545298 research outputs found
Sort by
Modeling ring current proton distribution using MLP, CNN, LSTM, and transformer networks
Get PDFThis study aims at developing ring current proton flux models using four neural network architectures: a multilayer perceptron (MLP), a convolutional neural network (CNN), a long short-term memory (LSTM) network, and a Transformer network. All models take time sequences of geomagnetic indices as inputs. Experimental results demonstrate that the LSTM and Transformer models consistently outperform the MLP and CNN models by achieving lower mean squared errors on the test set, possibly due to their intrinsic capability to process temporal sequential input data. Unlike MLP and CNN models, which require a fixed input history length even though proton lifetime varies with altitude, the LSTM and Transformer models accommodate variable-length sequences during both training and inference. Our findings indicate that the LSTM and Transformer architectures are well suited for modeling ring current proton behavior when GPU resources are available, and the Transformer slightly underperforms the LSTM model due to the restriction on the number of total heads. For resource-constrained environments, however, the MLP model offers a practical alternative, with faster training and inference times, while maintaining competitive accuracy
Net Influx Rather Than Directional Rates: Re-evaluating Transporter Characterization InVivo and InVitro for Renal and Hepatic Clearance
Get PDFHere we challenge the assumption that hepatic uptake can be rate-limited solely by transporter-mediated influx clearance, as commonly concluded in interpretations based on the Extended Clearance Concept (ECC). We initially review the derivation of renal and hepatic clearance independent of differential equations based on adaptation of Kirchhoff’s Laws from physics, incorporating clinically relevant aspects such as transporter activity, organ blood flow, and clearance from the site of drug delivery into systemic circulation. In doing so, we highlight the limitations of the ECC framework, which does not adequately capture all aspects of these mechanistic elements and note that no experimental data or PBPK analyses definitively support its validity. In contrast, we show that all hepatic clearance data can be adequately explained based on our derivations, which provide a more robust and mechanistically consistent framework for interpreting renal and hepatic clearance. The derived equations define the net difference between influx and efflux clearances as the key determinant of transporter involvement in hepatic drug disposition, rather than considering influx alone as in the ECC. This approach is analogous to the treatment of secretion and reabsorption clearances as opposing processes in renal clearance, where their difference determines the net transport. We also question the mechanistic accuracy of determining hepatic influx and efflux clearances in vitro using initial rates of membrane passage, as is typically done for passive processes. Such measurements inherently reflect the net intramembrane difference between influx and efflux clearances, and do not allow for independent quantification of directional transport, due to the inability to measure drug concentrations within the membrane. Finally, while we acknowledge the utility of ECC and PBPK analyses for predicting changes in pharmacokinetic exposure due to DDIs (or other variables such as disease state, pharmacogenomics, etc.), we caution that model-based data fitting, however useful, does not constitute mechanistic validation.Graphical Abstrac
Physical integrity and residual bio-efficacy of PBO-pyrethroid synergist-treated and pyrethroid-only LLINs after 1.5 years of field use in Western Kenya
No full textBACKGROUND: Long-lasting insecticidal nets (LLINs) are vital for malaria control in sub-Saharan Africa, but their durability is challenged by fabric decay and pyrethroid resistance. This study assessed the physical integrity and bioefficacy of piperonyl butoxide-LLINs (PBO-LLINs) and pyrethroid-only LLINs (pyrethroid-LLINs) after 1.5 years of use in western Kenya, where resistance is widespread.
METHODS: A survey on net integrity and insecticide efficacy was conducted in randomly selected households (101-107 per group per visit) from three villages per net type group in Muhoroni Sub-County, Kisumu County. Physical integrity surveys were done after every six months while residual bio-efficacy was after every three months for 18 months. Physical integrity and residual bio-efficacy studies were conducted following WHO guidelines.
RESULTS: PBO-LLINs exhibited higher physical integrity than pyrethroid-LLINs over time. At 18 months, 45.2% (61/135) of pyrethroid-LLINs and 21.8% (31/142) of PBO-LLINs were torn, with pHI values of 2494.1 ± 1696.4 and 1618.6 ± 1056.7, respectively. Net type, net age and house wall structures significantly influenced net integrity (p < 0.05). Torn nets were significantly more common in pyrethroid-LLIN households with mud-unplastered [OR=5.323 (95% CI = 1.685-16.816), p = 0.004] and corrugated iron walls [OR=6.31 (95% CI = 2.10-18.93), p < 0.001] and in PBO-LLIN households with mud-unplastered walls [OR=9.823 (95% CI = 1.487-64.898), p = 0.018]. Against the Kisumu susceptible Anopheles gambiae s.s, both net types decreased in mortality at baseline (when new) from 97.6% to 18.4% and 98.6% to 18.5% for pyrethroid and PBO-LLINs respectively at 18 months. Against a Bungoma pyrethroid-resistant Anopheles gambiae s.s, mosquito mortality with pyrethroid-LLINs declined from 36.9% when new to 6.8% at 18 months, while PBO-LLINs dropped from 55.6% to 11.8%.
CONCLUSION: Both physical integrity and bioefficacy of LLINs declined significantly within 18 months. The findings demonstrate that not all nets in the field offer maximum protection by this time point, calling for net care education and further evaluation of PBO-LLINs especially in pyrethroid-resistant regions
Incidence and predictors of immune checkpoint inhibitor treatment–related cognitive impairment in a racial and ethnic diverse population
Get PDFIntroductionImmune checkpoint inhibitors (ICI) have revolutionized cancer therapy in recent years. In addition to rejuvenating anti-cancer immunity, ICI may cause immune dysregulation, impacting homeostasis, including brain functions. Thus, the association of ICI with cognitive function needs further investigation. Using NIH’s PROMIS system, this study investigates self-reported cognitive impairment within a diverse cohort of ICI-treated patients. Additionally, we explore risk factors influencing self-reported cognitive function, including concurrent symptoms and racial/ethnic background.MethodsThis was a prospective, longitudinal study conducted between July 2021 and June 2023. Included patients were ≥ 18 years old, newly diagnosed with cancer, and scheduled to receive ICI therapy. Serial patient-reported outcomes (PROs) were collected from self-reported patient surveys at therapy initiation and while receiving treatment. Clinically significant cognitive impairment was defined as mild to severe symptoms measured on computer adaptive tests using the PROMIS Bank v2.0 — Cognitive Function. Multi-step analysis utilizing generalized estimating equations (GEE) was implemented to evaluate characteristics associated with cognitive function T-scores.ResultsOur study included 51 ICI patients, with 51% being of Asian or Hispanic descent. Of 126 PROMIS symptom survey sets collected, 16.7% reported clinically significant cognitive impairment, with incidence peaking in survey sets collected 1–2 months removed from therapy initiation at 26.1%. All concurrent PROMIS symptom scores significantly correlated with cognitive function, including physical function (r = 0.33, p < 0.001), fatigue (r = − 0.61, p < 0.001), depression (r = − 0.56, p < 0.001), and anxiety (r = − 0.56, p < 0.001). Multi-variable regression demonstrated impaired physical function (Coef = − 4.01, p = 0.007), fatigue (Coef = − 5.15, p = 0.005), and anxiety (Coef = − 4.45, p < 0.001) are associated with decreased cognitive function scores, after adjusting for other patient characteristics.ConclusionPatients receiving ICI therapy experience significant cognitive impairment with therapy initiation and in subsequent weeks and months during their therapy course. Managing and monitoring concurrent symptoms and inflammatory biomarkers may help identify at risk patients and alleviate cognitive impairments
Advances in the microbial biosynthesis of therapeutic terpenoids
Get PDFTerpenoid natural products and their derivatives exhibit bioactivity that is utilized in FDA-approved drugs and candidates for future drugs; however, the widespread utilization of terpenoids has been limited by complex, low-yielding native biosyntheses and chemical syntheses. Microbial total/semi-biosynthesis of natural and new-to-nature terpenoids from sustainable feedstocks is scalable and can achieve economically viable cost targets. Herein, this review describes foundational advances in synthetic biology and metabolic engineering as exemplified by efforts to biosynthesize prominent terpenoids (i.e. artemisinin, taxol, vinblastine, QS-21, and cyclopamine) in engineered microorganisms (e.g. Escherichia coli and Saccharomyces cerevisiae). Emerging methods that accelerate microbial biosynthesis campaigns (i.e. automation, machine learning, artificial intelligence, and combinatorial screening) are then discussed
Targeting CHRM3 for REM Sleep Modulation: A Potential Strategy to Mood Stabilization in Bipolar Disorder
Get PDFEssays on Machine Learning in Causal Inference and Prediction
Get PDFMy research bridges the fields of machine learning (ML) and econometrics, addressing both predictive and causal inference challenges. The growing intersection of these two fields is transforming how we approach data analysis in economics and social sciences. While machine learning excels at pattern recognition and prediction tasks, econometrics provides a rigorous framework for understanding causal relationships. My work focuses on integrating these two approaches to enhance the power and interpretability of machine learning models in economics. In particular, I investigate how ML techniques can be adapted for causal inference, allowing for more robust predictions and better policy decisions. Chapter 1 outlines an overview of the research work in the dissertation.Chapter 2 provides the innovative architecture, theoretical foundations and simulation results of deep learning to handle individual heterogeneity and address endogeneity with generated regressors.Chapter 3 conducts a detailed comparative analysis of various tree-based and deep learning models, focusing on their prediction capabilities. Additionally, it explores different prediction combination techniques to evaluate whether combining predictions from multiple models enhances predictive accuracy.Chapter 4 explores the use of machine learning techniques for estimating partial derivatives, which is a critical step towards understanding causal relationships in econometric analysis. Using modern machine learning methods, such as tree-based models and deep neural networks, we assess their effectiveness in recovering regression functions and estimating partial derivatives.Chapter 5 presents a comparative study of double machine learning and balancing approaches to estimate the effects of continuous treatment in the context of large-scale tech applications. It utilizes the semi-synthetic data based on Snapchat user data to evaluate these methods’ performance in terms of scalability, flexibility, and precision in handling high-dimensional and highly non-linear causal relationships. The comparative study focuses mainly on dose-response curves and marginal effects for potential applications of studying the impact of continuous business metrics such as ad frequency and app latency on user engagement
Investigation of Therapeutic Interventions Addressing Demyelination and Axonal Injury in Mouse Models of Multiple Sclerosis
No full textVisual dysfunction is a common feature in multiple sclerosis (MS), an autoimmune inflammatory demyelinating disease of the central nervous system. The visual pathway, highly susceptible due to its dense myelination, often experiences inflammatory demyelination and progressive axonal damage, commonly presenting as optic neuritis. Optic neuritis is one of the earliest and most frequent manifestations of MS, typically preceding motor and cognitive decline. Optic neuritis in mouse models of MS can be assessed using non-invasive, functionally relevant measures, making it an ideal platform for studying the mechanisms of demyelination and axonal injury, as well as for developing targeted neuroprotective therapies.There remains an unmet need for treatments that achieve both functional remyelination and neuroprotection in MS. This dissertation tests the hypothesis that novel estrogen receptor beta (ERβ) ligands are safe and effective in promoting functional remyelination while modulating immune responses in mouse models of MS. Additionally, treatment with SARM1 inhibitors may reduce axonal damage and improve visual function. The outcomes of this research aim to advance dual-target therapies that simultaneously enhance remyelination and neuroprotection, ultimately preserving neurological function and improving quality of life for individuals with MS