545298 research outputs found
Sort by
Doubly Robust Methods for Policy Learning: Observational and Data Combination Settings
Get PDFLearning optimal individualized treatment rules (ITRs) is an important problem in causal inference, with applications in healthcare, education, public policy, and economics. Experimental data facilitate reliable causal inference but are often expensive to collect and may lack long-term outcomes. Observational data are more accessible but subject to unmeasured confounding. In addition, operational constraints, such as fairness, privacy, or feasibility, may restrict the covariates available at decision time. This dissertation develops doubly robust methods for policy learning under these practical limitations.
Chapter 1 addresses policy learning under runtime confounding, where policies must be based on a limited subset of covariates. We propose two doubly robust estimators that ensure consistent policy learning when either the outcome model or the propensity score model is correctly specified. Finite-sample regret bounds are established and extended to sparse high-dimensional settings using Lasso. The methods are evaluated through simulations and an application to the ACTG175 clinical trial.
Chapter 2 and Chapter 3 consider combining observational and experimental data to learn ITRs for long-term outcomes in the presence of persistent unmeasured confounding. Long-term outcomes are observed only in the observational data, while the experimental data provide short-term outcomes. In Chapter 2, we propose three classification-based policy estimators, including a doubly robust method that remains consistent under partial misspecification. Regret bounds are established under various settings, and the methods are validated through simulations and employment data. Chapter 3 shifts focus to estimating long-term causal effects and the associated ITRs. We develop three estimation strategies compatible with general machine learning methods and provide theoretical guarantees. We further illustrate our approaches using deep neural networks and derive theoretical results specific to this implementation. The methods are evaluated through simulations and applications to employment data
Engineering targeted nanomaterials for the treatment of traumatic brain injury
No full textTraumatic brain injury (TBI) is a significant public health challenge impacting millions worldwide annually. The initial trauma triggers a secondary injury cascade involving blood-brain barrier (BBB) disruption, neuroinflammation, and extracellular matrix (ECM) remodeling. Clinical translation of TBI therapeutics is hindered by poor pharmacokinetics and passive accumulation of systemically administered therapeutics within injured brain tissue. To address this, this dissertation explores active targeting strategies for therapeutic and nanomaterial delivery in mouse and ex vivo models of TBI.
Peptide therapeutics often face rapid clearance and poor stability. CAQK is a therapeutic peptide that also targets the ECM in injured brain tissue. To improve its pharmacokinetics, we engineered CAQK onto a polyethylene glycol (PEG) scaffold using two linker chemistries: maleimide-thiol and DBCO-azide. PEGylation extended circulation half-life and enhanced injured brain accumulation compared to free peptide. Both conjugates colocalized with tenascin-C, the putative receptor of CAQK, in the perilesional cortex. However, maleimide-linked conjugates exhibited reduced off-target retention in filtration organs, emphasizing the impact of conjugation chemistry on the pharmacokinetics of nanomaterials. In a prophylactic dosing, maleimide-linked CAQK-PEG conjugates significantly outperformed free CAQK peptide in injured brain accumulation. Building on these targeting strategies, we next developed lipid nanoparticles (LNP) for neuron-specific gene delivery after TBI, combining precise systemic dosing with RVG peptide functionalization to target neuronal receptors. Despite reduced bulk accumulation and activity in the injured brain, RVG-targeted LNPs significantly enhanced neuronal transfection over untargeted controls. This approach highlights a strategic trade-off between total transfection and cell-specific targeting for gene delivery after TBI. Finally, to extend these targeting strategies into mechanistic and therapeutic testing, we integrated an exploding bridgewire shockwave system (EBWS) with organotypic brain slice (OBS) cultures to model primary blast-induced shockwave (BIS) injury ex vivo. We characterized temporal changes in key physiological markers of cellular activation, neuroinflammation, oxidative stress, and ECM remodeling, capturing the complex molecular responses following blast injury. This model recapitulates fundamental in vivo blast TBI features whilst enabling reproducible, high-throughput studies without systemic effects. Collectively, we demonstrate promising approaches for targeted nanomaterial delivery across different models of brain injury, facilitating the advancement towards more effective clinical translation of TBI treatments
Unlocking New Pathways in Catalysis: Carbon is Key!
No full textUntil the isolation of the first phosphino silyl carbene in 1988 and the subsequent report of the first N-heterocyclic carbene (NHC) in 1991, carbenes were considered to be laboratory curiosities. Nearly four decades later, a vast library of carbenes has been isolated. The carbene’s highly tunable electronic and steric properties make them widely applicable in a variety of disciplines, which include but are not limited to: transition metal catalysis, material science, organocatalysis, and pharmacology. Prolific experimental and theoretical carbene research has also inspired exploration into a closely related class of divalent carbon species, carbones. While carbone research is still in its infancy, their unique electronic properties in comparison to carbenes has allowed for a rise in their catalytic applications in the past decade.The intricate interplay of electronic structure, stabilization, and reactivity of divalent carbon species illuminates the topic of this dissertation: the application of divalent carbon species in catalysis. Herein is an exploration of the transition metal complex facilitated valorization of feedstock molecules using carbene and carbone ligands, and the functions of carbene reductive properties as organocatalysts. Chapter 2 details the synthesis and metalation of bisimidazolium and functionalized cyclic (alkyl) (amino) carbene precursors in the pursuit of designing a more donating class of carbene-based ligands for chromium catalyzed ethylene tetramerization. Chapter 3 discusses the synthesis and preliminary results of the catalytic activity of cyclic bent allenes in rhodium catalyzed phenylacetylene cyclotrimerization. Finally, Chapter 4 explores the capacity of carbenes to function as catalysts in single electron transfer reactions
The memory of event sequences is mediated by hippocampal dynamics
No full textThe ability to temporally organize our memories and behaviors is critical to daily life functioning and is impaired across a wide range of cognitive disorders. This includes our ability to remember sequences of events from the past to solve problems in the present, but it also extends to the prediction of sequences of future steps needed to reach a goal. Considerable research has demonstrated that this capacity is conserved across species, applies across sensory modalities, and depends on a network of brain regions centered on the hippocampus. However, the neural mechanisms by which the hippocampus gives rise to this capacity remain unclear.Accumulating evidence from electrophysiological studies in rodents performing spatial navigation tasks point to potential mechanisms. Hippocampal neurons have been shown to code for the specific sequence of locations visited in a given trajectory in a maze, a phenomenon observed during both “online” and “offline” periods. Online periods refer to moments of active exploration during which hippocampal activity exhibits prominent theta oscillations (4-12 Hz), whereas “offline” periods refer to brief rest intervals between task events as well as sleep. It remains to be determined, however, whether these sequence coding properties extend to sequences of non- spatial events unfolding over several seconds, as in daily life episodes in humans.To address this critical issue, we analyzed hippocampal activity data recorded as rats performed a complex non-spatial sequence memory task with strong behavioral parallels in humans. Since existing decoding approaches to decode neural activity could not be applied to discontiguous event sequences, we developed statistical machine learning methods to quantify sequential neural representations during online and offline periods. During online periods (stimulus presentations), we found that hippocampal ensembles differentiated distinct types of task-critical information sequentially within events, and exhibit theta-associated reactivation of the sequential relationships among events. We also demonstrated that non-spatial event representations were sequentially organized within individual theta cycles and processed across successive cycles. During offline periods (intervals between stimuli), we discovered that hippocampal ensembles rapidly replayed the sequential relationships across events in either the forward or reverse direction. This was observed at the level of specific pairs of events and of the full sequence. Collectively, these findings suggest a fundamental function of the hippocampal network is to extract the sequential order of discontiguous experiences and to sequentially reactivate this information during online and offline periods to support goal-directed behavior
Causes and Consequences of Proteolytic Remodeling of Metabolism During Bacillus subtilis Sporulation
No full textCellular differentiation is a ubiquitous process in eukaryotes; however, it is seldom seen in bacteria. Sporulation in Bacillus subtilis is an example of cellular differentiation where in response to nutrient depletion, the vegetative cell differentiates into 2 distinct cell types: the larger mother cell and the smaller forespore which eventually becomes the mature spore. After the 2 cells are formed by polar septation, cell-specific sigma factors activate, sending the cells on different developmental pathways. The mother cell then engulfs the forespore into its cytoplasm, the coat and cortex assemble around the forespore and the forespore goes dormant, completing maturation. The mother cell then lyses, releasing the mature spore into the environment. Thus, Bacillus subtilis sporulation entails cell specific gene expression and the generation of morphologically and functionally differentiated cells that are hallmarks of eukaryotic differentiation.Recent research has found that sporulation includes another facet of cellular differentiation, metabolic differentiation. Shortly after polar septation, metabolic enzymes are depleted in the forespore resulting in the metabolic differentiation of the mother cell and forespore. These depleted enzymes are primarily in biosynthetic pathways such as amino acid synthesis, so metabolic differentiation requires that the mother cell nurtures the forespore by providing the metabolic precursors required for development. However, how and why metabolic enzymes are depleted from the forespore are still unknown.Here I used a combination of in vivo experiments and metabolic modeling to explore metabolic differentiation. Using genetic screens, we have discovered that metabolic enzymes are being degraded in the forespore by the protease complex ClpCP and identified a new sporulation adaptor protein for ClpC we named MdfA. Further tests on a representative MdfA-ClpCP substrate called CitZ revealed a degron sequence that is recognized by ClpC. However no direct interaction was found between MdfA and its substrates. We also used metabolic modeling to elucidate the requirements of metabolites in the forespore and mother cell and identified a potential new mechanism to produce energy in the forespore. Altogether this research has revealed a new mechanism during spore development and laid a foundation for future research
Abbreviated Analysis of California Senate Bill 242 Medicare Supplement Coverage: Open Enrollment Periods
Get PDFNovel origin of replication for environmentally isolated Pantoea strain enables expression of heterologous proteins, pathways and products
No full textPlasmids isolated or characterized from environmental samples serve as a resource that can be used to develop genetic tools for characterizing recently isolated or less-studied microbes. In this report, we leveraged sequences from a previously characterized groundwater plasmidome and developed a screen to identify novel plasmid origins. Putative origin sequences were used to construct a barcoded plasmid library, which contained both known and newly predicted origins. This library was tested against a panel of representative bacterial strains and led to the identification of 3 novel origins that putatively replicate in gram-negative bacteria not previously associated with these origin sequences. We empirically validated one of the newly identified origins, 6911, to be functional in both the model bacterial strain, Escherichia coli BW25113, as well as in Pantoea sp. MT58, a fast growing and metal tolerant, environmentally important bacterium from the widespread Pantoea genus. We confirmed that a plasmid bearing origin 6911 as the sole origin could replicate and had a copy number of 9 (± 2) in Pantoea sp. MT58. We successfully used a plasmid based on the new origin to express the reporter protein GFP, and two non-native metabolite pathways for the natural product, indigoidine and the terpenoid compound, isoprenol. By pairing functional novel origins of replication to non-model organisms this pipeline can expand the tool kit for genetic manipulations of both model and less-studied bacteria. Abstract Figure Abstract figure. We developed a host-origin pair identification pipeline by constructing a plasmid library containing both literature-sourced and computationally predicted origins of replication. By conjugating this library into non-model microbes, we identified functional origins in diverse bacteria, and could leverage these findings to develop genetically tractable hosts for microbial engineering
Take a deep breath: the important role of vitamin A in neonatal lung development and visiting the potential of aerosolized delivery for bronchopulmonary dysplasia prevention
Get PDFDistinct somatic mutation profiles in colon cancer by behavioral comorbidity
Get PDFBackgroundTobacco use, obesity, and type 2 diabetes are risk factors for colorectal cancer, but whether they generate distinct tumor mutation patterns is unclear. Tobacco is a known mutagen, while obesity and diabetes may act through metabolic and inflammatory pathways.MethodsWe analyzed colon cancer patients from the University of California Health Data Warehouse, linking clinical sequencing data to diagnosis-based indicators of tobacco dependence, obesity, and type 2 diabetes. For each gene-behavior pair, we conducted reverse logistic regressions and calculated a combined score reflecting the strength and specificity of association adjusting for demographic covariates and cancer stage. Multidimensional scaling and clustering assessed behavioral differentiation.ResultsOf 981 gene-behavior tests, 87 pairs exhibited a behavioral association at p < 0.001 in adjusted models. Of these, 60 tobacco, 12 obesity, and 9 diabetes pairs had affinity ≥0.5; 48 tobacco pairs exceeded 1.0. Mean (SD) combined scores were 1.39 (0.79) for tobacco, 1.24 (0.88) for obesity, and 0.74 (0.39) for diabetes. Exemplars included KEAP1 and CDKN2A (tobacco), ASPSCR1 and PGR (obesity), and a smaller diabetes signal led by MAF.ConclusionsTobacco dependence is associated with a more mutagenic and distinct somatic mutation profile in colon cancer, suggesting fundamental differences in behavioral mechanisms of carcinogenesis
Women’s self-help group participation and discussion of reproductive coercion: Associations with past experiences of violence among family planning clients in Kenya
Get PDFReproductive coercion (RC) and intimate partner violence (IPV) are prevalent in Kenya and undermine women's reproductive agency; community-based women's groups may offer an opportunity to address these issues. We quantitatively examined women's participation in women-only self-help groups, described discussions of RC within these groups, and assessed differences based on prior experiences of RC and IPV among a clinic-based sample of Kenyan women seeking family planning services to inform future programming. Data were collected from 659 women of reproductive age seeking family planning services at six private clinics in Nairobi as part of the baseline for an intervention to address RC and IPV within family planning counseling. We used descriptive statistics, bivariate hypothesis testing, and adjusted mixed-effect logistic regression models to examine the relationship between self-help group participation and lifetime experience of RC and IPV. Using the same methods, we analyzed group discussions about RC in relation to lifetime experiences of abuse among women who had recently participated in groups. Additionally, we explored associations between RC discussions and the type of self-help group (e.g., economic, charitable) using bivariate tests. Over half of women reported prior participation in women-only self-help groups, with most participating in economic groups (63%). More than one in two women reported prior IPV and one in three reported prior RC. Self-help group participation was not significantly associated with prior experiences of RC or IPV. However, women who had ever experienced RC were more likely to report discussions about RC within groups (AOR 1.82 95% CI 1.06-3.14). Discussions of RC were less common in economic groups. Women-only self-help group participation is common and acceptable among Kenyan women, including those experiencing violence. However, economic groups-though widespread-discuss RC less often, indicating a key opportunity to integrate RC/IPV programming into these settings to strengthen community-based support