Dartmouth Institute for Health Policy and Clinical Practice
Dartmouth Digital Commons (Dartmouth College)Not a member yet
8214 research outputs found
Sort by
Skyline Sketches: The Certain Slant of Light
Full text linkConsidering Emily Dickinson while climbing Mount Willard in autumn
Advances in Indole Chemistry: Regiodivergent Annulations and Bridged Azacycle Synthesis
Full text linkAn important aspect for the research and development of new drug therapies is the ability to use modular synthetic methods to rapidly access structurally diverse drug-like building blocks. Nitrogen-containing heterocycles represent one of the most prevalent structural motifs found in pharmaceutical agents and biologically active natural products. However, despite their ubiquity, existing methods for accessing these structures often remain limited in scope, efficiency, and structural diversity, particularly when targeting densely functionalized or sterically demanding scaffolds from simple starting materials.
To address this problem, the research described in this dissertation focuses on the design and development of (3+2) annulation reactions for the direct synthesis of diverse cyclohexa-fused indoline compounds This reaction technology enables rapid and regioselective construction of indoline cores with unique substitution patterns that would be challenging or cannot be easily accessed with existing methods. Key to this approach is the use of mechanistically informed reaction design to control regiochemical outcomes under mild conditions with a broad substrate scope.
Moreso, using previously published methodologies by our lab, these annulative products serve as highly functionalized intermediates amenable to further structural elaboration. In particular, they can be modified to incorporate bridged azacycles within the gross structure; thereby expanding the accessible chemical space and enabling the exploration of novel three-dimensional architectures. This capacity for modular diversification aligns with the broader goal of generating compound libraries with maximized structural diversity, favorable for drug discovery.
Overall, the methodologies described herein not only expand the synthetic toolbox for nitrogen heterocycle construction but also invite new opportunities for accessing molecular architectures that are both challenging to prepare and relevant to pharmaceutical research. By enabling the streamlined synthesis of complex, stereochemically rich scaffolds from simple starting materials, these strategies contribute to the efficient exploration of underrepresented regions of biologically relevant chemical space
A Comprehensive Pipeline for Autonomous Surface Vehicle Navigation in Challenging Aquatic Environments
No full textThe primary objective of this thesis is to develop and validate innovative and robust navigation methods for autonomous surface vehicles (ASVs) in challenging scenarios. These efforts aim to establish a complete autonomy pipeline for robotic decision-making systems, enabling high-level tasks such as environmental monitoring and autonomous transportation with broader impacts. The ocean economy contributes over 1.5 trillion USD annually, supporting diverse cultures and economies through tourism, fisheries, shipping, and renewable energy. The global marine industry handles over 90% of the world’s cargo transportation, underscoring its critical importance. Despite this significance, current maritime navigation relies heavily on human decision-making, which is prone to error under uncertain conditions. Recent advancements in autonomous mobility, including efforts by the International Maritime Organization (IMO) and technology companies, highlight the importance of developing autonomous systems for the marine domain. These systems promise reduced costs, improved safety, and enhanced efficiency. However, navigation safety and robustness remain major barriers to the widespread adoption of ASVs due to unstructured waterway conditions, vehicle dynamics, uncertain sensor information and intention, and ambiguous traffic regulations. Such challenges make common techniques from self-driving cars not directly applicable, motivating the need for aquatic-specific autonomy. This thesis addresses these challenges through contributions in two key areas: planning and perception. For planning, contributions include: (1) adaptive and proactive collision avoidance using risk-vector-based near-miss strategies; (2) multiple obstacle avoidance via holistic motion attribute-based clustering, theoretical analysis, and multi-objective optimization; (3) learning-augmented active collision avoidance with topological modeling of passing and intent-awareness, validated in real robot deployments; and (4) risk- and energy-aware global path planning across topologically distinct options under dynamic disturbances. For perception, contributions include: (5) an efficient LiDAR-based in-water obstacle detection framework for unknown aquatic environments that operates in real time onboard ASVs; and (6) the first multi-modal maritime dataset, publicly released to advance sensor fusion frameworks. The proposed holistic framework advances autonomy “in the wild” and serves as proof of concept using a custom ASV in real-world deployments
Beyond the Priesthood: Patriarchy, Faith, and the Feminist Movement with the LDS Church
Full text link
EXPLORING THE FACETS OF RESPONSIBLE AI: INTERPRETABILITY, BIASES, AND MORALITY OF LARGE LANGUAGE MODELS
No full textThis thesis investigates critical aspects of responsible artificial intelligence (AI) — specifically model interpretability, bias detection and mitigation, and moral alignment in large language models (LLMs) — due to their pivotal role in the deployment of transparent, fair, and ethical AI systems. By addressing these dimensions of responsible AI, we hope to foster the increased trust and understanding necessary for wider AI adoption.
We begin by surveying the existing landscape of interpretability metrics and critically assess the effectiveness of interpretability methods designed to generate reliable explanations. Building upon this evaluation, we introduce novel model architectures and frameworks explicitly developed to enhance the interpretability of LLMs. On the front of bias mitigation, we adapt established bias benchmarks to focus specifically on racial and LGBTQ+ biases within healthcare contexts. Our evaluations demonstrate substantial biases embedded in multiple LLM architectures and highlight the nuanced effects of debiasing strategies, showing minimal performance trade-offs for general NLP tasks but notable impacts in specialized biomedical applications. In exploring the moral dimension of AI, we conduct extensive experiments assessing how LLMs align with established normative ethical frameworks. These investigations reveal systematic patterns in moral reasoning and significant inconsistencies influenced by scenario framing.
Collectively, our contributions offer a cohesive approach to responsible AI, effectively integrating interpretability, bias reduction, and moral alignment strategies. The insights and practical tools provided by this thesis contribute meaningfully to the development of AI systems that are transparent, equitable, and ethically consistent, establishing a foundation for responsible AI deployment
Deep Learning for Fine-Grained Digital Histopathology Image Analysis
No full textAs digital pathology becomes increasingly popular, it is critical to develop machine learning solutions to utilize this data. While other image modalities have seen exponential increases in methodology availability, the same has not been true for histopathology images. This is likely in part because histopathology whole slide images possess unique characteristics that prevent simply applying existing methods as-is.
In this thesis, we identify and propose solutions to 3 open problems with histopathology images: 1. large raw image size (up to 150,000×150,000 pixels in size), 2. low class-positivity (low ratio of positive to negative patches), and 3. limited image availability with existing images having weak or no labels. We address the large raw image size problem by designing a knowledge distillation-based approach to reduce computational cost significantly with a modest decrease in classification performance. The computational cost reductions are substantial enough to enable real time use in clinical scenarios. For the low class-positivity issue, we develop a custom view generation approach for self-supervised representation learning. This view generation approach takes advantage of the low class-positivity to increase possible view pairings and produce better classification outcomes. Lastly, we present an image generation approach using existing image-spatial transcriptomics pairs to generate synthetic histopathology patches. We demonstrate these generated patches are clinically useful through evaluations including nuclei distribution quantification and downstream tasks
CHARACTERIZING THE PRODUCTION OF LIPID AND ISOTOPE BIOSIGNATURES BY EXTREMOPHILIC ARCHAEA IN LAB EXPERIMENTS AND TERRESTRIAL HOT SPRINGS
No full textUnderstanding how biological signals are recorded and preserved in organic molecules is central to reconstructing Earth’s history and guiding the search for life beyond our planet. Lipid biomarkers – organic molecules produced in cellular membranes – offer unique insights due to their structural diversity, taxonomic specificity, and long-term preservation potential. The hydrogen isotope values (δ²H) of lipid biomarkers track water in the growth environment with an often large offset due to biosynthetic effects. Certain lipids may retain their original H-isotopic composition for up to 108 years, enabling studies of past hydroclimate, metabolism, and ecology. The underlying controls on the δ²H composition of lipids produced by the domain Archaea remain poorly constrained. This dissertation investigates the δ²H values of archaeal isoprenoid glycerol dibiphytanyl glycerol tetraether lipids (iGDGTs), combining laboratory experiments, field observations, and planetary analog studies to evaluate their potential as hydroclimate proxies and biosignatures suitable for life detection beyond Earth. In lab experiments with the thermoacidophilic archaeon Sulfolobus acidocaldarius, I show that lipid-water H-isotope fractionation (2εL/W) is consistently large and negative across a wide range of environmental conditions, indicating that environmental water is the dominant H-source for lipids (Chapter 1). Isotope labeling experiments and mass-balance models further illuminate the role of metabolic pathways in determining lipid δ²H values (Chapter 2). A field survey of hydrothermal springs in Yellowstone National Park (USA) and El Tatio Geyser Field (Chile) reveals that archaeal lipid δ²H values are consistently depleted relative to environmental waters and provides the first environmental calibration of archaeal lipid δ²H (Chapter 3). Patterns in fractionation correlate with spring geochemistry, likely reflecting variation in community metabolism. Lipid hydrogen stable isotope probing (LH-SIP) experiments constrain lipid turnover to decadal timescales, confirming the persistence of these biosignatures in the sediments of hydrothermal springs (Chapter 4). Finally, I assess how these findings can inform life detection efforts on Mars, where hydrothermal deposits are considered high-priority targets for astrobiology exploration (Chapter 5). I highlight strategies that couple orbital-, drone-, and rover-scale methods to identify promising landing and sampling sites. Together, these results advance archaeal lipid δ²H as a geochemical and astrobiological tool and provide a framework for interpreting molecular biosignatures in terrestrial and planetary contexts