Autonomous systems have the potential to improve safety and efficiency in safety-critical domains such as transportation, and medicine. Since human lives are at risk in these applications, we require rigorous safety validation before deployment. Traditional safety validation approaches such as real-world testing and scenario-based testing in simulation are not scalable to complex systems and environments and may miss unforeseen failures. Formal verification techniques also lack the scalability required for large scale autonomy. The thesis address the safety validation problem with black-box sampling techniques, which assume no knowledge of the design of the autonomous system. The system takes actions in a stochastic environment and failures are discovered by sampling environmental disturbances. The black-box assumption allows for better scalability to complex autonomous systems and sampling can be combined with machine learning to discover unforeseen failures. Previous black-box safety validation approaches have been based on optimization, path-planning, reinforcement learning and importance sampling. Although successful for many safety validation applications, existing algorithms may have poor interpretabiliy, scalability, and efficiency. Black-box sampling approaches can provide example failure trajectories but do not provide a high-level description of failures, as scenario-based approaches do. We present a new technique for generating failure descriptions in the form of signal temporal logic specifications on the environment disturbances. The specifications are optimized with genetic programming to produce failure examples and can used to gain insight into why a failure occurred. A key contribution of this thesis is the proposal and analysis of a state-dependent sampling distribution to approximate the distribution over failures. The use of the state of the environment produces a more efficient sampling distribution than baseline importance sampling approaches, but may be limited by the size of the state space. To improve scalability, we propose a decomposition technique for multi-agent validation tasks. Each subproblem is solved independently and the results are combined for better performance than learning from scratch. During the design of an autonomous system, safety validation is performed repeatedly, requiring a large computational expense. We propose a transfer learning technique that can reduce the number of required samples and lead to better performance. Knowledge from previous validation tasks is transferred to new tasks in the form of value functions that are combined using a learned set of attention weights. Results show improved knowledge transfer between tasks compared to baseline techniques. The safety validation algorithms presented in this work are tested on two gridworld scenarios and two driving scenarios. A simple gridworld scenario is used to illustrate important safety validation concepts while a gridworld with multiple adversaries is used as a test case for multi-agent validation. A rules-based autonomous driving policy is tested in a crosswalk scenario with a pedestrian and a T-intersection scenario with multiple vehicles. It is shown that the presented algorithms can improve the interpretability, scalability, and efficiency of safety validation

The three-dimensional (3D) organization of the genome is important for cellular function, such as gene expression and differentiation throughout development. Both the spatial and temporal expression of a gene are largely regulated by non-coding sequences in the genome. The genome is folded into compartments, topological associated domains (TADs), and loops, as determined by sequencing-based technology such as Hi-C. Many of the differences in cell type arise from specific interactions between distal enhancers and their target promoters, which are typically located thousands to hundreds of thousands of basepairs apart. Long-range enhancer and promoter activity and the specific of enhancer-promoter interactions are believed to arise from the cell-type specific genome folding. How this genome organization is established and regulated during development is not well understood. Hi-C and other sequencing-based assays lack information pertaining to the spatial organization of cells in tissues, and largely provide population-level information, not single cell, which makes it challenging to understand how genome folding might contribute to differences among cell types. Thus, there is a great need for approaches that provide a view of the chromatin organization and transcriptional activity in single cells. Here, I present my work developing and using a super-resolution technique to gain such an unprecedented view. Our novel super-resolution microscopy approached termed Optical Reconstruction of Chromatin Architecture (ORCA) to trace the DNA path in steps from 30 kb to 2 kb at the single-cell level. We discovered that single cells do have TAD-like structures that are heterogeneous across cells. However, the boundary positions of these single cell TADs do preferentially lie at insulator boundary protein CTCF and cohesin binding sites. Although depletion of cohesin is crucial for the presence of TADs at the population-level, we found that the TAD-like domains in single cells are not dependent on cohesin. Thus, my findings using ORCA in cultured cells (Chapter 2) shed important new light to genome organization in single cells. My interest in gene regulation led me to expand our microscopy approach by making ORCA compatible with multiplex RNA imaging to enable direct correlation between chromatin structure and gene expression on a cell-by-cell basis. Furthermore, I expanded our experimental system by applying ORCA to cryosectioned Drosophila embryos to investigate the role of 3D genome structure in loci, such as in the bithorax complex (BX-C), with well-studied enhancers. I discovered that cell-type specific 3D DNA folding of the BX-C correlates with BX-C expression patterns in different embryonic body segments. Using embryos with genetic perturbations allowed me to determine that the genetic elements at TAD boundaries drive proper cell-type specific enhancer-promoter contacts and gene expression. My results (Chapter 3) suggest that architectural proteins, such as CTCF and cohesin, at TAD boundaries are responsible for the establishment of 3D organization during development. Additionally, my results emphasize the need to study cell-type specific chromatin structures on a cell-by-cell and cell type basis, an area that is still largely unexplored. To facilitate such exploration, I worked towards making our approach accessible to other researchers that are interested in 3D genome architecture and transcriptional activity (Chapter 4). To determine the role of architectural proteins in genome organization (Chapter 5), I took advantage of Drosophila genetics and obtained null allele mutant embryos that lacked zygotic expression of architectural proteins such as Rad21, Wapl, CTCF, and CP190. Using ORCA, I found that these mutants have BX-C TADs that are similar to that of WT in mid to late staged embryos. However, as the maternal transcripts for these architectural proteins were present throughout embryogenesis, the maternally encoded proteins appeared to be sufficient to retain genome structure in the zygotic null mutants. I also observed BX-C TAD structure that looks similar to that of wild-type in the central nervous system (CNS) of mutant CTCF L3 larvae where maternal gene products were fully absent. My results raise the probability that other Drosophila insulator binding proteins, such as CP190, may play a redundant insulation function. To examine the role of various cis-acting insulator elements, I have begun preliminary studies in investigating how inserting insulators into the genome affects long-range cis-regulatory interactions (Chapter 6). Overall, the development of ORCA has enabled us to begin understanding the mechanisms underlying genome organization and their role in regulating transcription in a complex tissue. As our techniques improve and becomes more accessible to other researchers in the field, we are certain that the methods we have developed will play a role in un-covering the function of various chromatin components, such as transcription factors and epigenetic state, in establishing the 3D genome organization during development

Chromatin architecture plays an important role in gene regulation. Recent advances in super-resolution microscopy have made it possible to measure chromatin 3D structure and transcription in thousands of single cells. However, leveraging these complex datasets with a computationally unbiased method has not been achieved. In this dissertation, I present a deep learning-based approach to better understand to what degree chromatin structure relates to the transcriptional state of individual cells. Furthermore, I explore methods to "unpack the black box" to determine in an unbiased manner which structural features of chromatin regulation are most important for gene expression state. I apply this approach to the Optical Reconstruction of Chromatin Architecture dataset of the Bithorax gene cluster in Drosophila and show it significantly outperforms previous contact-focused methods. This work finds the structural information is distributed across the domain, overlapping and extending beyond domains identified by prior genetic analyses. Individual enhancer-promoter interactions are a minor contributor to predictions of activity

Allogeneic hematopoietic stem cell transplant (HSCT) can be used as a curative treatment for patients with hematologic malignancies; however, acute graft-versus-host disease (aGVHD) is a highly inflammatory disease and a major complication that can occur following this therapy. Two populations capable of regulating immune responses are invariant natural killer T (iNKT) and regulatory T (Treg) cells, but their biology in the clinical aGVHD setting is not fully understood. Thus, I examined how iNKT and Treg cells regulate aGVHD in the clinical setting. To study human iNKT cells, I used a high-dimensional, data-driven approach to devise a framework for parsing human iNKT cell heterogeneity. A population of iNKT cells characterized by HLA-DR expression was found to be enriched in aGVHD patients and associated with exhaustion markers and a Th1 profile, suggesting a loss of immunoregulatory mechanisms in aGVHD. This phenotype, along with other discovered phenotypic markers, were also associated with aGVHD outcomes, suggesting that iNKT cells can potentially serve as a biomarker and therapeutic target in allogeneic HSCT. Beyond these potential clinical implications, a new framework was discovered for understanding human iNKT cell heterogeneity that advances our knowledge of human iNKT cell biology. As our understanding of fundamental human Treg cell biology is more advanced, a phase I/II study of the infusion of Treg cells with hematopoietic stem cells to prevent aGVHD in adult patients undergoing myeloablative allogeneic HSCT for hematological malignancies has been started and is ongoing. Clinically, greater than grade II aGVHD was noted in two patients in the first cohort of five patients who received cryopreserved Treg cells. In a second cohort of seven patients who received fresh Treg cells and single agent aGVHD prophylaxis, none developed aGVHD. A key finding is that patients in the second cohort showed immune reconstitution comparable to patients who underwent allogeneic HSCT on standard-of-care. Together, these studies reveal new insights into the biology of human iNKT and Treg cells in the clinical aGVHD setting

Meiotic recombination plays dual roles in the evolution and stable inheritance of genomes: recombination promotes genetic diversity by reassorting variants, and it establishes temporary connections between pairs of homologous chromosomes that ensure their future segregation. Meiotic recombination is initiated by generation of double-strand DNA breaks (DSBs) by the conserved topoisomerase-like protein Spo11. Despite strong conservation of Spo11 across kingdoms, auxiliary complexes that interact with Spo11 complexes to promote DSB formation are very poorly conserved. Here, I identify DSB-3 as a DSB promoting protein in C. elegans. Mutants lacking DSB-3 are proficient for homolog pairing and synapsis but fail to form meiotic crossovers. Lack of crossovers in dsb-3 mutants reflects a requirement for DSB-3 in meiotic DSB formation. DSB-3 concentrates in meiotic nuclei with timing similar to DSB-1 and DSB-2 (putative homologs of yeast/mammalian Rec114/REC114), and DSB-1, -2, and -3 are interdependent for this localization. Bioinformatics analysis and interactions among the DSB proteins support the identity of DSB-3 as a homolog of MEI4 in conserved DSB-promoting complexes. This identification is reinforced by colocalization of pairwise combinations of DSB-1, -2, and -3 foci in structured illumination microscopy images of spread nuclei. However, unlike their counterparts in mouse and yeast, DSB-1 can interact directly with SPO-11, and C. elegans DSB protein complexes are not concentrated at meiotic chromosome axes. I suggest that rapid divergence of protein complexes that serve essential, yet auxiliary, roles in meiotic recombination may reflect an immediate impact of changes in such proteins on processes that directly affect reproductive success

Chytrids are one of the only known groups of ciliated fungi and have important ecological roles in soil and aquatic environments. However, very little is known about their cell biology. In order to make chytrids more amenable to study, I have developed various tools to facilitate the use of chytrids as model organisms. Focusing on the non-amoeboid chytrid Rhizoclosmatium globosum and the amoeboid chytrid Spizellomyces punctatus, I have characterized these chytrids' life cycles, optimized imaging techniques, and made preliminary attempts at transformation. I have also identified select genes in their genomes and found that both R. globosum and S. punctatus have most of the expected centrosome and cilium associated genes. However, only S. punctatus has a full set of branched actin nucleation related genes associated with amoeboid motility. After developing these chytrid related tools and resources, I next sought to use these chytrids to study cilium disassembly. Cilia are protrusions from cells containing a tubulin-based structure called an axoneme. They can be involved in motility and sensory functions; however, due to the modified centrioles found at their bases that are used in mitosis, cilia need to be disassembled for progression through the cell cycle. Cilium disassembly is an understudied process, and most of what has been discovered has been found in Chlamydomonas reinhardtii and cultured mammalian cells, which undergo two types of cilium disassembly: gradual shortening and fast severing. However, while chytrids can sever their cilia under stressful conditions, they generally undergo a separate mechanism where their cilia are retracted into their cell body. Thus, a non-model organism is required to study a process that cannot be found in common model systems. I have found that R. globosum and S. punctatus have a common cilium disassembly mechanism of reeling in their axonemes into their cell bodies. However, amoeboid S. punctatus has additional, faster mechanisms of ciliary compartment loss and lash-around retraction, and its initiation of retraction is at least partially actin dependent. Once the axoneme is internalized, its tubulin is degraded over the course of about two hours in both species in a partially proteasome dependent manner. Axoneme disassembly also seems to be partially proteasome dependent, indicating that it is coupled with axonemal tubulin degradation. Overall, I was able to develop tools for chytrid fungi to better characterize cilium retraction and the fate of internalized axonemes

Health care costs in the United States exceed $3.5 trillion annually, with between $760 billion and $935 billion considered waste. Data-driven analytics could reduce costs and provide higher quality care to patients by more efficiently allocating limited resources, just as analytics has done in other industries such as logistics, manufacturing and aviation. In this dissertation, I demonstrate three levels at which analytics provide value in health: clinical decision making, healthcare operations management and public health policy. Clinical decision making refers to decisions at the individual patient level: for example, determining which treatment to provide a patient or predicting an individual's risk of disease. Healthcare operations management refers to decisions about the system that delivers care to patients: for example, determining how to organize patient flow through a hospital or schedule procedures. Finally, public health policy refers to decisions about the overall health of a population: for example, determining how to control an infectious disease or distribute limited resources across different diseases

Tactile sensing is paramount for robots operating in human-centered environments to help in understanding interaction with objects. To enable robots to have sophisticated tactile sensing capability, researchers have developed different kinds of tactile sensors for robotic hands to realize the 'sense of touch'. In this study, we are focused on the incipient slip detection problem for robots which is known as one of the most challenging issues in robotic tactile sensing. Currently, most of the slip detection sensors are passive sensors which provide limited information about the sensing parameters. Therefore, this will usually require large amount of data and extra computation effort in accurately classifying slip conditions of robotic hands. Other sensing mechanisms such as optical approaches which can provide enriched sensing parameters for slip detection often suffer from complex sensor configurations and being inflexible in terms of customization. Active sensing, on the other hand, has the advantage of simple sensor configurations, and in the meantime can provide more sensing parameters which will improve the overall efficiency of the tactile sensing capabilities for incipient slip detection. In this thesis, by using the active sensing method, a novel active sensing smart skin technique is developed for incipient slip detection which leverages piezoelectric transducers as actuators/sensors. With this method, a robotic fingertip with the embedded actuator and sensor were created in which the actuator generates ultrasonic guided waves received by the sensor during a slip scenario. By analyzing the received signal using an attenuation-based method, we can monitor the entire contact area evolution during a slip scenario. Therefore, this method can serve as an excellent indicator for early slip detection with the advantage of accurately monitoring the contact condition. In addition, the frustrated total internal reflection method was used to validate the signal attenuation increases with the growing of the contact area. Built on these results, a unique robotic skin was then designed and fabricated which demonstrated robust and sensitive response for incipient slip detection. Finally, an LED slip alert system on a real gripper was developed to demonstrate the capability of our method to be applicable to real robotic finger situations

The widespread prevalence of rising economic inequality across western democracies has led to immense academic and policy interest, as well as the rapid development of the tools required to study it. Researchers are now equipped with rich data and advanced computational methods which are well-suited to analyzing the processes underlying the extensive differences exhibited across individuals and groups within and between societies. To an extent, diverse outcomes reflect an intrinsic natural variation in individual tastes and preferences. However, in many cases we rather consider inequalities, particularly economic inequalities, to reflect injustice, misallocation and constrained opportunities. When considering labor market earnings, a substantial proportion of the variation across individuals can be explained by a single predictor: a worker's gender. In the first chapter of this dissertation I study a policy explicitly designed to reduce this association, in which employers are required to publicly report gender pay gap statistics. Proponents argue that increasing the information available to workers and consumers places pressure on firms to close pay gaps, but opponents argue that such policies are poorly targeted and ineffective. I contribute to the debate by analyzing the UK's recent reporting policy, in which employers are mandated to publicly report simple measures of their gender pay gap each year. Exploiting a discontinuous size threshold in the policy's coverage, I apply a difference-in-difference strategy to linked employer-employee payroll data. I find that the introduction of reporting requirements led to a 1.6 percentage-point narrowing of the gender pay gap at affected employers. This large-magnitude effect is primarily due to a decline in male wages within affected employers and is not caused by a change in the composition of the workforce. To explain this effect, I propose that a worker preference against high pay gap employers induces the closing of pay gaps upon information revelation. Newly-gathered survey evidence shows that female workers in particular exhibit a significant preference for low pay gap employers. In a hypothetical choice experiment, over half of women accept a 2.5\% lower salary to avoid a high pay gap employer. I also demonstrate substantial heterogeneity in the interpretation of pay gap statistics across workers and show that this affects their valuation of jobs at employers with different pay gaps. Does the importance of your family background on how far you get in adulthood also depend on where you grow up? For England and Wales, a paucity of data has made this a difficult question to reliably answer. My second chapter, co-authored with Brian Bell and Stephen Machin, presents a new analysis of intergenerational mobility across three cohorts in England and Wales using linked decennial census microdata. These data permit the study of different mobility outcomes in occupation, home ownership and education, at the spatial level through time. As well as showing national results consistent with previous studies, we find strong sub-regional patterns in mobility, with four main results emerging. First, area-level differences in upward occupational mobility are highly persistent over time. Second, consistent with evidence from other countries, absolute and relative mobility are positively correlated for all measures and particularly strongly for home ownership. Third, there is a robust relationship between upward educational and upward occupational mobility. Last, there is a small negative relationship between upward home ownership mobility and upward occupational mobility, revealing that social mobility comparisons based on different outcomes can have different trends. Social scientists have long been interested in the relationship between parental factors and later child income. Finding the best characterization of this relationship for the question at hand is however fraught with choices. In my third chapter, co-authored with Erling Risa, we use machine learning methods to assess the `completeness' of one popular modelling approach. Here, completeness refers to how well the model summarizes the total predictive relationship between multiple parental factors and a single child outcome. Machine learning methods enable us to depart from functional form assumptions, allowing flexible interactions between a large set of possible parental factors. Using our most flexible complete model as a benchmark, we assess the popular `rank-rank' model relating parent and child incomes. Applying our approach to high-quality Norwegian administrative data, we demonstrate that the rank-rank model explains 68\% of the total explainable variation in child income rank, based on a broad set of potential parental factors entering a neural network. Parental wealth and parental education explain the majority of the remaining explainable variation. At the regional level, we estimate homogeneous completeness across areas. Rankings of areas based on rank-rank slope estimates align with those based on the predictive fit of the broader flexible model

Modeling the hydromechanical behavior of geological materials infiltrated by one or more types of fluids is critical in many scientific and engineering applications, including geologic carbon sequestration, hydrocarbon recovery, and geotechnical analysis. For some of these applications, it is sufficient to consider only the solid deformation problem or the fluid flow problem separately. Even then, the solution could encounter difficulties, both theoretically and numerically. Other applications require simultaneous solution of solid deformation and fluid flow phenomena, thus creating enormous computational challenges in that one needs to solve different physical problems at the same time. This thesis addresses some of the computational challenges associated with modeling coupled multiphase and multi-physical systems in poromechanics. The first part of this thesis focuses on how to improve the solution accuracy for a particular type of discretization that uses the finite element method to solve the solid deformation problem and the finite volume scheme for the fluid flow. This type of approximation is quite common; however, it may lead to unphysical oscillations in the pore pressure field. To suppress the spurious oscillations, we devise a stabilized formulation. This work is one of very few studies of stabilization procedures addressing multiphase poromechanics in the context of mixed finite element-finite volume method. The designed stabilization supplements the mass balance equations with stabilizing fluxes inside patches of elements called macroelements. The method is simple to implement in an existing code and conserves mass within the macroelements. We show that the stabilization effectively treats the pervasive oscillations associated with the presence of near-singular modes. Additionally, a direct result of removing the near-singular modes is the dramatical improvement on the convergence of iterative linear solvers. The second part of this work aims to develop an efficient solver for coupled solid deformation-fluid flow problems where capillarity forces are strong. For these cases, the saturation of the invading fluid spreads more diffusively. This change in the behavior of saturation variables has to be taken into account in the design of efficient solvers. In a coupled simulation, a major part of the overall computational time is devoted to solving the large system of linear equations that arises from the discretization of the governing equations. By designing a more efficient iterative linear equation solver for the problem, we reduce the computational cost of large-scale coupled simulations. In this work, the efficiency of the iterative linear solver is improved by designing three preconditioning approaches that are more suitable to the strong capillary pressure regime. The approaches are based on a block factorization of the coupled system with sparse Schur complement approximations. The strategies differ on how they approximate the Schur complement used to correct the saturation unknowns. When the approximation relies on a multilevel preconditioner, we show that the linear solver reduces the computing time by a factor of two relative to the base case strategy, which uses a Jacobi approximation. The final part of this work builds upon the previously developed numerical framework to analyze a potential offshore carbon storage site in the Gulf of Mexico. The goal of the study is to assess appropriate deformation monitoring techniques that have sufficient sensitivity to measure the seabed uplift induced by fluid injection into the reservoir. Before deploying such measuring instruments in an offshore environment, it is essential that we first simulate the coupled solid deformation-fluid flow phenomena as realistically as possible to gain some insight into the precision of instruments needed for such field study. To this end, we create a finite element-finite volume model of a potential offshore site that incorporates a spatially heterogeneous permeability distribution expected at this site. We conduct a sensitivity analysis on the geomechanical parameters and investigate the influence of fault sealing on the fluid flow. Results of the simulations suggest that the reservoir has a remarkable injectivity, allowing for a large amount of fluid to be injected without a substantial increase in pore pressure. The computed seabed deformations are equally small and could be monitored with distributed fiber optic cables and ocean bottom pressure recorders

I have been broadly interested in the molecular, cellular, and developmental mechanisms underlying neural circuit assembly in mammals. My graduate research has focused on the cerebellum, and I have pursued a variety of genetic and other approaches to investigate the mechanisms underlying the assembly of cerebellar circuits. My work has had three primary directions: (1) characterizing the role of birth timing in the wiring and function of cerebellar granule cells; (2) understanding the role of synapse formation in cerebellar Purkinje cell dendrite morphogenesis; and (3) developing new methods for profiling the cell surface proteomes of specific cell types in mice and applying these methods to characterize the cell surface proteomes of developing and mature Purkinje cells. In the first line of research, I probed the role of birth timing in cerebellar granule cell wiring and function. Cerebellar granule cells comprise over half of all neurons in the mammalian brain, yet there are no reports of molecularly-defined subtypes within cerebellar regions. However, previous work revealed that granule cells project their axons in a manner dictated by their birth timing. Thus, I developed strategies to gain genetic access to early- and late-born granule cells and performed viral tracing and two-photon imaging in vivo to reveal the wiring and functional properties of early- and late-born granule cells. Retrograde monosynaptic rabies virus tracing revealed different patterns of mossy fiber input to granule cells in different lobules, as well as to early- and late-born granule cells of the same lobule. Imaging revealed representations of diverse task variables and stimuli by both populations, with differences in the proportions of early- and late-born cells encoding a subset of movement and reward parameters. Taken together, these data suggest neither organized parallel processing nor completely random organization of mossy fiber--> granule cell circuitry, but instead a moderate influence of birth timing on granule cell wiring and encoding. In the second line of research, I addressed the role of synapse formation in dendrite morphogenesis of cerebellar Purkinje cells. The synaptotrophic hypothesis of dendrite growth, a hypothesis more than 30 years old, posits that dendrites preferentially grow to-ward rich synaptogenic fields. However, this hypothesis had not been causally tested in the mammalian central nervous system in vivo. We reasoned that by disrupting GluD2, a synaptogenic protein expressed in Purkinje cells and involved in formation and maintenance of parallel fiber--> Purkinje cell synapses, we could test this hypothesis in the mammalian brain. Strikingly, sparse but not global knockout of GluD2 resulted in a curious phenotype in which Purkinje cell dendritic arbors grow into an inverted triangular shape, under-elaborating in the deep molecular layer and overelaborating in the superficial molecular lay-er. Developmental, overexpression, structure-function and genetic epistasis experiments, along with modeling, revealed the importance of GluD2's synaptogenic activity in dendritic morphogenesis, supporting an important role for molecular interactions crucial for synapse formation and function in dendrite morphogenesis in the mammalian brain. In the third and final line of research described here, I developed methodology featuring a new mouse line and accompanying protocols for unbiased proteomic profiling of extracellular cell surface proteins of defined cell types in mice via proximity labeling. To date, there have been no methods for cell-type specific, unbiased capture of cell-surface proteomes in situ. Our method works in diverse organs and cell types and has allowed us to generate high quality cell surface proteomes of developing and mature cerebellar Purkinje cells. This work is being followed by a proteome-instructed CRISPR-mediated loss-of-function in vivo screen to discover novel molecular regulators of dendrite morphogenesis. This method constitutes a valuable new approach for profiling cell surface proteins in de-fined cell types in mice. In summary, my thesis work utilizes genetic and other approaches to pursue mechanistic interrogation and technological development to gain insight into the molecular, cellular, and developmental mechanisms of cerebellar circuit assembly

Donated blood is a critical component of health systems around the world, but its collection and transfusion involve risk for both donors and recipients. Transfusion-transmitted diseases and non-infectious adverse events pose a risk to transfusion recipients, and repeat blood donation can cause or exacerbate iron deficiency among donors. This dissertation describes four decision-analytic modeling projects that inform blood safety policy. In Chapter 2, I integrate epidemiological, health-economic, and biovigilance data to estimate the efficacy and cost-effectiveness of a 2016 policy mandating that all blood donations are screened for Zika virus in the U.S. The analysis uses a novel microsimulation of individual transfusion recipients that captures the relationship between disease exposure risk and the number and type of blood components transfused. In Chapter 3, I develop the first health-economic assessment of whole blood pathogen inactivation. The analysis is for Ghana and improves on prior blood safety assessments for sub-Saharan Africa by considering the likelihood and timing of clinical detection for chronic viral infections. In Chapter 4, I develop an optimization-based framework for identifying the optimal portfolio of blood safety interventions that overcomes some limitations of traditional cost-effectiveness analyses for blood safety. By applying this framework retrospectively to evaluate U.S. policies for Zika and West Nile virus, I show that the optimal policy can vary by geography, season, and year. Chapter 5 focuses on how frequently donors are allowed to give blood. I develop a machine learning-based decision model that tailors the inter-donation interval to each donor's risk of iron-related adverse outcomes to balance risks to donors against risks to the sufficiency of the blood supply. Together, these model-based analyses introduce novel methods and provide guidance for efficient and effective use of resources for blood safety

For billions of years, organisms and their enzymes have been evolving and adapting in response to selection pressures and opportunities presented by their environments. The mechanisms that underly enzyme adaptation are fundamental to our understanding of how enzymes work and the process of molecular evolution. Enzyme adaptation occurs through changes in the molecular interactions that specify enzyme structure and function. In particular, hydrogen bonds are central to how enzymes have evolved and are ubiquitous in enzyme structures, where they are often found in extended networks. Given the ubiquity and importance of hydrogen bonds, this dissertation seeks to understand the physical principles that govern hydrogen bond structure and energetics and their contribution to enzyme function and adaptation. I first define the physical principles that determine hydrogen bond lengths and energies. I then investigate the extent that hydrogen bonds in networks are coupled, structurally an energetically, to one another, improving our understanding of hydrogen bonds and the local and collective structural dynamics that exist within proteins and providing quantitative benchmarks for the design of hydrogen bonds and their networks. Second, to address the interplay between hydrogen bonds and other molecular interactions and enzyme evolution, I investigated enzyme temperature adaptation, as temperature profoundly influences enzyme stabilities and activities, providing a direct link between an environmental selection pressure and enzyme function. Performing deep mechanistic and structural studies and broad phylogenetic and sequence analyses, I identify new molecular mechanisms of enzyme temperature adaptation, test and revise nearly all prior proposals for enzyme adaptation to increased temperature, provide evidence that enzymes frequently adapt to temperature at sites of low epistasis, favoring parallel adaptation. This work improves our understanding of the molecular and evolutionary mechanisms underly enzyme evolution and suggests design principles that can be applied to enzyme engineering

Anthropogenic carbon dioxide (CO2) emissions have driven widespread ocean acidification (OA). OA has reduced surface ocean pH by at least 0.1 pH units since the beginning of the industrial era and global models forecast a further decrease of 0.3 to 0.4 pH units by the end of the century. Submerged aquatic vegetation, such as kelp forests and seagrass beds, has the potential to locally ameliorate OA by removing CO2 during photosynthesis and storing it as fixed carbon. Thus, understanding the contribution of these habitats to local biogeochemistry is essential to inform coastal management and policy, especially as the impacts of anthropogenic climate change become more prevalent. The following work describes high resolution spatiotemporal variability in seagrass and kelp forest biogeochemistry (Chapters 1 and 2) and in the surface canopy extent of a giant kelp forest (Chapter 3). In order to understand the contributions of kelp forest and seagrass metabolism to their respective local biogeochemistry, we must determine the natural variability in these systems and disentangle the physical and biological drivers of local biogeochemical variability. In Chapter 1, I deployed an extensive instrument array in Monterey Bay, CA, inside and outside of a kelp forest to assess the degree to which kelp locally ameliorates present-day acidic conditions, which we expect to be further exacerbated by OA. Interactions between upwelling exposure, internal bores, and biological production shaped the local biogeochemistry inside and outside of the kelp forest. Significantly elevated pH, attributed to kelp canopy productivity, was observed at the surface inside the kelp forest. This modification was largely limited to a narrow band of surface water, implying that while kelp forests have the potential to locally ameliorate ocean acidification stress, this benefit may largely be limited to organisms living in the upper part of the canopy. In Chapter 2, I quantified net community production (NCP) over a mixed seagrass-coral community on Ngeseksau Reef, Ngermid Bay, Republic of Palau. We observed a net heterotrophic diel signal over the deployment, but dissolved oxygen (O2) fluxes during the day were largely positive, illustrating daytime autotrophy. pH, O2, and temperature followed a clear diel pattern with maxima typically occurring in the afternoon. The relationship between tidal regime and time of day drove the magnitude of the signals observed. The case studies described in Chapters 1 and 2 emphasize the importance of high-resolution measurements (high temporal frequency as well as high horizontal and vertical spatial resolution) and consideration of the multiple drivers responsible for shaping the observed biogeochemical variability. In addition to the photosynthetic biomass (kelp and seagrass) at the center of these studies, the physical environment played an important role in dictating the signals observed, in particular water circulation and residence time. Biogeochemical studies rarely look beyond a few deployment sites, but the ecosystem contributing to the local biogeochemical variability includes influences from beyond those discrete points. Describing the area around these discrete points is important for accurate assessment of factors driving the signals observed at those points. Remote sensing can help us capture and describe the spatial patterns of biomass contributing to changes observed in our chemical records. In Chapter 3, I established a low altitude unmanned aerial vehicle (UAV) record of giant kelp surface areal extent over 18 months on the wave-protected side of Cabrillo Point (Hopkins Marine Station) in Monterey Bay, CA. This was the same canopy responsible for elevating pH in Chapter 1; however, in this case, the kelp canopy mapping did not overlap in time with biogeochemical measurements in the kelp forest. I compared the UAV kelp classification to canopy cover determined from Landsat satellite images obtained over the same period. There was a linear relationship between the drone kelp ratio and Landsat kelp canopy fraction for spatially-matched pixels; a Landsat kelp fraction of 0.64 was equivalent to 100% kelp cover in the drone data. The level of resolution provided by UAV, compared with Landsat images, could allow more detailed mapping of kelp responses to environmental change. Future studies should pair mapping flights with biogeochemical measurements to quantify the relationship between changes in canopy area and the relative surface canopy modification of pH.  

Wave-equation-based parameter estimation techniques can retrieve accurate and high-resolution subsurface physical properties from seismic data acquired close to the surface of the Earth. In fact, multiple acoustic full-waveform inversion methods have been proposed over the years to retrieve the P-wave velocity of the subsurface. Moreover, researchers have extended full-waveform inversion approaches to estimate anisotropic and absorption parameters as well. Nowadays, some applications of elastic full-waveform inversion can also be found. However, given its prohibitive computational cost compared to the acoustic counterpart, elastic wave-equation inversion workflows still have limited applicability within seismic exploration datasets. To tackle this challenge, I propose a novel wave-equation-based elastic parameter estimation workflow based on wave-equation operators. I refer to the entire approach as target-oriented elastic full-waveform inversion. The method is composed of two steps. In the first one, I apply an extended linearized waveform inversion to the surface data. The obtained subsurface image is then employed to synthesize data as if they were acquired close to a target area. Finally, this dataset is inverted using an elastic full-waveform inversion workflow to estimate the subsurface elastic parameters. I demonstrate its efficacy on a 2D synthetic test and an ocean-bottom-node dataset acquired in the Gulf of Mexico, showing its ability to retrieve the elastic parameters of potential subsurface prospects. Compared to the elastic inversion of the surface dataset, the proposed method has a computational cost lower by two orders of magnitude

Today's data-driven applications, such as big data analytics, neural networks, and machine learning, require huge memory and computing resources. The latency and energy incurred for data movement between processing unit and memory has become a significant limitation (known as the "memory wall") as traditional techniques such as caching are no longer effective for the data-intensive applications that dominate modern computing. Solving the "memory wall" problem requires future computer technology to directly integrate memory and transistor devices with high density vertical interconnect accesses (vias) to provide parallel, high bandwidth memory access. However, the high temperature of silicon processing is not compatible with this 3D monolithic integration. A desired low temperature integration can be achieved by using two-dimensional (2D) materials which have intrinsically layered and flat atomic structures, because devices made of 2D materials can be fabricated at lower temperatures. This thesis introduces three exploratory studies on advancing 2D materials for developing 3D monolithic integrated systems. First, I focused on improving transistor performance and studied high-mobility 2D material black phosphorous (BP) transistors using various metals as the contact metal. This work achieved unipolar n-type BP transistors by using ultra-low work function metals, demonstrating record high n-type current. Furthermore, the study revealed the physical mechanisms of controlling doping and de-pinning effects for n- and p- type BP transistors. Beyond transistor studies, I then demonstrated the first 3D sequential monolithic integrated two levels of 1-transistor-1-resistor (1T1R) memory cells. The cell is fabricated entirely using 2D materials: hexagonal-Boron Nitride (hBN) serves as the resistive switching memory cell and monolayer Molybdenum Disulfide (MoS2) serves as the channel material for the transistor selector. In order to achieve a high-density memory array, a two-terminal selector with a high nonlinear current-voltage characteristic is necessary. I developed a new two-terminal selector utilizing a 2D material heterostructure with an H-shape energy barrier. This new device design in theory should have high on-state current density and virtually unlimited endurance due to its quantum tunneling mechanism. Our results characterized the first out-of-plane current through an ultra-thin (3 monolayer) heterojunction and provide the critical foundation for a high endurance selector using a 2D heterojunction. Due to the transferable feature of 2D materials, these processes can be easily adopted for 3D monolithic integrated transistors and memories in multiple logic and memory layers connected vertically by fine-grained nanoscale vias, thereby overcoming the "memory wall.".