The number of three-dimensional structures of potential protein targets available in several platforms such as the Protein Data Bank is subjected to a constant increase over the last decades. This observation should be an additional motivation to use structure-based methodologies in drug discovery. In the recent years, different success stories of Structure Based Drug Design approach have been reported. However, it has also been shown that a lack of druggability is one of the major causes of failure in the development of a new compound. The concept of druggability can be used to describe proteins with the capability to bind drug-like compounds. A general consensus suggests that around 10% of the human genome codes for molecular targets that can be considered as druggable. Over the years, the protein druggability was studied with a particular interest to capture structural descriptors in order to develop computational methodologies for druggability assessment. Different computational methods have been published to detect and evaluate potential binding sites at protein surfaces. The majority of methods currently available are designed to assess druggability of a static structure. However it is well known that sometimes a few local rearrangements around the binding site can profoundly influence the affinity of a small molecule to its target. The use of techniques such as molecular dynamics (MD) or Metadynamics could be an interesting way to simulate those variations. The goal of this thesis was to design a new computational approach, called JEDI, for druggability assessment using a combination of empirical descriptors that can be collected ‘on-the-fly’ during MD simulations. JEDI is a grid-based approach able to perform the druggability assessment of a binding site in only a few seconds making it one of the fastest methodologies in the field. Agreement between computed and experimental druggability estimates is comparable to literature alternatives. In addition, the estimator is less sensitive than existing methodologies to small structural rearrangements and gives consistent druggability predictions for similar structures of the same protein. Since the JEDI function is continuous and differentiable, the druggability potential can be used as collective variable to rapidly detect cryptic druggable binding sites in proteins with a variety of MD free energy methods.

Bioactive glasses are a class of synthetic inorganic material that have wide orthopedics, dentistry, tissue engineering and other biomedical applications. The origin of the bioactivity is closely related to the atomic structures of these novel glass materials, which otherwise lack long range order and defies any direct experimental measurements due to their amorphous nature. The structure of bioactive glasses is thus essential for the understanding of bioactive behaviors and eventually rational design of glass compositions. In this dissertation, molecular dynamics (MD) and reverse monte carlo (RMC) based computer simulations have been used to systematically study the atomic structure of three classes of new bioactive glasses: strontium doped 45S5 Bioglass®, ZnO-SrO containing bioactive glasses, and Cao-MgO-P2O5-SiO2 bioactive glasses. Properties such as ionic diffusion that are important to glass dissolution behaviors are also examined as a function of glass compositions. The accuracy of structure model generated by simulation was validated by comparing with various experimental measurements including X-ray/neutron diffraction, NMR and Raman spectroscopy. It is shown in this dissertation that atomistic computer simulations, when integrated with structural and property characterizations, is an effective tool in understanding the structural origin of bioactivity and other properties of amorphous bioactive materials that can lead to design of novel materials for biomedical applications.

Currently the treatment of alcohol use disorder is very difficult and often requires the combination of therapy and medications, with many who undertake treatment experiencing relapse over time. There is also no treatment in use to prevent the development of alcohol use disorder. It is the aim of this work to provide information that may be useful for the development of a preventative treatment for developing alcohol use disorder by elucidating more of the acute effects of alcohol use. It is known that these effects originate in the brain. Within the brain are circuits made up of neurons that communicate with each other through chemical synapses. These chemical synapses involve the release of neurotransmitters from one neuron that are detected by another neuron, which initiates its own response. It is known that ethanol can change how much neurotransmitter is released from a neuron, depending on the specific neuron tested, and many researchers have implicated the "release machinery" as a target. It is also known that alcohol can affect lipid membrane properties that are important for the fusion of the vesicle membrane, encapsulating the neurotransmitter, with the cell membrane for release of the neurotransmitter outside of the neuron. It is not known if alcohol directly affects the SNARE proteins ("release machinery") or the lipid membranes to initiate the change in neurotransmitter release previously observed. Within this work you will find a discussion of the steps of neurotransmitter release and the known effects of anesthetics on components of this process, as an introduction to the topic (Chapters 1 and 2). In Chapters 3-5 you will find studies that successively dive deeper and deeper into the effects of alcohol on the SNARE proteins and lipid membranes. We show that ethanol is effective at a dose of 0.4% v/v or 64 mM at increasing fusion probability in a model of neurotransmitter release that uses the 3 SNARE proteins to drive fusion of a vesicle with a supported membrane. We also show that alcohol has little direct effect on the SNARE proteins themselves. In addition, we provide evidence that alcohol alters fusion oppositely, depending on which membrane leaflet it has most direct access to. In Chapter 5 we show that alcohol increases the probability of lipid tail protrusion in silico. Previously it has been shown that protrusion of one fatty acid tail of one lipid can initiate fusion of that membrane with an apposing membrane. These data provide further insight into the effects of alcohol on a neuron and we would argue are valuable to research pursuing treatment and prevention of alcohol use disorder.

No description available.

Preparing molecular coordinate files for molecular dynamics (MD) simulations can be a very time-consuming process. Herein we present the development of a user-friendly program that drastically reduces the time required to prepare these molecular coordinate files for MD software packages such as AmberTools. Our program, known as charge atomtype naming (CAN), creates and uses a library of structures such as amino acid monomers to update the charge, atom type, and name of atoms in any molecular structure (mol2) file. We demonstrate the utility of this new program by rapidly preparing structural files for MD simulations for polypeptides ranging from small molecules to large protein structures. Both native and non-native amino acid residues are easily handled by this new program. Proteins and enzymes generally achieve their function by creating well-defined 3D architectures that pre-organize reactive functionalities. Mimicking this approach to supramolecular preorganization is leading to the development of highly versatile artificial chemical environments, including new biomaterials, medicines, artificial enzymes, and enzyme-like catalysts. The use of beta-turn and alpha-helical motifs is one approach that enables the precise placement of reactive functional groups to enable selective substrate activation and reactivity/selectivity that approaches natural enzymes. Our recent work has demonstrated that helical peptides can serve as scaffolds for pre-organizing two reactive groups to achieve enzyme-like catalysis. In this study, we used CYANA and AmberTools to develop a computational approach for determining how the structure of our peptide catalysts can lead to enhancements in reactivity. These results support our hypothesis that the bifunctional nature of the peptide enables catalysis by pre-organizing the two catalysts in reactive conformations that accelerate catalysis by proximity. We also present evidence that the low reactivity of monofunctional peptides can be attributed to interactions between the peptide-bound catalyst and the helical backbone, which are not observed in the bifunctional peptide.

Doctor of Philosophy / Department of Chemistry / Paul Edward Smith / The activity and function of many macromolecules in cellular environments are coupled with the binding of ions such as alkaline earth metal ions and poly oxo anions. These ions are involved in the regulation of important processes such as protein crystallization, nucleic acid and protein stability, enzyme activity, and many others. The exact mechanism of ion specificity is still elusive. In principle, computer simulations can be used to help provide a molecular level understanding of the dynamics of hydrated ions and their interactions with the biomolecules. However, most of the force fields available today often fail to accurately reproduce the properties of ions in aqueous environments. Here we develop a classical non polarizable force field for aqueous alkaline earth metal halides (MX₂) where M = Mg²⁺, Ca²⁺, Sr²⁺, Ba²⁺ and X = Cl⁻, Br⁻, I⁻, and for some biologically important oxo anions which are NO₃⁻, ClO₄⁻, H₂PO₄⁻ and SO₄²⁻, for use in biomolecular simulations. The new force field parameters are developed to reproduce the experimental Kirkwood-Buff integrals. The Kirkwood-Buff integrals can be used to quantify the affinity between molecular species in solution. This helps to capture the fine balance between the interactions of ions and water. Since this new force field can reproduce the experimental Kirkwood-Buff integrals for most concentrations of the respective salts, they are capable of reproduce the experimental activity derivatives, partial molar volumes, and excess coordination numbers. Use of these new models in MD simulations also leads to reasonable diffusion constants and dielectric decrements. Attempts to develop force field parameters for CO₃²⁻, HPO₄²⁻ and PO₄³⁻ ions were unsuccessful due to an excessive aggregation behavior in the simulations. Therefore, in an effort to overcome this aggregation behavior in the simulations, we have investigated scaling the anion to water interaction strength, and also the possibility of using a high frequency permittivity in the simulations. The strategy of increasing relative permittivity of the system to mimic electronic screening effects are particularly promising for decreasing the excessive ion clustering observed in the MD simulations.

The non-reactive Dreiding and the reactive ReaxFF atomic potentials were applied within a family of atom molecular dynamics (MD) simulations to investigate and understand interfacial adhesion in graphene/vinyl ester composites. First, a liquid vinyl ester (VE) resin was equilibrated in the presence of graphene surfaces and then cured, resulting in a gradient in the monomer distribution as a function of distance from the surfaces. Then the chemically realistic relative reactivity volume (RRV) curing algorithm was applied that mimics the known radical addition regiochemistry and monomer reactivity ratios of the VE monomers during three-dimensional chain-growth polymerization. Surface adhesion between the cured VE resin and the graphene reinforcement surfaces was obtained at a series of VE resin “crosslink densities.” Both pristine and oxidized graphite sheets were employed separately in these simulations using a Dreiding potential. The pristine sheets serve as a surrogate for pure carbon fibers while oxidizing the outer graphene sheets serve as a model for oxidized carbon fibers. Hence, the effects of local monomer distribution and temperature on the interphase region formation and surface adhesion can be investigated. Surface adhesion was studied at various curing conversions and as a function of temperature. Uniaxial loading simulations were performed at different curing conversions for both models to predict the composites’ modulus of elasticity, Poisson’s ratio, and yield strength. The same analysis was performed for the neat cured matrix. The glass transition temperature (Tg) for the homogenized composite and neat VE matrix was determined at different degrees of curing. Subsequent MD simulations were performed to predict structural damage evolution and fracture in the neat VE matrix. The ReaxFF potential was used to quantify irreversible damage due to bond breakage in the neat VE matrix for different degrees of cure, stress states, temperatures, and strain rates. The predicted damage mechanisms in the bulk VE thermosetting polymer were directly compared to those for an amorphous polyethylene (PE) thermoplastic polymer.

Improvement in the in vitro oxidative folding of disulfide-containing proteins, such as extracellular and pharmaceutically important proteins, is required. Traditional folding methods using small molecule aliphatic thiol and disulfide, such as glutathione (GSH) and glutathione disulfide (GSSG) are slow and low yielding. Small molecule aromatic thiols and disulfides show great potentiality because aromatic thiols have low pKa values, close to the thiol pKa of protein disulfide isomerase (PDI), higher nucleophilicity and good leaving group ability. Our studies showed that thiols with a positively charged group, quaternary ammonium salts (QAS), are better than thiols with negatively charged groups such as phosphonic acid and sulfonic acid for the folding of bovine pancreatic trypsin inhibitor (BPTI). An enhanced folding rate of BPTI was observed when the protein was folded with a redox buffer composed of a QAS thiol and its corresponding disulfide. Quaternary ammonium salt (QAS) thiols and their corresponding disulfides with longer alkyl side chains were synthesized. These QAS thiols and their corresponding disulfides are promising small molecule thiols and disulfides to fold reduced BPTI efficiently because these thiols are more hydrophobic and can enter the core of the protein. Conformational changes of disulfide-containing proteins during oxidative folding influence the folding pathway greatly. We performed the folding of BPTI using targeted molecular dynamics (TMD) simulation and investigated conformational changes along with the folding pathway. Applying a bias force to all atoms versus to only alpha carbons and the sulfur of cysteines showed different folding pathways. The formation of kinetic traps N' and N* was not observed during our simulation applying a bias force to all atoms of the starting structure. The final native conformation was obtained once the correct antiparallel β-sheets and subsequent Cys14-Cys38 distance were decreased to a bond distance level. When bias force was applied to only alpha carbons and the sulfur of cysteines, the distance between Cys14-Cys38 increased and decreased multiple times, a structure similar to the confirmation of N*, NSH were formed and native protein was ultimately obtained. We concluded that there could be multiple pathways of conformational folding which influence oxidative folding.

Defects in the functional oxides play an important role in electronic devices like metal oxide semiconductor field effect transistors (MOSFETs) and resistive random-access memories (ReRAMs). The continuous scaling of CMOS has brought the Si MOSFET to its physical technology limit and the replacement of Si channel with Ge channel is required. However, the performance of Ge MOSFETs suffers from Ge/oxide interface quality and reliability problems, which originates from the charge traps and defect states in the oxide or at the Ge/oxide interface. The sub-oxide layers composed of GeII states at the Ge/GeO2 interface seems unavoidable with normal passivation methods like hydrogen treatment, which has poor electrical properties and is related to the reliability problem. On the other hand, ReRAM works by formation and rupture of O vacancy conducting filaments, while how this process happens in atomic scale remains unclear. In this thesis, density functional theory is applied to investigate the defect behaviours in oxides to address existing issues in these electronic devices. In chapter 3, the amorphous atomic structure of doped GeO2 and Ge/GeO2 interface networks are investigated to explain the improved MOSFET reliability observed in experiments. The reliability improvement has been attributed to the passivation of valence alternation pair (VAP) type O deficiency defects by doped rare earth metals. In chapter 4, the oxidation mechanism of GeO2 is investigated by transition state simulation of the intrinsic defect diffusion in the network. It is proposed that GeO2 is oxidized from the Ge substrate through lattice O interstitial diffusion, which is different from SiO2 which is oxidized by O2 molecule diffusion. This new mechanism fully explains the strange isotope tracer experimental results in the literature. In chapter 5, the Fermi level pinning effect is explored for metal semiconductor electrical contacts in Ge MOSFETs. It is found that germanides show much weaker Fermi level pinning than normal metal on top of Ge, which is well explained by the interfacial dangling bond states. These results are important to tune Schottky barrier heights (SBHs) for n-type contacts on Ge for use on Ge high mobility substrates in future CMOS devices. In chapter 6, we investigate the surface and subsurface O vacancy defects in three kinds of stable TiO2 surfaces. The low formation energy under O poor conditions and the +2 charge state being the most stable O vacancy are beneficial to the formation and rupture of conducting filament in ReRAM, which makes TiO2 a good candidate for ReRAM materials. In chapter 7, we investigate hydrogen behaviour in amorphous ZnO. It is found that hydrogen exists as hydrogen pairs trapped at oxygen vacancies and forms Zn-H bonds. This is different from that in c-ZnO, where H acts as shallow donors. The O vacancy/2H complex defect has got defect states in the lower gap region, which is proposed to be the origin of the negative bias light induced stress instability.

Machine Learning interatomic potentials (ML-IAP) are currently the most promising Non-empirical IAPs for molecular dynamic (MD) simulations. They use Machine Learning (ML) methods to fit the potential energy surface (PES) with large reference datasets of the atomic configurations and their corresponding properties. Promising near quantum mechanical accuracy while being orders of magnitudes faster than first principle methods, ML-IAPs are the new “hot topic” in material science research. Unfortunately, most of the available publications require advanced knowledge about ML methods and IAPs, making them hard to understand for beginners and outsiders. This work serves as a plain introduction, providing all the required knowledge about IAPs, ML, and ML-IAPs from the beginning and giving an overview of the most relevant approaches and concepts for building those potentials. Exemplary a gaussian approximation potential (GAP) for amorphous carbon is used to simulate the defect induced deformation of carbon nanotubes. Comparing the results with published density-functional tight-binding (DFTB) results and own Empirical IAP MD-simulations shows that publicly available ML-IAP can already be used for simulation, being indeed faster than and nearly as accurate as first-principle methods. For the future two main challenges appear: First, the availability of ML-IAPs needs to be improved so that they can be easily used in the established MD codes just as the Empirical IAPs. Second, an accurate characterization of the bonds represented in the reference dataset is needed to assure that a potential is suitable for a special application, otherwise making it a 'black-box' method.:1 Introduction 2 Molecular Dynamics 2.1 Introduction to Molecular Dynamics 2.2 Interatomic Potentials 2.2.1 Development of PES 3 Machine Learning Methods 3.1 Types of Machine Learning 3.2 Building Machine Learning Models 3.2.1 Preprocessing 3.2.2 Learning 3.2.3 Evaluation 3.2.4 Prediction 4 Machine Learning for Molecular Dynamics Simulation 4.1 Definition 4.2 Machine Learning Potentials 4.2.1 Neural Network Potentials 4.2.2 Gaussian Approximation Potential 4.2.3 Spectral Neighbor Analysis Potential 4.2.4 Moment Tensor Potentials 4.3 Comparison of Machine Learning Potentials 4.4 Machine Learning Concepts 4.4.1 On the fly 4.4.2 De novo Exploration 4.4.3 PES-Learn 5 Simulation of defect induced deformation of CNTs 5.1 Methodology 5.2 Results and Discussion 6 Conclusion and Outlook 6.1 Conclusion 6.2 Outlook