with the usual functional form of a force field. We then apply the newly developed technology to study the liquid mixture of tert-butanol and water. We are able to obtain, after 4 iterations, the correct phase behavior and accurately predict the value of the Kirkwood Buff (KB) integrals. We further illustrate that a potential that is determined solely by KB information, or the pair correlation function, is not necessarily unique.

Construction of accurate potential energy (PE) surfaces for molecular systems is one of the primary tasks performed by theoretical physical chemists. Once in hand, these PE functions can be used to study the dynamics and spectroscopies, as well as the structures and properties of molecular systems. This study focuses on approximating many-body electronic inductionin order to improve the accuracy of existing potentials and improve the efficiency of {em{ab initio}}methods in order to allow “on-the-fly” energy and force evaluations in dynamical calculations. The majority of the work reported here focuses on the solvated electron. We initiate a study aimed at understanding the effects of explicitly including the (ultrafast) electron--solvent electronic induction, or polarization. We construct a single electron potential in which the coarse grained electronic degrees of freedom of the solvent are treated self-consistently along with the electronic wave function. Predictions of the binding energy of an excess electron in water clusters obtained using this potential compare well to {em{ab initio}} electronic structure theories. Subsequently, this potential was used to investigate the behavior of the excess electron in liquid water. The explicit treatment of induction appears to have a minimal impact on the structure and solvation dynamics of the excess electron (in the ground state) but does have a large impact on the vertical detachment energy and the optical absorption spectrum. In these latter cases there is an abrupt change in the charge distribution of the excess electron. In such cases the electronic response from the solvent can be large and should be taken into account. The electronic response of the solvent occurs on the time scale of electronic excitation.This introduces technical complications when solving for orthogonal eigenstates of this system since the model Hamiltonianis state dependent. We describe a simple method of solving this problem and discuss the possibility of generalizing this scheme to many-electron theories(such and density functional and Hartree-Fock theories). This procedure may potentially enable the study of non-adiabatic excited state relaxation dynamics including the electronic response of the solvent. Construction of empirical potential energy surfaces, such as the one developed here, is a time intensive process and one questions whether or not they have found the optimal set of parameters. We would prefer to use accurate electronic structure theories to compute energies and forces. Of particular interest is the use of {em{ab initio}} methods which offer a systematically improvable route to the exact energy. Currently this is only feasible for small systems and short time scales. A class of algorithms called fragment methods are currently being developed to extend these approaches to condensed phase environments. Our strategy has been to efficiently approximate electronic induction and fold this into the description of the single molecular fragments. The fragments are then coupled to one another through a version of symmetry adapted perturbation theory. This yields an accurate and efficient method that scales linearly in the large system limit.

In this work, existing force field descriptions of poly (lactic acid), or PLA, were improved by modifying the torsional potential energy terms to more accurately model the bond rotational behavior of PLA. Extensive calculations were carried out using density functional theory (DFT), for small PLA molecules in vacuo, and also using DFT with a continuum model to approximate the electronic structure of PLA in its condensed phase. From these results, improved force field parameters were developed using a combination of the OPLS and CHARMM force fields. The new force field, PLAFF2, is an update to the previously developed PLAFF model developed in David Bruce's group, and results in more realistic conformational distributions during simulation of bulk amorphous PLA. It is demonstrated that the PLAFF2 model retains the accuracy of the original PLAFF in simulating the crystalline α polymorph of PLA. The PLAFF2 model has superior performance to any other publicly available force field for use with PLA; hence, we recommend its use in future modeling studies on the material, whether in its crystalline or amorphous form.

This work describes the development of the polarizable force field for proteins based on classical mechanics, electronic structure theory and the combined quantum mechanical molecular mechanical method. In the first model, the classical force field is augmented with the explicit polarization energy term yielding a polarizable intermolecular potential function (PIPF). The polarizable atom sites in the system respond to the electric field by generating induced dipole moment at each site. The PIPF potential is optimized for amides and alkanes which are building blocks of proteins. The molecular dynamics simulations using the PIPF potential yield comparable or better energetics and geometries of the model compounds. In order to speed up the convergence of the induced dipole moments in the PIPF potential when all intramolecular interactions are included, we proposed a coupled polarization matrix inversion and iteration method (CPII). We were able to achieve convergence within 15 iterations for all the systems under consideration in which the iterative method shows divergence or oscillation. The second model, called the explicit polarization model (X-Pol), accounts for the polarization and charge transfer effects by treating all the fragments of the system using electronic structure theory. A variational version of the X-Pol potential is derived which facilitates the calculation of analytical gradient of energy needed for molecular dynamics simulations. Furthermore, the X-Pol potential is augmented with the buffer zone by calculating the two-electron Coulomb integrals between adjacent fragments in protein. The introduction of buffer zone in the X-Pol potential improves the convergence and the transferability of semiempirical parameters in the X-Pol potential. The molecular dynamics simulation of a solvated bovine pancreatic trypsine inhibitor reveals significant polarization and charge transfer effects in the protein.

The potential energy surfaces of both calcium binding sites in calbindin D9k were determined in DFT calculations using the Gaussian031 program. The potential energy surfaces were also determined using the CHARMM2 program; the default force field and a charge modified force field were used in the classical calculations. The potential energy surface was found for the interaction energy between the Ca2+ ion and the ligand oxygens plus limited neighbors. The binding sites are in the form of a pentagonal bipyramid around the calcium where 6 of the ligands are contributed by the protein and one is an oxygen atom from a water molecule. A large and significant difference is seen between the classical and quantum results; this dictates that any classical calculations involving the calcium ion in this and other EF-hand structures need to take this into account. A potential has been calculated using a Metropolis Monte Carlo simulated annealing procedure. Future work will implement this potential into a classical molecular dynamics program.

Public strategy is a widely discussed concept, which especially in an election year is often proposed as a means to improving administrative operations. Over the last 40 years, there has been an explosion in the number of books written about strategy. The academic community and the public have been bombarded with countless, theories, models, frameworks, approaches and schools about strategy. We have seen the rise and fall of strategic planning and the movement from strategic management to strategic leadership to strategic thinking. Despite the plethora of solutions, some to problems yet to be discovered, both the private and public sectors continue to see strategies fail. How can it be that strategies still fail when we have so many people available who know how to make them succeed? This study uses an information technology project in the Department of Defense as a field study to determine why the strategies for that project succeeded or why they failed. The research draws on concepts from the Delphi Technique, Force Field Theory, classic strategy literature and public administration to identify and map various levels of forces extant in the implementation environment for those strategies.