95038 research outputs found
Sort by
Efficient Multi-Modal Learning under Resource Constraints
Get PDFLeveraging information from multiple domains is crucial for applications such as health care and robotics, where integrating diverse data sources enables more accurate and robust decision making. However, in many real-world scenarios, access to all data types is limited due to incomplete inputs or constrained resources. This presents significant challenges for existing approaches, which typically assume the full availability of all modalities during both training and inference. To address these limitations, this thesis develops strategies for learning under missing or constrained multi-modal conditions. Specifically, we explore two complementary directions: 1) retrieval-based methods for bridging missing information across domains, and 2) adaptive learning for multi-modal analysis under resource constraints. For retrieval-based methods, we leverage a large-scale database to collect relevant cross-modal samples that serve as surrogates for the missing modality, thereby avoiding the need for potentially inaccurate synthesis. This strategy proves effective across various downstream tasks, providing strong approximations of the absent data. However, these methods can still struggle with highly heterogeneous modalities, such as between tabular and image data. To address this, we develop adaptive learning approaches that allow decision-making under limited resources. Given partially observed multi-modal data, our method can adapt to various patterns of missingness. Furthermore, when data is yet to be collected, we allow adaptive modality selection to balance between prediction accuracy and acquisition cost.Doctor of Philosoph
ESSAYS ON DEVELOPMENT ECONOMICS: SPECIAL TAX REGIMES IN INFORMAL LABOR MARKETS AND SOCIAL RETURNS TO INFRASTRUCTURE INVESTMENTS
Get PDFThis dissertation contains two independent chapters on major constraints in developing economies: pervasive labor‑market informality and underinvestment in infrastructure. Chapter 1 — Special Tax Regimes in Informal Labor Markets. Using a two‑sector equilibrium search model with wage posting, estimated for low‑educated men in São Paulo, the chapter compares two size‑dependent policy instruments: stronger, progressive enforcement against informality and a smooth, revenue‑based special tax regime (STR) for small formal firms that phases out with size. In the model, stronger enforcement shifts activity toward formality and raises wages, output, and welfare with little change in unemployment. The STR’s effects are sensitive to design and market equilibrium: with fixed offer arrivals it fragments the formal sector and lowers GDP and welfare; when arrivals respond to market tightness, moderate relief yields small gains mainly via improved match quality, while deeper relief erodes revenue and dissipates gains. Chapter 2 — The Social Rate of Return on Road Infrastructure Investments. Motivated by the fact that roughly one billion people live more than two kilometers from an all‑season road, this chapter estimates country‑specific social rates of return to an additional kilometer of two‑lane highway across 55 emerging-market and developing economies (EMDEs) by combining macro production‑function estimates of the marginal product of public capital with hedonic, country‑specific construction costs from the World Bank’s ROCKS database. Returns are high but heterogeneous, median 55% (mean 97%), and decline with income; relative to a 7% U.S. benchmark for the social return on private capital, the median excess is about eightfold (roughly five times the pre‑liberalization equity premium that created Emerging Markets equity funds). A Dual‑Hurdle map summarizes where domestic and foreign efficiency conditions are met, and macro estimates are read alongside micro and financial evidence.Doctor of Philosoph
Discovery and characterization of novel chikungunya virus nsP2 inhibitors
Get PDFInfectious diseases, including those caused by viral pathogens, remain a major cause of global morbidity and mortality. Chikungunya virus (CHIKV) is an emerging arbovirus responsible for explosive outbreaks of acute febrile illness that often gives rise to chronic, debilitating arthralgia. Despite its significant public health impact, there are currently no FDA-approved antivirals for CHIKV or other alphavirus infections. The viral non-structural protein 2 (nsP2) protease is an attractive target for therapeutics due to its essential role in viral replication and pathogenesis.Here, we report the development and optimization of a high-throughput screening platform for CHIKV nsP2 protease using an internally quenched fluorescent peptide substrate. Following assay optimization and validation, screening of diversity-oriented chemical libraries (>100,000 compounds) and a cysteine-reactive covalent fragment library identified the vinyl sulfone inhibitor RA-0002034 as a potent inhibitor of nsP2 protease enzymatic activity and alphavirus replication. Multipronged characterization efforts demonstrated selectivity of the inhibitor for the active site cysteine, as well as minimal off-target inhibition of human cysteine proteases. Subsequent medicinal chemistry approaches established the structure-activity relationship of the scaffold and yielded analogs with improved potency, covalent reactivity, and pharmacokinetic profiles. Additional discovery efforts utilizing display technologies and targeted protein degradation strategies, as well as efforts targeting the nsP2 helicase domain with small molecule inhibitors, are also presented, providing complementary approaches for the development of nsP2-targeted antivirals.Doctor of Philosoph
BIOCHAR BASED FILTRATION SYSTEM FOR SEPTIC TANK EFFLUENT: PERFORMANCE EVALUATION AND NUMERICAL MODELLING
Get PDFSeptic tank effluent (STE) contains elevated nutrients, organic matter, and pathogens, posing environmental and public health risks. In this study, STE was treated using oak woodchip biochar in fixed-bed filtration systems. A 4 × 2 factorial design assessed effects of particle size (fine, medium, coarse, mixed) and flow regime (continuous, intermittent) on removal of COD, TOC, TN, NH₄⁺-N, and K⁺. Biochar produced at 700°C exhibited a highly aromatic, alkaline, porous structure with oxygen functional groups, confirmed by SEM–EDS and XPS, with carbon increasing from 49% to 91%. Particle size significantly influenced removal (p < 0.001), while flow regime effects were minimal. Fine biochar achieved the highest reductions (10.07–76.84%). Random Forest models accurately predicted removal efficiencies (R² > 0.97 for COD, TOC; 0.95–0.96 for TN; 0.86–0.89 for NH₄⁺-N; 0.65–0.82 for K⁺). These results support fine and mixed biochar as promising polishing media for decentralized STE management.Master of Scienc
The Effect of Prolonged Sitting and Sitting Interruption on Cerebral Hemodynamics
Get PDFIntroduction: Sedentary behavior (SB) is associated with an increased risk of Alzheimer’s Disease and Related Dementias. Increased cerebral arterial stiffening (AS) during acute bouts of SB could be a factor explaining this association. However, there is limited understanding of how acute SB affects cerebral AS, and whether interrupting SB can reduce cerebral AS. Studies & Methods: Data from 22 sedentary (>8 hours sitting daily), physically inactive (<150 min/week moderate-vigorous physical activity), middle aged (30-60 years old) adults was analyzed from an ongoing trial. Participants underwent four, 4-hour sitting conditions: 1) uninterrupted sitting, 2) once/hour 15-minute standing breaks, 3) once/hour 5-minute walking breaks, 4) both 15-minute standing and 3 of sitting with scheduled interruption (standing breaks, walking breaks, and both hourly). AS data measured using heart-to-middle cerebral pulse wave velocity (hmPWV) from these participants was used for Study 1 (reliability of hmPWV) and Study 2 (using hmPWV to evaluate changes during sitting) of the proposed dissertation. For Study 3, trial participants and a supplemental participant pool were recruited to conduct focus groups related to SB and SB interruption. Results: The Aim 1 study found the reliability of hmPWV measured using electrocardiogram (ECG) and impedance cardiogram (ICG) to be moderate (ICC = 0.632 with ECG, ICC = 0.523 with ICG) but both were below our acceptability criteria (ICC = 0.700). The Aim 2 study found a large effect of time for hmPWV (b = 0.979 m/s, d = 2.56, p < 0.001) irrespective of condition. Lastly, insights from our focus groups indicated that sedentary but physically active adults likely have lower barriers to sedentary behavior interruption than physically inactive, sedentary adults. Conclusions: While promising, additional work should be performed to standardize hmPWV measures to improve reliability. More intense and frequent SB interruption strategies can be trialed for effectiveness. Translating SB interruption strategies trialed in samples of sedentary, physically inactive adults can likely be easily translated to sedentary but physically active adults due to lower perceived barriers to behavior change in this realm.Doctor of Philosoph
A neutralizing human antibody induces movement of the HCoV-229E receptor binding domain
Get PDFHCoV-229E is an endemic Alphacoronavirus that typically causes common cold-like disease in most healthy adults, but can also cause severe respiratory disease in the very young and the elderly. Although the virus was discovered over sixty years ago and undergoes continuous antigenic drift, remarkably little is known about the humoral immune response to HCoV-229E infection. Here we report the isolation of two receptor binding domain-targeting neutralizing human antibodies raised in response to natural HCoV-229E infection. One of these, DH1533, potently neutralizes HCoV-229E, binds to spike with sub-nanomolar affinity and prevents the association between the RBD and the host cell receptor aminopeptidase N. Structural characterization of this antibody bound to HCoV-229E spike delineated a neutralization-sensitive epitope on the RBD and revealed that DH1533 induces conformational flexibility in neighboring RBDs, reminiscent of the “up-and-down” kinetics observed in the related Betacoronavirus spikes. These findings provide insight into the humoral immune response to HCoV-229E infection and will serve as a guide for the design of future therapeutic interventions
Development and Validation of a Simple-to-Use Nomogram of In-Hospital Heart Failure in Patients with Acute Myocardial Infarction
Get PDFBackground: Patients with acute myocardial infarction (AMI) who experience in-hospital heart failure (HF) would present a higher risk for fatal events. This study aims to develop and validate a simple-to-use diagnostic nomogram to identify high-risk individuals for in-hospital HF in patients with AMI. Methods: Using data from CCC-ACS (Improving Care for Cardiovascular Disease in China-Acute Coronary Syndrome) project (2014–2019), this study included 74,697 patients with ST elevation myocardial infarction (STEMI) or non-STEMI (NSTEMI) who admitted within 24 h after symptom onset, without HF, cardiac arrest, or cardiac shock at admission. Independent predictors were identified through univariate logistic regression analyses and least absolute shrinkage and selection operator (LASSO) regression. A nomogram was subsequently constructed based on multivariate logistic regression. The model’s performance was evaluated by its discrimination and calibration, assessed using Harrell’s C-index and calibration curves with Hosmer–Lemeshow goodness-of-fit tests, respectively. Results: Six predictors were selected for the final nomogram, including age, heart rate, history of atrial fibrillation, history of chronic obstructive pulmonary disease, history of chronic HF, and history of chronic kidney disease. The nomogram demonstrated a C-index of 0.68 (95% CI: 0.66–0.69) in the training cohort and 0.67 (95% CI: 0.66–0.69) in the validation cohort. The calibration curves of the nomogram showed a strong calibration, as Hosmer–Lemeshow goodness-of-fit tests yielded chi-squares of 11.00 (p = 0.21) and 8.48 (p = 0.39) for the training and validation cohort, respectively. Conclusions: This simple-to-use nomogram for effectively predicting the risk for in-hospital HF may be used as a helpful tool in clinical decision-making during treatment and management in patients with AMI
AI for evidence-based treatment recommendation in oncology: a blinded evaluation of large language models and agentic workflows
Get PDFBackground: Evidence-based medicine is crucial for clinical decision-making, yet studies suggest that a significant proportion of treatment decisions do not fully incorporate the latest evidence. Large Language Models (LLMs) show promise in bridging this gap, but their reliability for medical recommendations remains uncertain. Methods: We conducted an evaluation study comparing five LLMs’ recommendations across 50 clinical scenarios related to multiple myeloma diagnosis, staging, treatment, and management, using a unified evidence cutoff of June 2024. The evaluation included three general-purpose LLMs (OpenAI o1-preview, Claude 3.5 Sonnet, Gemini 1.5 Pro), one retrieval-augmented generation (RAG) system (Myelo), and one agentic workflow-based system (HopeAI). General-purpose LLMs generated responses based solely on their internal knowledge, while the RAG system enhanced these capabilities by incorporating external knowledge retrieval. The agentic workflow system extended the RAG approach by implementing multi-step reasoning and coordinating with multiple tools and external systems for complex task execution. Three independent hematologist-oncologists evaluated the LLM-generated responses using standardized scoring criteria developed specifically for this study. Performance assessment encompassed five dimensions: accuracy, relevance, comprehensiveness, hallucination rate, and clinical use readiness. Results: HopeAI demonstrated superior performance across accuracy (82.0%), relevance (85.3%), and comprehensiveness (74.0%), compared to OpenAI o1-preview (64.7, 57.3, 36.0%), Claude 3.5 Sonnet (50.0, 51.3, 29.3%), Gemini 1.5 Pro (48.0, 46.0, 30.0%), and Myelo (58.7, 56, 32.7%). Hallucination rates were consistently low across all systems: HopeAI (5.3%), OpenAI o1-preview (3.3%), Claude 3.5 Sonnet (10.0%), Gemini 1.5 Pro (8.0%), and Myelo (5.3%). Clinical use readiness scores were relatively low for all systems: HopeAI (25.3%), OpenAI o1-preview (6.0%), Claude 3.5 Sonnet (2.7%), Gemini 1.5 Pro (4.0%), and Myelo (4.0%). Conclusion: This study demonstrates that while current LLMs show promise in medical decision support, their recommendations require careful clinical supervision to ensure patient safety and optimal care. Further research is needed to improve their clinical use readiness before integration into oncology workflows. These findings provide valuable insights into the capabilities and limitations of LLMs in oncology, guiding future research and development efforts toward integrating AI into clinical workflows
Bridging water, health and climate: A call to action
Get PDF2024 was the warmest year on record. Projections indicate that increasing temperatures will have a profound impact on the hydrological cycle with cascading effects on ecosystems, economies, and societies. For people experiencing these changes, the combined effects can result in multiple and varied societal health risks, particularly for the most vulnerable. These risks can only be effectively managed through a holistic approach that directly addresses the interconnected dynamics of water, health, and climate
Myeloid Cell Adhesion Molecule 1 (CADM1) Promotes Proinflammatory Signaling in Human Inflammatory Bowel Diseases.
Get PDFCytotoxic T cells have been postulated to facilitate the destruction of intestinal epithelium in inflammatory bowel diseases (IBDs). CADM1, which encodes a membrane adhesion protein that can bind the T cell receptor CRTAM, was markedly up regulated in colons of IBD patients compared with non-IBD (NIBD) patients.We performed comprehensive small RNA and RNA profiling on colon tissue from IBD and NIBD control patients in addition to characterizing the serum cleaved ectodomain of CADM1 (sCADM1) function in lamina propria mononuclear cells isolated from these patients. Last, a conditional loss-of-function mouse was developed to assess Cadm1 function in the myeloid compartment during chemical-induced colitis.We identified CADM1 enrichment in multiple immune cell clusters including macrophages and dendritic cells in the colons of IBD patients. Increased numbers of CADM1+ myeloid cells were measured adjacent to CD8+ T cells within colons of ulcerative colitis patients compared with NIBD patients. Conditional deletion of Cadm1 in myeloid cells resulted in reduced numbers of activated T cell populations and protected mice from chemical-induced colitis. Similarly, administration of a Cadm1 "neutralizing" antibody, which binds its extracellular domain reduced tissue inflammation and breakdown of the intestinal epithelium and crypts after induction of colitis in mice. Last, serum levels of sCADM1 were elevated in IBD patients compared with NIBD control subjects and treatment of lamina propria mononuclear cells with recombinant sCADM1 enhanced inflammatory STAT3 phosphorylation.CADM1 is a mediator of proinflammatory signaling cascades in the colon and a potential therapeutic target for the IBDs