%{search_type} search results

Deep Learning hat bereits in vielen Bereichen bemerkenswerte Ergebnisse erzielt. Jetzt hält es Einzug in die Wissenschaften, insbesondere in die Biowissenschaften. Dieses praxisorientierte Buch bietet Programmierern und Wissenschaftlern einen Überblick darüber, wie Deep Learning in Genomik, Chemie, Biophysik, Mikroskopie, medizinischer Analyse und Arzneimittelforschung eingesetzt wird. Das Buch vermittelt Deep-Learning-Grundlagen und führt in die Arbeit mit der Python-Bibliothek DeepChem ein. Sie erfahren, wie Deep Learning z.B. zur Analyse von mikroskopischen Bildern, für molekulare Daten und bei medizinischen Scans genutzt wird. Abschließend zeigen Bharath Ramsundar und seine Co-Autoren anhand einer Fallstudie Techniken für die Entwicklung neuer Therapeutika, eine der größten interdisziplinären Herausforderungen der Wissenschaft.

The development of small molecule drugs is a lengthy and expensive process that could potentially be improved by new technologies. Lead optimization is an important part of small molecule drug discovery, where initial hit molecules are gradually developed into suitable drug candidates. It can be described as an iterative cycle of design, make and test phases, which can be further broken down into a series of concrete sub-problems, namely: molecular property prediction, molecule generation, chemical synthesis planning, experimental chemical synthesis, and experimental testing. This thesis explores machine learning methods to tackle a few of the sub-problems in small molecule lead optimization, with a focus on the early design phases

Deep learning has already achieved remarkable results in many fields. Now it's making waves throughout the sciences broadly and the life sciences in particular. This practical book teaches developers and scientists how to use deep learning for genomics, chemistry, biophysics, microscopy, medical analysis, and other fields. Ideal for practicing developers and scientists ready to apply their skills to scientific applications such as biology, genetics, and drug discovery, this book introduces several deep network primitives. You'll follow a case study on the problem of designing new therapeutics that ties together physics, chemistry, biology, and medicine-an example that represents one of science's greatest challenges. Learn the basics of performing machine learning on molecular data Understand why deep learning is a powerful tool for genetics and genomics Apply deep learning to understand biophysical systems Get a brief introduction to machine learning with DeepChem Use deep learning to analyze microscopic images Analyze medical scans using deep learning techniques Learn about variational autoencoders and generative adversarial networks Interpret what your model is doing and how it's working.
(source: Nielsen Book Data)

Atomistic simulations provide detailed information about a molecular system's dynamics at finer time and length scales than experiments can access. However, the modeling and interpretation of simulation datasets in an unbiased and statistically sound way require dedicated algorithms. One analysis method is the so-called variational approach to conformational dynamics, which introduces an objective framework for the ranking of models. One such class of models are Markov state models (MSMs), which separate the configuration space explored by a simulated molecule, such as a protein, into discrete, disjoint states between which the transitions can be modeled as Markovian. In Chapter 1, I summarize the essentials of MSM construction and the variational approach. In Chapters 2 and 3, I describe systematic studies in varying MSM parameters, in pursuit of general trends for variationally optimal models of proteins, with a focus on clustering into states. In Chapters 4 and 5, I present algorithmic advances in comparing multiple related models and in coarse-graining MSMs, also using clustering. In Chapter 6, I summarize the current state of the art in MSM methods and identify frontiers in their application and methods development. Finally, in Chapters 7 and 8, I apply clustering to different problems, such as in fluid dynamics. I discuss the possibility of extending the methods motivated here to a broader class of dynamical systems in Chapter 9. Overall, the methods presented in this work focus on interpretability and statistical robustness.

• Cover; Copyright; Table of Contents; Preface; Conventions Used in This Book; Using Code Examples; O'Reilly Online Learning; How to Contact Us; Acknowledgments;
• Chapter 1. Why Life Science?; Why Deep Learning?; Contemporary Life Science Is About Data; What Will You Learn?;
• Chapter 2. Introduction to Deep Learning; Linear Models; Multilayer Perceptrons; Training Models; Validation; Regularization; Hyperparameter Optimization; Other Types of Models; Convolutional Neural Networks; Recurrent Neural Networks; Further Reading;
• Chapter 3. Machine Learning with DeepChem; DeepChem Datasets

• Training a Model to Predict Toxicity of MoleculesCase Study: Training an MNIST Model; The MNIST Digit Recognition Dataset; A Convolutional Architecture for MNIST; Conclusion;
• Chapter 4. Machine Learning for Molecules; What Is a Molecule?; What Are Molecular Bonds?; Molecular Graphs; Molecular Conformations; Chirality of Molecules; Featurizing a Molecule; SMILES Strings and RDKit; Extended-Connectivity Fingerprints; Molecular Descriptors; Graph Convolutions; Training a Model to Predict Solubility; MoleculeNet; SMARTS Strings; Conclusion;
• Chapter 5. Biophysical Machine Learning

• Protein StructuresProtein Sequences; A Short Primer on Protein Binding; Biophysical Featurizations; Grid Featurization; Atomic Featurization; The PDBBind Case Study; PDBBind Dataset; Featurizing the PDBBind Dataset; Conclusion;
• Chapter 6. Deep Learning for Genomics; DNA, RNA, and Proteins; And Now for the Real World; Transcription Factor Binding; A Convolutional Model for TF Binding; Chromatin Accessibility; RNA Interference; Conclusion;
• Chapter 7. Machine Learning for Microscopy; A Brief Introduction to Microscopy; Modern Optical Microscopy; The Diffraction Limit

• Electron and Atomic Force MicroscopySuper-Resolution Microscopy; Deep Learning and the Diffraction Limit?; Preparing Biological Samples for Microscopy; Staining; Sample Fixation; Sectioning Samples; Fluorescence Microscopy; Sample Preparation Artifacts; Deep Learning Applications; Cell Counting; Cell Segmentation; Computational Assays; Conclusion;
• Chapter 8. Deep Learning for Medicine; Computer-Aided Diagnostics; Probabilistic Diagnoses with Bayesian Networks; Electronic Health Record Data; The Dangers of Large Patient EHR Databases?; Deep Radiology; X-Ray Scans and CT Scans; Histology

• MRI ScansLearning Models as Therapeutics; Diabetic Retinopathy; Conclusion; Ethical Considerations; Job Losses; Summary;
• Chapter 9. Generative Models; Variational Autoencoders; Generative Adversarial Networks; Applications of Generative Models in the Life Sciences; Generating New Ideas for Lead Compounds; Protein Design; A Tool for Scientific Discovery; The Future of Generative Modeling; Working with Generative Models; Analyzing the Generative Model's Output; Conclusion;
• Chapter 10. Interpretation of Deep Models; Explaining Predictions; Optimizing Inputs; Predicting Uncertainty

Deep learning has already achieved remarkable results in many fields. Now it's making waves throughout the sciences broadly and the life sciences in particular. This practical book teaches developers and scientists how to use deep learning for genomics, chemistry, biophysics, microscopy, medical analysis, and other fields. Ideal for practicing developers and scientists ready to apply their skills to scientific applications such as biology, genetics, and drug discovery, this book introduces several deep network primitives. You'll follow a case study on the problem of designing new therapeutics that ties together physics, chemistry, biology, and medicine-an example that represents one of science's greatest challenges. Learn the basics of performing machine learning on molecular data Understand why deep learning is a powerful tool for genetics and genomics Apply deep learning to understand biophysical systems Get a brief introduction to machine learning with DeepChem Use deep learning to analyze microscopic images Analyze medical scans using deep learning techniques Learn about variational autoencoders and generative adversarial networks Interpret what your model is doing and how it's working.
(source: Nielsen Book Data)

This dissertation covers work discussed in the following papers: "Spatial Graph Convolutions for Drug Discovery" describes new deep neural network architectures for modeling drug-receptor interactions. We argue that the future of predicting the interactions between a drug and its prospective target demands more than simply applying deep learning algorithms from other domains, like vision and natural language, to molecules. "Machine Learning Harnesses Molecular Dynamics to Discover New Opioid Chemotypes" describes an algorithm that leverages protein motion to enrich the search for active molecules. We then applied the method to find a new chemical scaffold that we experimentally verified is an agonist for the μ Opioid Receptor. "Kinetic Machine Learning Unravels Ligand-Directed Conformational Change of Opioid Receptor" describes differential pathways of deactivation and differential conformational states sampled by the μ Opioid Receptor in response to different opioid ligands.

Understanding the structure and dynamics of biological macromolecules is a central focus of biological research. To be able to study and gain insights into these systems, it is first necessary to have an accurate and informative model for the system of interest. However, such a model is often difficult to build. For example, during protein folding, many proteins collapse into transient kinetic intermediates on timescales too fast for high-resolution experimental techniques to detect, preventing structural characterization of these species. Alternatively, current algorithms for RNA design (i.e. predicting a sequence that folds into a desired target structure) cannot accurately model structure-sequence relationships and rely primarily on brute force stochastic search, leading to poor performance on complex targets. Here, we show that it is possible to improve the quality of models for biological systems by applying a common Bayesian approach to building them, i.e. incorporating prior information to impose informative constraints on the model parameters. Through this approach, it is possible to build high-resolution models of protein dynamics given limited experimental data, as well as a state-of-the-art computational RNA design agent that outperforms all currently existing algorithms.

DNA gyrase is an essential bacterial molecular motor that uses ATP hydrolysis to drive the directional introduction of DNA supercoils. The enzyme employs a duplex strand passage mechanism that requires coordinating the opening and closing of three protein "gates": the N-gate, DNA-gate, and Exit-gate. The N-gate is formed by the dimerization of ATPase domains and acts as a nucleotide-dependent clamp that captures DNA for subsequent strand passage. Dynamic measurements of N-gate conformational changes are necessary to understand how gyrase harnesses chemical energy to direct changes in DNA topology. Here, we report real-time single molecule measurements of E. coli gyrase N-gate conformational dynamics under varying DNA and nucleotide conditions. We identify a landscape of distinct conformational intermediates whose populations can be shifted upon DNA and nucleotide binding. The N-gate is primarily open in the absence of DNA and nucleotide, but transiently samples closed conformations. The non-hydrolyzable ATP analog AMPPNP, but not ADP, induces stable N-gate dimerization, where FRET values are consistent with a closed conformation seen in crystal structures based on in silico modeling of dye positions. In the presence of DNA, the enzyme samples a distinct high FRET conformation of the N-gate that is consistent with an intermediate conformation previously described in studies of B. subtilis gyrase. Our measurements support a loose-coupling model in which N-gate conformations are highly dynamic and depend on both DNA and nucleotide binding. Substrate-induced N-gate conformational changes appear to be conserved across divergent bacterial species and could extend to other enzymes in the Gyrase-Hsp90-MutL (GHL) ATPase family. This work sets the stage for detailed structural modeling and for multimodal measurements that directly correlate protein and DNA dynamics in this complex molecular machine.

The advent of atomistic molecular dynamics simulations held the promise of a complete understanding of biomolecular dynamics. However, this goal has remained elusive, as increased computational power has brought with it larger systems to simulate and an overwhelming number of observables to analyze. In this work, I describe how recent advancements in Markov state modeling have helped overcome this dimensionality problem and enabled the characterization of complex phenomena, such as the folding-upon-binding processes of intrinsically disordered peptides. But is it possible to produce even more insightful models? To this end, I present a method that exploits Markov state models to infer statistically causal drivers of protein dynamics. Finally, I discuss a neural network alternative to Markov models, which yields physically interpretable insights and has the potential to replace expensive atomistic simulations.

Proteins are frequently characterized as molecular machines, with atomic-level motions driving biological function. Past decades have witnessed a dramatic increase in the tools available to probe these dynamics, but few methods enable us to resolve these collective motions with high spatial resolution. This dissertation investigates the potential of x-ray diffuse scattering from protein crystals to meet this critical need. Specifically, I review the models of correlated disorder that have previously been suggested to account for this signal and describe algorithms for processing the diffuse scattering in experimental diffraction data. These models and algorithms are applied to dissect the physical origins of the diffuse scattering observed from three protein crystals. Though considerable progress is still required for the analysis of diffuse scattering to become a routine biophysical method for studying protein dynamics, the framework and findings described in this dissertation make concrete steps toward that end.

Molecular dynamics (MD) simulations are a computational technique capable of providing detailed atomic level understanding of molecular processes. They do this by numerically solving the microscopic interactions that govern these processes. Akin to a "computational microscope", MD simulations are used to predict and understand these protein free-energy landscapes. Given simulations from related mutant proteins, MD simulations could even predict the atomic level effects of mutations. These mutations might be oncogenic thereby increasing cell proliferation or they might abrogate an inhibitors' binding affinity or they might be completely benign. A structural and quantitative model for these mutations would be invaluable in both understanding the system and potentially designing personal therapeutics. However, MD simulations are very computationally expensive to converge and it is not immediately obvious how to compare multiple mutant simulations to one another in a statistically significant and efficient manner. In this thesis, I describe methods for analyzing and sampling mutant free energy landscapes. The first few chapters are dedicated towards building Markov state models (MSMs) for multiple protein kinases using milliseconds of aggregate simulation data. Currently, these are some of the largest kinase simulation datasets ever recorded. The latter chapters present several methodological advances on how to more efficiently predict mutational effects using a combination of enhanced sampling simulations and existing MSMs. This thesis attempts to merge the fields of Machine learning, enhanced sampling, and Markov state modeling for the efficient analysis and sampling of protein mutation landscapes.

Organic photovoltaics (OPVs) have emerged as a promising alternative to conventional PV technology due to their low cost and industry-level scalability with high-volume production through solution-based processing. OPVs combine the unique flexibility and versatility of plastics with electronic properties, making them amenable to applications in the "Internet of Things" and distributed generation applications. The current key challenge for wide adaptation of OPVs is the lack of high power conversion efficiency (PCE) in large scale roll-to-roll processed devices. A key factor is the morphology: there exists disorder between the electron donor and electron acceptor materials in the active layer, and the mechanisms by which the morphology can be tuned are not well understood. Simulation is a promising inexpensive technique for exploring OPVs in the large parameter space of both processing methods and chemical components. In this work, we leverage and improve upon these computational approaches to reduce the need for iterative design for OPVs. First, we develop a multiscale molecular dynamics (MD) model to provide understanding of morphology evolution during solution processing. In addition, we train and utilize a predictive deep learning model to study the correlation of performance with the chemical and engineering design considerations. These parallel approaches allow for an accelerated sampling of the parameter space of OPV conditions, which in turn leads to targeted experiments.

Chemical information systems answer questions about chemistry to support decision-making across the spectrum of investment, research, operations, and governance in industries as diverse as agriculture, energy, and healthcare. Chemoinformatics systems facilitate question answering by providing three core functions: indexing, search, and inference. In chapter 1, we introduce the basic concepts, lexicon, use cases, and unmet needs in chemoinformatics. Chemical structure is a key index for molecules because it is unique physical description that determines properties and biological activities, offering powerful search and quantitative structure-activity relationship prediction (QSAR) capabilities. In chapter 2 we, give provide review of representations and algorithms in chemoinformatics. The computable descriptions of molecules that support search and QSAR are a key area of innovation in chemoinformatics, and the recent success of deep and shallow learning architectures in other fields has sparked interest in representation learning. In chapter 3, we investigate the transfer of representations learned by kernel methods employed by support vector machines to linear regression models and characterize the tradeoffs in classification performance and computation time. Analog analysis focuses on chemical structure transformations and relationships and requires specialized databases and techniques for indexing and search, which are limited by their inability to flexibly abstract over the molecular contexts of analogous transformations. In chapter 4 we develop a conceptual algebra of molecules that supports search for chemical analog relationships with flexible abstraction over their molecular contexts. In chapter 5, we apply our method of representing chemical transformations using algebraic expressions of molecules to characterize the space of drug metabolism reactions catalyzed by microbes in the human gut. Finally, in chapter 6 we conclude our work by summarizing our contributions and outlining future directions for our research.

Models are tools used to interpret and draw conclusions from nature. Molecular dynamics (MD) simulation is a powerful technique for modeling complex atomistic systems such as biomolecules. In this dissertation, I discuss how one can improve and apply MD simulation in order to learn about biophysical phenomena. I first discuss how to improve the representation of the underlying physical interactions in a simulation. Chapter 2 discusses the optimization method, and 3 discusses how to rigorously characterize a resultant potential function. I then discuss how to use Markov state modeling to derive an interpretable mechanistic characterization of a simulation dataset. Chapters 4 and 5 apply this framework to study the conformational dynamics of the TREK-2 and CLC-2 ion channels, respectively. A brief introduction to the topics of MD simulation, force field optimization, and Markov state modeling is given in chapter 1. There remains a lot of work to be done before simulations are able to mimic reality with high fidelity. However, I am optimistic that with increasing data availability and improvements in optimization methodology, simulation will prove itself progressively more useful for studying dynamics at atomic resolution.

Biology is the ultimate emergent phenomenon, and we largely lack a full picture of its function at the smallest scales. Molecular dynamics purports to model biomolecules like proteins with all-atom resolution. Among other challenges, merely analyzing the large quantities of data that result from a simulation has become a bottleneck. In this dissertation, I present my work towards building reduced-complexity models that faithfully capture the relevant functional dynamics of biomolecular simulations. In chapter 1, I introduce a mathematical language for dealing with stochastic processes and show the connection to established modeling methods like Markov modeling and tICA. Chapter 2 develops and characterizes a method for including solvent degrees of freedom in Markov state models. In chapter 3, we apply state-of-the-art MSM modeling to understand multi-scale conformational dynamics of a potassium ion channel. Chapter 4 provides an overview of a curated selection of modeling building blocks accessible through our carefully designed software package. Chapter 5 introduces a new non-linear basis which unites the MSM and tICA approaches. Finally, in chapter 6, I introduce parameterized sets of basis functions and use the variational principle directly to optimize the basis set itself. It is my hope that these novel algorithms aided by well-engineered software implementations and validated by characterization on real biomolecular systems will lead the field closer towards truly robust dynamical models of biomolecules.

Bryostatin 1 is a marine natural product that has been of great interest to chemists and clinicians due to its highly complex structure and its remarkable biological activity. Bryostatin has been investigated for the treatment of many indications, most notably cancer, HIV, and Alzheimer's disease, collectively for which it has been entered into over 40 clinical trials. Notwithstanding the immense potential impact of bryostatin's biological activity, its current supply is nearly exhausted and future supply is uncertain. All bryostatin that has been used clinically has been sourced from one GMP isolation in 1991, and all subsequent efforts to isolate more (through isolation from source organism, aquaculture, engineered biosynthesis, or total synthesis) have been unsuccessful or not scalable. One solution to bryostatin's supply problem is the design of a shorter, supply-impacting synthesis. If accomplished, this solution would allow for a rapid replenishment of bryostatin's supply and an immediate clinical impact. A second solution to bryostatin's supply problem is the design of new bryostatin analogs. Because bryostatin was not optimized for its therapeutic use in humans, new analogs can be designed that retain or even improve upon bryostatin's biological activity while also reducing its immense complexity. Such an effort, however, is complicated by the fact that there exists little structural information about bryostatin's target, protein kinase C (PKC), in its active, membrane-associated state. Thus, a substantial portion of the work described here has used both molecular dynamics (MD) simulations and solid state NMR experiments to more fully understand the structure and function of membrane-associated PKC. Chapter 1 provides a survey of the structure, function, and membrane interactions of PKC. This chapter contains a brief overview of the different PKC isoforms, their various functions within the cell, and the biological indications that are tied to PKC regulation (such as cancer, HIV, and Alzheimer's disease). It examines the bryostatin analogs that have been synthesized in order to target these indications. Of particular emphasis is that design of new PKC activators has been complicated by the fact that while a few X-ray and NMR structures of PKC fragments exist, there are no structures of membrane-associated PKC. The importance of the membrane in PKC function is described, as are the efforts thus far to examine the role of the membrane in the activity of PKC activators. Chapter 2 details the use of molecular dynamics (MD) simulations in elucidating the membrane-associated structure of ligand-bound PKC. These simulations examine how different PKC activators differentially position the ligand-bound PKC complex in the membrane, and the role of waters and lipid headgroups at the interface of the membrane and cytosol. These simulations also provide an explanation for why bryostatin's northern region is important to its activity despite not being in contact with the binding pocket, thus providing a hypothesis for future design of new bryostatin analogs. Chapter 3 details the synthesis of a new library of greatly simplified bryostatin analogs, and the development and use of a new assay to test the PKC binding affinity of these and other compounds across all conventional and novel PKC isoforms. These greatly simplified compounds remove bryostatin's complex northern region entirely by replacing the A- and B-rings with a short diester chain, thus reducing a 20-membered macrocycle to a 14-membered one, and are synthesized in only 19 linear steps (20 total). It is also shown that while some compounds in this library bind to PKC as strongly as bryostatin 1 across all isoforms, others exhibit unprecedented selectivity between conventional and some novel PKCs. Chapter 4 addresses the lack of any existing experimental membrane-associated structure of the PKC-ligand complex. This problem is addressed through the use of solid-state REDOR NMR studies, in which interatomic distances are measured between different isotopes. These experiments used an isotopically-labeled bryostatin analog bound to the PKCδ C1b domain in the presence of phospholipid vesicles. In doing so, this represents the first experimental determination of the bound conformation of any PKC activator in a phospholipid membrane. These experiments are coupled with MD simulations to use the measured interatomic distances to construct a full picture of the ensemble of conformations that exist in this PKC-ligand-membrane complex. Chapter 5 details the total synthesis of bryostatin 1. Through the collaborative work of 8 co-workers in the Wender lab, we have accomplished the shortest reported synthesis of bryostatin 1 at 19 linear steps (29 total). My key contributions to this collaborative effort are highlighted. This short synthesis is scalable and has thus far produced more than 2 grams of bryostatin 1. This chapter also describes how such a synthesis fundamentally alters the landscape of bryostatin supply; all bryostatin that had ever been used in the clinic was from one GMP isolation in 1991 and is almost entirely exhausted. Subsequent efforts to isolate bryostatin and replenish this supply have proven either unsuccessful or not scalable. Our accomplishment in producing a short, scalable synthesis breaks through this barrier and finally provides a new, renewable source of bryostatin 1.

The study of microbe morphology uniquely offers a collision of the simple and hyper complex, both in the model organisms like E. coli, and the tools used to dissect them. At first glance, a microbe needs to do nothing more than incorporate nutrients and replicate itself. Yet observations dating back nearly six decades demonstrate that bacteria exhibit precise regulation of growth and form robust across multi-fold changes in mass. From these simple observations, the model that a bacterium is nothing more than a bag of enzymes becomes insufficient to explain the rich and robust landscape of morphology they can achieve. In this work, we seek to expand and expound the molecular mechanisms by which bacteria regulate their shape, using a set of diverse experimental and computational techniques. In the first chapter, we offer a colloquial introduction to bacterial morphology and the major molecular components known to be involved. The following three chapters include their own, more technical introductions. In the second chapter, we present a molecular model of MreB, the bacterial cytoskeletal protein thought to be responsible for spatially organizing the growth of the cell shape-determining cell wall. This model, based on analysis of molecular dynamic simulations, was published in 2014 and subsequently confirmed by an independent lab via crystallography. Our molecular models provide a reference point from which to interpret a growing diversity of disparate experimental results. In the third chapter, we explore how direct molecular interaction between MreB and other cell wall synthesis proteins -- namely RodZ -- can drive changes in the biophysical properties of the bacterial cytoskeleton. This chapter relies on simple but crucial set of observational experiments to reveal that the properties of the bacterial cytoskeleton vary widely in the course of normal bacterial growth. We identify potential protein interactions that appear to directly bind to and alter the biophysical properties of the MreB cytoskeleton. This work is currently in review at Nature Microbiology. In the final chapter, we expand the scope of focus to broadly identify possible interactions between all essential genes in the model bacteria Bacillus subtilis using a variety of experimental, genetic and bioinformatics techniques. In particular, we offer a novel insight into how the depletion of nearly 300 unique proteins drives a diverse set of changes in microbial morphology, highlighting a large number of molecular players previously not associated with changes in growth. Taken together, these chapters represent three unique scopes -- molecular, subcellular and cellular, that work in concert to answer questions about bacterial morphology and its regulation. The results presented herein, and the novel experimental assays developed in the process, present tangible advances in our understanding of bacterial morphology, and expands the list of known unknowns left to pursue.

We are in the midst of a machine learning revolution. From self-driving cars to clinical diagnostics, machine learning promises to change the way we live our lives and make decisions. Applied to drug discovery, machine learning enables us to build upon existing experimental data and more effectively explore the vastness of chemical space for new therapeutics. In particular, virtual screening allows us to evaluate many more compounds and biological targets than we can test experimentally, helping to identify starting points for further development. In this dissertation, I present several applications of machine learning to drug discovery. Much of the work presented here focuses on multitask neural networks---variants of the models that have transformed computer vision and beaten some of the world's best Go players. Applications of these models to ligand-based virtual screening demonstrate improvements over standard machine learning methods such as random forest and logistic regression. I also describe neural network models built on simple encodings of the molecular graph, moving beyond traditional fingerprint-based screening methods to a richer and more flexible input representation.

Chaperonins are an essential family of molecular chaperones present in all three domains of life. They are large oligomeric complexes comprising two rings of 7-9 subunits related by twofold symmetry. Chaperonins engage unfolded or misfolded client proteins and encapsulate them in an isolated folding chamber in an ATP-dependent manner. The chaperonins are subdivided into two phyla, known simply as group I and group II. The group I chaperonins are found in bacteria and the endosymbiotic organelles mitochondria and chloroplasts, whereas group II chaperonins are principally encoded by archaeal and eukaryotic genomes. Less frequently, group II chaperonins may be found in bacterial genomes. The two groups of chaperonin share the same domain architecture, differing primarily in complex stoichiometry, inter-ring register, and the requirement of cofactors for client folding. This thesis concerns the structural dynamics of the terminal regions of both the group I and group II chaperonins. In particular, this work presents evidence from molecular dynamics simulations that the C-termini of the group I chaperonin undergo an ATP dependent conformational cycle with implications for substrate binding. Using a combination of crystallography and solution state nuclear magnetic resonance spectroscopy, the final chapter of the thesis establishes that the C-terminal conformational cycle hypothesized for the group I chaperonin is observed in a model group II chaperonin. Furthermore, that chapter introduces a heretofore unobserved N-terminal conformation of the group II chaperonin and demonstrates that both the N and C-termini influence the nucleotide usage of the enzyme.

Here we report that proper treatment of nonbonded interactions is essential for the accuracy of molecular dynamics (MD) simulations, especially in studies of lipid bilayers. The use of the CHARMM36 force field (C36 FF) in different MD simulation programs can result in disagreements with published simulations performed with CHARMM due to differences in the protocols used to treat the long-range and 1-4 nonbonded interactions. In this study, we systematically test the use of the C36 lipid FF in NAMD, GROMACS, AMBER, OpenMM, and CHARMM/OpenMM. A wide range of Lennard-Jones (LJ) cutoff schemes and integrator algorithms were tested to find the optimal simulation protocol to best match bilayer properties of six lipids with varying acyl chain saturation and head groups. MD simulations of a 1,2-dipalmitoyl-sn-phosphatidylcholine (DPPC) bilayer were used to obtain the optimal protocol for each program. MD simulations with all programs were found to reasonably match the DPPC bilayer properties (surface area per lipid, chain order parameters, and area compressibility modulus) obtained using the standard protocol used in CHARMM as well as from experiments. The optimal simulation protocol was then applied to the other five lipid simulations and resulted in excellent agreement between results from most simulation programs as well as with experimental data. AMBER compared least favorably with the expected membrane properties, which appears to be due to its use of the hard-truncation in the LJ potential versus a force-based switching function used to smooth the LJ potential as it approaches the cutoff distance. The optimal simulation protocol for each program has been implemented in CHARMM-GUI. This protocol is expected to be applicable to the remainder of the additive C36 FF including the proteins, nucleic acids, carbohydrates, and small molecules.