12023 research outputs found
Sort by
Computation-Aided Protein Engineering for Targeted Therapeutic Delivery
Get PDFMy Ph.D. projects centered on using computational structural biology tools to develop protein engineering methods for targeted therapeutic delivery, emphasizing delivering molecules to the brain. In this thesis, I focus on three main projects. First, utilizing computational structural biology techniques, I investigate the molecular mechanism that enables engineered adeno-associated viral (AAV) capsids to cross the blood-brain barrier (BBB). I develop a pipeline to model the vast and dynamic complex between engineered AAV capsids and their BBB receptors. I also apply a tool, recently developed by myself and discussed in Chapter 3, to distinguish capsids that bind to different receptors. The findings of this study can lead to novel approaches for developing chemicals and biologicals that can penetrate the human brain (Chapter 2). Second, I describe the development of Automated Pairwise Peptide-Receptor AnalysIs for Screening Engineered proteins (APPRAISE). This computational pipeline predicts the receptor binding propensity of engineered proteins based on competitive modeling and physics-grounded analysis. I show that APPRAISE is capable of distinguishing between receptor-dependent and receptor-independent adeno-associated viral vectors and ranking various engineered proteins, such as miniproteins binding to the SARS-CoV-2 spike and nanobodies binding to a G-protein-coupled receptor. A top performer in an in silico screening using APPRAISE was validated experimentally (Chapter 3). Third, I show an example to engineer a genetically encoded transmitter indicator (GETI), which may eventually be a cargo delivered to the brain. The GETI has a novel scaffold based on bacterial repressors, a class of transcriptional regulators that are critical for bacteria to respond to environmental chemicals. I repurposed an antibiotic-sensing repressor protein to bind a neurotransmitter, melatonin, using machine-learning-guided directed evolution. A melatonin indicator was then created by integrating the repurposed receptor with a fluorescent protein. This engineering platform may be adapted to create bio-orthogonal GETIs for various neurotransmitters (Chapter 4)
Grasp, Speech, and Internal Speech Representation in the Human Cortical Grasp Circuit
Get PDFThe ability to move freely and to connect with others through communication is invaluable for human independence. In this thesis, we explore how brain-machine interfaces (BMIs) can help patients affected by movement or speech deficits to recover lost human experiences. This work builds on previous findings indicating that premotor and posterior parietal areas are involved in movement generation and language processes. These higher-level brain areas do not only engage in movement execution, but also during planning, representing rich behavioral patterns that can be leveraged for BMI applications. In this work, we investigated how the ventral premotor cortex (PMv), the posterior parietal cortex (PPC), and the sensorimotor cortex (S1) represent grasp and speech processes at a single-neuron level. Using multielectrode Utah arrays, neuronal populations were recorded in tetraplegic human participants. We found that the supramarginal gyrus (SMG), PMv and S1 significantly encode motor imagery of grasping. By studying the cognitive processes underlying neural activity during the cue phase of grasping, we found a transition from cue-modality dependent to cue-modality independent grasp representation in SMG, the anterior intraparietal cortex (AIP) and PMv. Our findings suggest SMG integrates audio, written, and image cue modalities, but more similarly represents audio and written cues, indicating language involvement. We confirmed this hypothesis by demonstrating that SMG encodes spoken words, engaging different motor plans for speech compared to grasping even when the semantic content remained unchanged. These results suggest a BMI could be trained to decode both grasp motor imagery and speech from one brain area. Lastly, we showed that SMG is highly involved in language processes, modulating for written word recognition, auditory tones, vocalized speech, and internal speech. As a proof-of-concept, we built a real-time internal speech BMI from signals recorded in SMG that can decode eight words with high accuracy. This work is the first of its kind, demonstrating internal speech can be robustly decoded from an implant in a single brain area. We find high neural SMG generalization between seeing a written word, saying it internally and vocalizing the word, suggesting shared cognitive functions between different language processes. Furthermore, words in different languages are represented in SMG. This thesis advances the BMI field by providing a better understanding of the neural processes that underly grasp motor imagery and language. To summarize, our findings suggest that studying higher-level brain areas can lead to the development of more effective and versatile brain-machine interfaces
Ion Transport in Temperature Sensitive Polyelectrolytes
Get PDFTemperature sensors are widely employed and play a key role in many industries, such as automotive vehicles, medical devices, environmental monitoring, and process control. The state-of-the-art thermal sensing elements are made of rigid and costly inorganic materials, such as vanadium oxide and platinum. These materials have limitations for emerging applications such as wearable devices and prosthetic devices. Ideal temperature sensing materials for such applications need to be flexible, reliable under mechanical deformation, and suitable for large-area production. Electrical conductive polymers were found to be a promising solution because of their flexibility and solution processability. However, they often lag in temperature resolution compared to their inorganic counterparts.
A recent discovery revealed that the ionic conductivity of crosslinked pectin, a biopolymer extracted from plant cell walls, has a record-high temperature response. It is biocompatible, flexible when hydrated, and solution-processable, making it a strong candidate for wearable temperature sensing and conformal temperature mapping. However, open questions remain about the origin of its temperature sensitivity and the principles governing its ion transport. Furthermore, the heterogeneity of the complex molecular structure of pectin presents challenges to its integration in sensing devices.
In this thesis, we study the origin of the high thermal sensitivity in pectin and develop a synthetic polyelectrolyte that mimics its key structure and properties. In Chapter 3, we focus on the ion transport mechanism in crosslinked pectin. We show that the binding between multivalent ions and certain chemical functional groups of pectin plays a critical role in its temperature sensitivity. In Chapter 4, the impact of water content on the ion transport and dielectric processes in crosslinked pectin is also investigated. In the following chapter, we present a novel synthetic polyelectrolyte designed to mimic pectin with a simpler structure. It has superior flexibility, high temperature sensitivity, and is stable under mechanical deformation. To further study this new material, we examine its ion transport dynamics under varying humidity and temperature conditions in Chapter 7. We discover that temperature and humidity have a similar effect on ion transport. Overall, we showed a biomimetic approach to design temperature sensitive polymers where the strong ion-polymer binding is the key to the ultrahigh temperature response.</p
Engineering of Second-Generation Acoustic Reporter Genes
Get PDFA major outstanding challenge in the fields of biological research, synthetic biology, and cell-based medicine is visualizing the functions of natural and engineered cells noninvasively inside opaque organisms. Ultrasound imaging has the potential to address this challenge as a widely available technique with a tissue penetration of several centimeters and spatial resolution below 100 µm. Recently, the first genetically encoded acoustic reporters were developed based on bacterial gas vesicles (GVs) to link ultrasound signals to molecular and cellular function. However, the properties of these first-generation acoustic reporter genes (ARGs) resulted in limited sensitivity and specificity for imaging gene expression in vivo.
The goal of my thesis work has been to engineer second-generation ARGs with improved acoustic and expression phenotypes compared to the existing first-generation constructs. I took two complementary engineering approaches to developing these constructs: homolog screening and directed evolution, sometimes referred to as the “nature and nurture” of protein engineering. The resulting constructs offer major qualitative and quantitative improvements, including much stronger ultrasound contrast, the ability to produce nonlinear signals distinguishable from background tissue in vivo, stable long-term expression, and compatibility with in vitro multiplexed imaging. In collaboration with others in the lab, we demonstrate the capabilities of these next-generation ARGs by imaging in situ gene expression in mouse models of breast cancer and tumor-homing therapeutic bacteria, noninvasively revealing the unique spatial distributions of tumor growth and colonization by therapeutic cells in living subjects and providing real-time guidance for interventions such as needle biopsies.
This thesis is organized as follows: in the first two chapters, I introduce the key background needed to understand both the importance and properties of ARGS, and how they have been and could be engineered. In the next two chapters, I detail specific efforts to engineer them—one involving the construction of a high-throughput, semi-automated setup for acoustic phenotyping of cells and its application to ARG directed evolution, and another involving the screening of several GV cluster homologs to identify ones suitable for use as improved ARGs. Finally, I conclude with insights gleaned from these two ARG engineering projects and suggestions for future ones.
The approaches, results, and ideas presented in this thesis represent the current state-of-the-art in ARG engineering and application. While recent technology development in this field has unlocked exciting new use cases for ARGs in noninvasive biological imaging, most of their potential for basic science and disease diagnosis and treatment has yet to be realized.</p
Interparticle Forces and Stress Transfer in Saturated and Unsaturated Granular Systems
Get PDFGranular systems are ubiquitous in nature and engineering applications. The macroscopic behavior of such systems is governed by the behavior at the grain-scale, including force transfer between adjacent grains. The correlation between continuum behavior and interparticle forces in granular systems is yet to be fully understood. For a saturated or unsaturated granular system under external load, it is important to decode stress partition and transfer in the solid, fluid, and gas phases. In the meantime, the presence of the fluid phase and gas phase greatly increases the difficulty of measuring interparticle forces in opaque granular systems. This thesis describes the theoretical and experimental works on interparticle forces and effective stresses in two types of granular systems: i) fully saturated granular media, and ii) unsaturated granular media.
The first part of the thesis focuses on the direct measurement of interparticle forces and the experimental validation of the concept of effective stress introduced by Karl Terzaghi. The grain-scale expression of Terzaghi's effective stress for saturated granular media under small deformation and quasi-static state is derived using stress decomposition and balance of forces and moments. For the experimental validation of the analytical solution, an experimental setup was designed to study 2D saturated rubber rod packing under classic 1D consolidation. A hybrid optical-mechanical method based on the Granular element method (GEM) and Digital image correlation (DIC) is applied. The interparticle forces are directly computed from 2D strain distribution of the grains, and the effective stress is calculated using the grain-scale forces. With pore water pressure measured by a pressure sensor, the summation of the effective stress and the pore water pressure is then compared with the external load applied in the 1D consolidation experiment, which is the core of Terzaghi's principle. The 1D consolidation experiment is also compared with the 1D consolidation model and matches the results from Discrete element simulations (DEM).
The second part of the thesis investigates the measurement of interparticle forces in more complex unsaturated granular systems consisting of solid, pore fluid, and pore air phases. In the case of quasi-static, point contact, and low saturation, an expression for the partition of stress is derived as a function of interparticle forces. To simplify the expression of the stress partition equation, capillary bridges, which are integral parts of unsaturated systems under low saturation condition, are simulated numerically using 2D finite element method (FEM) to further understand the influence of gravity on pore fluid clusters. As the original GEM for fully saturated systems focuses on interparticle interactions, the GEM is further developed for unsaturated systems based on the original GEM and considering capillary forces. Finally, a hybrid optical-mechanical approach combined with the granular element method (GEM) is developed to extract interparticle forces in a classic 1D consolidation experiment. The partition of stresses is determined by experimental results and compared with the analytical results.
The major contributions of this thesis are the theoretical derivation and experimental validation of the link between the grain-scale properties (interparticle forces, branch vectors, etc.) and the stress transfer in fully saturated and unsaturated systems. The theoretical and experimental methodology employed in the thesis could pave the way for exploring the mechanics and physics behind the constitutive behaviors of a variety of poromechanical systems.</p
Learning and Control of Dynamical Systems
Get PDFDespite the remarkable success of machine learning in various domains in recent years, our understanding of its fundamental limitations remains incomplete. This knowledge gap poses a grand challenge when deploying machine learning methods in critical decision-making tasks, where incorrect decisions can have catastrophic consequences. To effectively utilize these learning-based methods in such contexts, it is crucial to explicitly characterize their performance. Over the years, significant research efforts have been dedicated to learning and control of dynamical systems where the underlying dynamics are unknown or only partially known a priori, and must be inferred from collected data. However, much of these classical results have focused on asymptotic guarantees, providing limited insights into the amount of data required to achieve desired control performance while satisfying operational constraints such as safety and stability, especially in the presence of statistical noise.
In this thesis, we study the statistical complexity of learning and control of unknown dynamical systems. By utilizing recent advances in statistical learning theory, high-dimensional statistics, and control theoretic tools, we aim to establish a fundamental understanding of the number of samples required to achieve desired (i) accuracy in learning the unknown dynamics, (ii) performance in the control of the underlying system, and (iii) satisfaction of the operational constraints such as safety and stability. We provide finite-sample guarantees for these objectives and propose efficient learning and control algorithms that achieve the desired performance at these statistical limits in various dynamical systems. Our investigation covers a broad range of dynamical systems, starting from fully observable linear dynamical systems to partially observable linear dynamical systems, and ultimately, nonlinear systems.
We deploy our learning and control algorithms in various adaptive control tasks in real-world control systems and demonstrate their strong empirical performance along with their learning, robustness, and stability guarantees. In particular, we implement one of our proposed methods, Fourier Adaptive Learning and Control (FALCON), on an experimental aerodynamic testbed under extreme turbulent flow dynamics in a wind tunnel. The results show that FALCON achieves state-of-the-art stabilization performance and consistently outperforms conventional and other learning-based methods by at least 37%, despite using 8 times less data. The superior performance of FALCON arises from its physically and theoretically accurate modeling of the underlying nonlinear turbulent dynamics, which yields rigorous finite-sample learning and performance guarantees. These findings underscore the importance of characterizing the statistical complexity of learning and control of unknown dynamical systems.</p
Essays in Mechanism Design and Contest Theory
Get PDFThis dissertation contains three essays. They offer contributions to the fields of mechanism design (Chapters 1 and 2) and contest theory (Chapter 3).
Chapter 1, co-authored with Wade Hann-Caruthers, studies the problem of aggregating privately-held preferences for a facility to be located on a plane. We show that for a large class of social cost functions, the mechanism that locates the facility at the coordinate-wise median of the agent’s ideal points is quantitatively optimal (in the sense that it has the smallest worst-case approximation ratio) among all deterministic, anonymous, and incentive-compatible mechanisms. We also obtain bounds on the worst-case approximation ratio of the coordinate-wise median mechanism for an important subclass of social cost functions.
Chapter 2, co-authored with Wade Hann-Caruthers, studies a principal-agent project selection problem with asymmetric information and demonstrates the value for the principal in inducing partial verifiability constraints, such as no-overselling, on the agent. We consider a setting where the principal has to choose one among a set of available projects but the relevant information, such as each project's profitability, is held by a self-interested agent who might also have its own preference over the projects. If the agent is unconstrained in its ability to manipulate its private information, the principal can do no better than randomly choosing a project. But if the agent cannot oversell any of the projects, maybe because it must support its claims with evidence, we show that a simple cutoff mechanism (agent's favorite project is chosen among those that meet a cutoff profit level and a default project) is optimal for the principal. We also find evidence in support of the well-known ally-principle which says that principal delegates more authority to an agent with more aligned preferences.
Chapter 3 studies the effect of increasing the value of prizes and competitiveness of contests on the effort exerted by participants in an incomplete information environment. We identify two natural sufficient conditions on the distribution of abilities in the population under which the interventions have opposite effects on effort. We also discuss applications to the design of optimal contests in three different environments, including the design of grading contests. Assuming that the value of a grade is determined by the information it reveals about the agent's ability, we establish a link between the informativeness of a grading scheme and the effort induced by it.</p
Folding and Dynamic Deployment of Ultralight Thin-Shell Space Structures
Get PDFThin-shell structures are becoming increasingly popular for space missions due to their high stiffness-to-mass ratio, easy folding and coiling, and self-deployment using stored strain energy. Broadly, two deployment strategies exist: 1) controlled or deterministic, and 2) unconstrained. Controlled deployment involves carefully orchestrated events using control or guidance systems, while in unconstrained deployment, the structure is simply allowed to self-deploy with minimal guidance. Unconstrained deployment offers lighter deployment mechanisms and better packaging efficiency but the unpredictability of this process has been a significant obstacle to its adoption.
This study focuses on demonstrating the predictability of unconstrained dynamic deployment of thin-shell structures, using the Caltech Space Solar Power Project (SSPP) structures as a case study. The Caltech SSPP uses composite triangular rollable and coilable longerons as the primary building blocks to create large bending-stiff structures. The specific objective is to improve the predictability and robustness of the unconstrained dynamic deployment of the Caltech SSPP structures. Deployment is influenced by the initial conditions and the interaction between the structure and the mechanism during the deployment. To understand these effects, high-fidelity numerical simulations are developed and validated against experiments. The study also examines the sensitivity of deployment characteristics to various design parameters and external influences to ensure the robustness of deployment.
This research demonstrates that the interaction between the structure and the deployment mechanism must be minimal to ensure the predictability of deployment, as thin-shell structures can self-deploy using stored strain energy. This study's sensitivity analysis will inform the design of future SSPP deployment mechanisms and structures. Additionally, the numerical simulation techniques developed have broader applicability beyond this specific case study to any deployable thin-shell structure.
Due to the large aspect ratios of thin-shell structures, a very fine finite element mesh is required to model them accurately. A dense finite element mesh is also required to model the contact interactions between the structure and the rigid components of the deployment mechanism. As large spacecraft structures become increasingly complex, full-scale numerical modeling becomes impractical, necessitating the search for more computationally efficient finite element methods. In this study, NURBS-based isogeometric analysis is explored, and it is shown that it is not yet worth switching to NURBS-based elements for the analysis of thin-shell deployable structures. In addition, h-adaptive meshing for quadrilateral shell elements is investigated, and more efficient refinement indicators and solution mapping techniques for nonlinear analyses are proposed and their superior performance is demonstrated using a test case of quasi-static folding of a tape spring.
This thesis fills a gap in the literature on unconstrained dynamic deployment of space structures, providing crucial insights and numerical modeling tools for further research. It establishes a knowledge and resource foundation to advance space structure design and promote more frequent use of unconstrained deployment, marking a pivotal contribution to the field and enabling safe and efficient space structure deployment. Furthermore, the study provides insights into more computationally efficient finite element methods, such as h-adaptive meshing. These insights are broadly applicable and can inform the design of future deployable structures beyond the tested cases.</p
A Deep Dive into the Connections Between the Renormalization Group and Deep Learning in the Ising Model
Get PDFThe renormalization group (RG) is an essential technique in statistical physics and quantum field theory, which considers scale-invariant properties of physical theories and how these theories’ parameters change with scaling. Deep learning is a powerful computational technique that uses multi-layered neural networks to solve a myriad of complicated problems. Previous research suggests the possibility that unsupervised deep learning may be a form of RG flow, by being a layer-by-layer coarse graining of the original data. We examined this connection on a more rigorous basis for the simple example of Kadanoff block renormalization of the 2D nearest-neighbor Ising model, with our deep learning accomplished via Restricted Boltzmann Machines (RBMs). We developed extensive renormalization techniques for the 1D and 2D Ising model to provide a baseline for comparison. For the 1D Ising model, we successfully used Adam optimization on a correlation length loss function to learn the group flow; yielding results consistent with the analytical model for infinite N. For the 2D Ising model, we successfully generated Ising model samples using the Wolff algorithm, and performed the group flow using a quasi-deterministic method, validating these results by calculating the critical exponent \nu. We then examined RBM learning of the Ising model layer by layer, finding a blocking structure in the learning that is qualitatively similar to RG. Lastly, we directly compared the weights of each layer from the learning to Ising spin renormalization, but found quantitative inconsistencies for the simple case of nearest-neighbor Ising models
Impacts of Zonal Asymmetry on Southern Ocean Dynamics and Biogeochemistry
Get PDFThe Southern Ocean is a key region for the ventilation and formation of intermediate and deep water masses. Interactions of the Southern Ocean’s Antarctic Circumpolar Current (ACC) with bathymetry can result in the diversion and compaction of frontal currents, resulting in standing meanders associated with enhanced mesoscale eddy kinetic energy (EKE) and submesoscale vertical velocities. As a result, standing meanders are thought to shape uptake and sequestration of heat and carbon across the ACC. In this thesis, I use observations from remote sensing and various autonomous underwater vehicles to investigate how physical mechanisms, from the submesoscale to the basin scale, shape the biogeochemical properties and tracer distributions of the Southern Ocean.
Processes at the ocean's submesoscale can play a vital role in exchanging water across the base of the mixed layer, contributing to water mass ventilation. Data from over 20,000 profiles from biogeochemical-Argo floats across the ACC highlight that the high EKE regions associated with standing meanders have relatively reduced apparent oxygen utilization (AOU) values below the base of the mixed layer. This result, as well as larger AOU variance in deep potential density classes, suggests there is enhanced ventilation occurring in standing meanders. Further investigation suggests the observed increased ventilation is due to both along-isopycnal stirring and enhanced exchange across the base of the mixed layer by vertical velocities at the submesoscale, highlighting the importance of standing meanders in shaping temporal and spatial variability of biogeochemical cycles and air-sea exchange.
Observations with horizontal scales of 2-4 kilometers in the standing meander associated with Crozet Plateau show that submesoscale processes are indeed ubiquitous. In this region, processes on the submesoscale to mesoscale spectrum play a role in enhancing surface frontal gradients and heightening tracer variability at depth. A separate field program provided novel observations at the submesoscale in Drake Passage during wintertime. The Polar Front, one of the major fronts of the ACC, is shown to be eddy-suppressing, suggesting that along-isopycnal submesoscale processes contribute to ventilation at the front. These spatial variations in stratification may additionally impact carbon fluxes between the atmosphere, surface mixed layer, and interior ocean. This thesis presents evidence that ventilation is a heterogeneous process across the Southern Ocean, with contributions from processes at physical scales that are undersampled by current observational programs.</p