Cancer formation is a complex, multi-step process that allows cells to grow abnormally and potentially invade and spread throughout the body. A single genetic or structural alteration of a single protein in a cellular physiological process is enough to stimulate cancer formation. In treating cancer, a ‘targeted therapy’ approach is becoming increasingly common, where we can develop drugs that specifically target these altered proteins implicated in cancer. Two proteins that are heavily involved in several human cancers are protein kinases and p53, which are the focus of this dissertation work. I chose to use molecular dynamics simulations and in silico virtual screening, two methods from the computational chemistry toolbox in studying protein kinases and p53. I demonstrate that performing molecular dynamics is worthwhile in conducting virtual screens against protein kinases, because it may result in that at least one conformation is more predictive than the crystal structure. I also reveal key insight into the transcriptional activation mechanism of p53, and show how this mechanism is altered as a result of the R175H cancer mutation.

Rational drug design has been revolutionized by computational methods such as molecular dynamics (MD) simulations and virtual screening. From both the methods development and the application standpoints, the entry of computation into chemistry has greatly changed how biophysics and medicinal chemistry are approached. First, we have developed an interactive platform on which a broader array of scientists can efficiently run molecular dynamics simulations, presenting an opportunity for a more rigorous and more accessible approach to utilizing this method. Second, virtual screening has allowed us to foray into the design of a protein-protein inhibitor for a polymeric form of IκB kinase alpha. And third, the application of molecular dynamics simulations has aided us in understanding the dynamics of proteins. Through MD, our studies on NF-κB-inducing kinase have given us insight into the interplay of interactions among its catalytic domains and revealed a possible allosteric site.

Molecular Dynamic (MD) simulation is a powerful computational tool that can be applied to study biological systems at an atomic scale. Antibody 2D1 was isolated from the 1918 influenza virus surface glycoprotein hemagglutinin (HA) and was also known to cross-neutralize the 2009 pandemic influenza HA. Nevertheless, the detailed mechanism is unclear. We have conducted molecular dynamic (MD) simulations to study the interactions

between Ig-2D1 and the HAs from four different strains including its natural binder 1918HA, the 2009 HA, a seasonal 2006 strain and a 2009HA mutant. We found that in 09HA, a serine to asparagine mutation from the 18HA weakened one of the salt bridges, which leaded to the loss of hydrogen bonds and the formation of a water pocket between 09HA and Ig-2D1. Another system involves the cancer suppressor, full-length p53 protein, and its DNA counter-parts. In this system, we observed that the C-terminals contacted DNA and formed direct salt bridges. This observation supported previous research, which reported that the C-terminals interact with DNA nonspecifically to search for the binding sequence. Each of these observations was possible because MD simulations provide atomistic detail, which facilitates the study of protein-protein and protein-DNA interactions. Additionally, MD simulations can furnish refined results, but the simulation and analysis processes can be daunting. To ease the complications, we utilized the Kepler platform and developed several automated workflows that integrated multiple commands into one central process

Aerosols are solid or liquid particles suspended in the air that can have far-reaching impacts on climate and human health. Aerosols impact climate through their radiative properties and their ability to seed cloud droplets or ice crystals. They also provide surfaces at which heterogeneous multiphase reactions can occur and serve as sinks for atmospheric sulfur, carbon and nitrogen. From a human health perspective, the physical and chemical properties of aerosols including their size, shape, and composition, can impact their transfer and deposition into the lungs. Smaller particles in particular can contain pollutants and pathogens and are able to travel deeper into the bronchioles to trigger irritation and infection. This body of work applies molecular dynamics simulations to understand aerosol systems, investigating their morphologies, impacts on climate, and ultimately their role in transporting the airborne SARS-CoV-2 virus. Molecular simulation and analysis methods are integrated with experiment to first probe surfactant interfaces with varying levels of chemical complexity, then to explore whole aerosol dynamics and phase within the context of understanding impacts of sea spray aerosols (SSA) on climate. This work shows that 1) surfactant charge modulates the surface activity of Burkholderia cepacia lipase at lipid monolayer interfaces; 2) calcium enhances polysaccharide adsorption to fatty acid monolayers; and 3) divalent cations induce morphological changes in LPS-containing aerosols, hindering the reactive uptake of atmospheric nitric acid. This dissertation also describes methods for building large-scale, intact SSA models with full chemical complexity and shows how organic components distribute throughout the aerosol, suggesting that SSA may adopt microemulsion-like morphologies. Finally, a workflow is developed to build ultra-large systems for the study of airborne disease, demonstrating the successful construction and simulation of 1) the SARS-CoV-2 wild type envelope, and 2) a billion-atom respiratory aerosol containing the full breadth of chemical complexity, including the first all-atom model of the Delta SARS-CoV-2 envelope and never-before-modeled pulmonary mucins. The latter project presents the first atomic-level views of the SARS-CoV-2 virus within a respiratory aerosol and represents a novel approach to investigating the infection mechanisms of airborne pathogens.

The development of molecular dynamics (MD) simulations builds off the maturing field of structural biology to provide new insight into the mechanisms of disease on an atomic level. However, there are few established methods that use these insights to develop effective therapies. This thesis begins by discussing the creation of “POVME 3.0”, a novel method to generate drug design-relevant insights from MD simulations. POVME 3.0 takes as input a MD simulation of a binding pocket of interest, and returns a summary of how the pocket shape changes over time. We then discuss the application of POVME 3.0 and other analysis techniques to the APOBEC3 family of proteins. APOBEC3 proteins are a newly discovered driver of mutation in many cancers, and their inhibition could contribute to cancer treatments. Next, we review algorithms that could be productively used in the study of APOBEC3 enzymes, both to understand their essential dynamics and also to discover new modes of inhibition. Finally, we discuss CELPP, a community-driven analysis of computer-aided drug design algorithms, which aims to improve the quality of predictive models in drug design.

The passing of time is a constant. In the same way that time shapes our everyday lives, it also plays an essential role in the microscopic universe through its translation into macromolecular dynamics. Proteins, in particular, visit a variety of short-lived - or metastable - states, and the conformations and rate of transitions between these states dictate their function in the cell. However, many of these alternative states are not accessible to established biophysical techniques, such that many puzzling results and challenges arise when trying to use these static observations to understand and predict their properties. In this dissertation, we emphasize the importance of accessing and characterizing the intrinsic dynamics of these macromolecules through the application of computational techniques in four distinct projects that allow us to observe molecular motion at the atomic level to widen our understanding of their function, and furthermore enable us to make predictions on properties that would be inaccessible from static models. In conjunction with experimental validation, we study the role of electrostatic interactions in mediating the allosteric activation of Protein Kinase A, an ubiquitous enzyme involved in key signal transduction pathways, as well as develop a methodology that incorporates dynamic descriptors to improve the efficiency of the challenging process of novel protein design. Additionally, through the application of the exciting Markov State Model methodology to access longer timescales than typically available to computational simulations, we help set the stage for the interpretation of a novice experimental technique, diffuse scattering, in terms of atomic motions in crystalline environments, and explore the conformational landscape of the essential tumor suppressor p53 and its cancer-related mutant Y220C, which leads to the identification of a novel cryptic pocket. These test cases present significant advances to our understanding of protein dynamics in fields as varied as protein design and cancer therapeutics, and taken together, stress the necessity of putting time, and dynamics, in the center stage of protein study.

Computational tools provide the automation and power that enable detailed modeling and analysis of many biomolecular phenomena of interest. Open source programs and automated tools empower researchers and provide opportunities for improvement to existing software. In the past few years, I have developed several open-source scientific software packages for the purposes of automating difficult or menial tasks pertaining to computational biophysics. These software packages involve the analysis of molecular dynamics simulations, Brownian dynamics simulations, electrostatics, pocket volume measurement, solvent fragment mapping, binding site characterization, milestoning theory, and allosteric network communications. In addition to allowing my research group and me to approach biomedical challenges that would otherwise be intractable, I hope and intend that these tools will be useful to the computational and theoretical biophysics research community.

Proteins dance; the atoms of a protein move through time in our cellular aqueous environment, and each atom is important in this process. The dance of a protein is unique to that protein, and alone controls how that protein functions. In this way, our cells are a world unto their own, and as the proteins of a cell dance along, they orchestrate life. In this thesis, I discuss the dance, or dynamics, of five different proteins: the ankyrin repeat and SOCS box protein 9 (ASB9), creatine kinase (CK), the L99A mutant of T4 lysozyme, the Inhibitor of κB kinase 2 (IKK2), and the NFκB essential modulator (NEMO). Each of these proteins have unique sizes, shapes, dynamics, and roles in the cell. To study these proteins, I chose to use the computational microscope, performing molecular dynamics simulations at the atomic level. Through the solving of theoretical models of the bonds between atoms, iterated over and over again through time, a dance unfolded for each of these proteins, illuminating how these proteins perform their duties in our cells and in our test tubes. I demonstrate that ASB9 and CK interact through a highly disordered interface, uncover the dynamics which guide both excited state transitioning and benzene release in the L99A mutant of T4 lysozyme, and present a hypothesis for how NEMO appends itself to the IKK2 hexamer.

Background

Results

Conclusions

Advancements in computational and methodological approaches have revolutionized our understanding of complex biomolecular systems. This dissertation presents innovative computational methodologies for accurately predicting the kinetics and thermodynamics of complex biomolecular systems, aiming to provide efficient alternatives to traditional long-scale molecular dynamics (MD) simulations. The research focuses on multiscale milestoning approaches, weighted ensemble methods, deep learning algorithms, and obtaining precise force field parameters as efficient alternatives to traditional MD simulations. This work presents advancements accomplished in the SEEKR (Simulation enabled estimation of kinetic rates) framework, an advanced tool for estimating biomolecular kinetics and thermodynamics, particularly predicting ligand-receptor (un)binding estimates. The incorporation of quantum mechanical force field reparameterization through the QMrebind tool marks a significant step in enhancing the accuracy of milestoning simulations. Additionally, it explores hybrid methods such as GaMD-WE (Gaussian-accelerated MD with the weighted ensemble) and DeepWEST (Deep learning with the weighted ensemble simulation toolkit), which synergize the benefits of enhanced sampling methods and deep learning for more effective kinetic and thermodynamic sampling of biomolecules. These methodologies significantly reduce computational demands while accurately predicting complex biomolecular processes, including receptor-ligand binding/unbinding, protein-protein interactions, and (un)folding, offering substantial contributions to computational biology and enhancing the understanding of biomolecular behavior with greater efficiency and accuracy.