71446 research outputs found
Sort by
Approximate Bayesian Inference for Network Processes
No full textIn network science and epidemiology, simulations are valuable for understanding real-world systems, informing predictions, and evaluating intervention strategies. However, for simulations to be useful, inference or calibration must be conducted on the internal parameters of the model. In many situations, this inference can be difficult, as many simulated models, even with relatively simple mechanistic rules, do not have computationally tractable likelihoods.
For inferences on such models, one popular set of methods is Approximate Bayesian computation (ABC), an approach that has been utilized for models in population genetics (Tavaré et al., 1997; Beaumont et al., 2002), physics (Akeret et al., 2015), and ecology (Toni et al., 2009). In order to compare simulation outputs and observed data, ABC methods require the specification of a set of summary statistics. In this dissertation, we seek to combine ABC methods with literature on Mixture Density Networks (MDN), which are neural networks that aim to learn a parametrized approximation to the posterior distribution of the parameters, conditioned on observed data (Bishop, 1994).
In Chapter 1, we will investigate the use of Mixture Density Network-Augmented ABC (MDN-ABC) (Hoffmann and Onnela, 2022) for inferences on epidemics where the event times (times of infection and recovery) are not observed. By learning informative summary statistics through an MDN, we show how valid Bayesian inferences can be obtained while circumventing the summary statistic selection step that most ABC methods rely on. Furthermore, we discuss the interpretability of the summary statistics obtained from MDNs.
In Chapter 2, we will continue to explore the use of MDN-ABC for epidemics on networks, but in cases where the contact network itself is also unobserved. By adopting a framework for modeling noise and missingness on networks proposed by Young et al. (2020), we find that it is possible to account for contact network uncertainty in a statistically valid manner through an additional network sampling step. In this chapter, we apply this Network-Augmented MDN-ABC (NA-MDN-ABC) to conduct inferences on Tattoo Skin Disease (TSD) spreading among dolphins in Shark Bay, Australia (Powell et al., 2019) and estimate the per-contact transmissibility and infectious period of the disease.
In Chapter 3, we will discuss Bayesian inferences for mixture-of-mechanisms models for networks. Bayesian inferences on the relative importance of network formation mechanisms remains a difficult problem, as mechanistic network models do not generally yield tractable likelihoods. Existing methods focus on utilizing network summary statistics (Ratmann et al., 2007; Raynal and Onnela, 2022), but it is not guaranteed that such statistics are informative or optimal. In this chapter, we will discuss the use of an MDN that utilizes a Graph Neural Network (GNN) to extract network information and conduct
Bayesian inferences. Approximate Bayesian inference tools provide a flexible framework with which researchers can study complex systems through simulations. By leveraging neural networks to extract relevant information from datasets, the methods we present allow for the automated learning of summary statistics, which avoids the use of ad hoc summary statistics and circumventing the summary statistic selection step.
Simulations, in principle, can be designed to emulate reality as closely as possible, given the available computational budget. However, parameter inference and estimation for such simulations often require automated, statistically principled methods for likelihood-free inferences. By developing Bayesian methods capable of accommodating a wide range of data inputs, we seek to bridge this existing gap between realistic simulations and valid statistical inferences.Biostatistic
Developing Classifiers for tRNA Nanopore Sequencing: from DNA Barcodes to Amino Acid Identities
No full textNanopore sequencing expands the new frontier of sequencing with its single-molecule resolution and ability to handle diverse types of biomolecules without compromising molecular information, such as DNA/RNA modifications. However, it has primarily been developed and validated for use with long DNA/RNA thus far, and only within the past year has directly nanopore sequencing tRNAs become experimentally feasible with sufficiently high yield and read quality. Given how recently this progress has been made, there is currently a lack of computational tools to handle tRNA nanopore sequencing data.
A multiplexing strategy allows simultaneous sequencing of multiple barcoded experimental groups, with retrieval of barcode labels using a computational demultiplexing tool. In this thesis, we develop a demultiplexing tool capable of handling both barcoded tRNAs and long RNAs. This tool achieves an AUROC of 0.99 with tRNAs and 0.95 with long RNAs in a four-barcode classification task. It is capable of classifying 78.9% of tRNAs with 99.6% accuracy, and 70.2% of long RNAs with 95.5% accuracy. The release of this demultiplexing tool would be the first open-source demultiplexing tool equipped for direct tRNA nanopore sequencing. We illustrate its usage with the incorporation of our demultiplexing tool into a streamlined pipeline for identifying modifications by comparing experimental groups.
Finally, we demonstrate a proof-of-concept for direct nanopore sequencing of aminoacylated tRNAs. We optimize the reaction conditions and choice of heterobifunctional linker in a bioconjuation strategy to capture and prepare aminoacylated tRNAs for direct nanopore sequencing. We then train a support vector machine for binary classification of a dataset containing the same type of tRNA aminoacylated with two types of amino acids. This model achieves an AUROC of 0.72 (p 0.05). This model verifies that nanopore sequencing can capture statistically significant signal surrounding the amino acid identity of an aminoacylated tRNA. This thesis is, to our knowledge, the first demonstration that direct nanopore sequencing can be used to analyze aminoacylated tRNAs.Computer Scienc
Regenerative Agriculture Adoption among Brazilian Soybean Farmers
No full textIn the current context of climate change and ecosystem degradation, regenerative agriculture is a farming practice that is
gaining attention for its climate mitigation and biodiversity loss reversal potential. However, there is still no global definition for it. Without a global definition, measurements of its adoption vary, as do assessments of the belief that farmers have in the environmental and productivity benefits of regenerative agriculture and the barriers they perceive to adopting it. Furthermore, there are limited offerings of price premiums for the products of farmers who follow such practices. This research assumes that regenerative farming practices entail lower climate and environmental risks compared to traditional farming methods. Therefore, it is important to understand levels of adoption and belief on regenerative agriculture for a transition to regenerative agriculture occur.
To assess the adoption rate of four regenerative practices for soybean farming in Brazil (no tillage, cover cropping, crop-livestock systems, and the use of bio inputs), understand the environmental and productivity beliefs and the main barrier to adoption that farmers face, 22 Brazilian soybean producers were interviewed in a semi-structured format. Respondents answered questions about their adoption rates for the four practices, their perceptions of the environmental and productivity benefits of regenerative farming, the primary barrier to adoption they observe, broader challenges they face with soybean
farming, and whether they observe any price premium for adopting regenerative practices.Extension Studie
Discovery and Characterization of Gut Bacterial Polyphenol Dehydroxylases
No full textPolyphenols found in plant-based foods are associated with various health benefits, including reduced inflammation and healthy aging. Notably, gut bacteria extensively metabolize polyphenols, altering their bioactivities and bioavailabilities. However, the bacterial enzymes responsible for polyphenol metabolism remain largely unknown. In my thesis work, we sought to advance our understanding of gut bacterial polyphenol metabolism by discovering and characterizing enzymes that dehydroxylate polyphenols, a transformation of significant chemical interest. Using an interdisciplinary approach that integrated bioinformatics, protein biochemistry, spectroscopy, structural characterization, and microbiology, we discovered molybdenum-dependent polyphenol dehydroxylases, including several catechol dehydroxylases and a xanthine dehydrogenase, that generate health-relevant gut metabolites. We further characterized catechol dehydroxylases by elucidating the active site environment, identifying key amino acid residues involved in catalysis and substrate recognition, and proposing a catalytic mechanism. Altogether, my thesis establishes a foundation for future investigations into polyphenol dehydroxylases and their impacts on human health.
Chapter 2 describes our metatranscriptomics-guided discovery of a prominent catechol dehydroxylase from the human gut microbiome. We identified a highly expressed and prevalent uncharacterized catechol dehydroxylase (Gp Hcdh) in the human gut and determined that its substrate is hydrocaffeic acid (HCA), a common metabolite derived from polyphenols found in plant-based foods like coffee and cruciferous vegetables. Gp Hcdh activity was linked to anaerobic respiration in the encoding Gordonibacter strains and correlated with host inflammation in a human cohort. Together, these findings highlight the utility of metatranscriptomics-guided discovery and suggest a potential connection between polyphenol metabolism and host inflammation.
In Chapter 3, we used Gp Hcdh as a model to investigate the molybdenum environment and catalytic mechanism of catechol dehydroxylase. X-ray absorption spectroscopy and mutagenesis identified that the Cys157 thiolate ligates the molybdenum ion. These insights informed computational modeling, which suggested a unique mechanism involving dearomatization and a 1,2-hydride shift. Additionally, these analyses highlighted a catalytic role for the active site Asp210 carboxylate as a general base, which was confirmed through mutagenesis. The proposed mechanism explains the requirement for a catechol motif in catalysis and warrants further experimental validation.
In Chapter 4, we determined the first structure of a catechol dehydroxylase (Gp Hcdh) using cryo-electron microscopy (cryo-EM), revealing the active site environment and substrate-binding funnel. Guided by this structure, mutagenesis and docking identified ten key funnel residues, including Arg371 and Arg488, which likely engage the substrate via ionic interactions. Using these insights, we discovered a previously uncharacterized enzyme (Hpbh) that dehydroxylates 4-(3-hydroxypropyl)benzene-1,2-diol, a novel polyphenol metabolite. Expanding this approach, we identified and validated HCA dehydroxylases from diverse environmental bacteria, showing this activity extends beyond Coriobacteriia in the human gut.
Lastly, Chapter 5 describes our efforts to identify gut bacterial enzymes involved in sequential dehydroxylation of ellagic acid (EA) to urolithin A, an anti-inflammatory metabolite and dietary supplement. Using differential gene expression analyses, we identified five enzymes from three gut bacterial genera: a lactone hydrolase (Eah) and three catechol dehydroxylases (Eadh1/2/3) from Gordonibacter and Ellagibacter, and a distinct molybdenum-dependent enzyme (Ucdh) within the xanthine dehydrogenase family from Enterocloster. Biochemical characterization revealed that substrate specificity of each dehydroxylase shapes the EA metabolic pathway. Notably, Ucdh, which does not require catechol, likely operates via a different catalytic mechanism from catechol dehydroxylase. Finally, genes encoding urolithin A-producing enzymes were depleted in the gut microbiomes of inflammatory bowel disease (IBD) patients, suggesting that gut bacterial metabolism of EA may differentially influence host inflammation in this context.Chemistry and Chemical Biolog
Nanoscale Investigations of Monolayer Thin Films and Heavy Element Materials
No full textDuring my Ph.D, my work were divided into two parts. First, I spent a lot of time and effort using
molecular beam epitaxy (MBE) to grow monolayer thin films that have novel quantum properties.
Second, I used the scanning tunneling microscope (STM) to study cleavable single crystals, including materials with a strong surface Rashba effect and Kondo effect.
This dissertation focuses on a subset of the work that I have done in my Ph.D. Part I is about film
growth using MBE, which includes three chapters. Chapter 1 gives a general introduction to the
characterization tools I used for film growth, including STM, X-ray photoelectron spectroscopy
(XPS), atomic force microscope (AFM), scanning transmission electron microscope (STEM), and
electron energy loss spectroscopy (EELS). Chapter 2 appears in its entirety in the manuscript:
Samantha O’Sullivan, Ruizhe Kang, Jules A. Gardener, Austin J. Akey, Christian
E. Matt, and Jennifer E. Hoffman ”Imaging Se diffusion across the FeSe/SrTiO3
interface.” Physical Review B 105, 165407 (2022)
There has been a long debate on the exact structure of the FeSe/SrTiO3 interface. Some groups
reported a clean interface between the FeSe and the SrTiO3 surface while others observed an additional Se layer. In this chapter, we provided evidence aiming to put an end to this debate. Even
though we didn’t observe an ordered Se layer between the film and the substrate, we discovered a significant amount of Se diffused across the monolayer FeSe/SrTiO3 interface using EELS. This work shines light for a possible factor that affect the high-temperature superconductivity at FeSe/SrTiO3
interface. Chapter 3 demonstrates my efforts in growing a monolayer honeycomb bismuth film (bismuthene) on hydrogen-passivated SiC substrates. In this chapter, I have demonstrated the importance of the hydrogen passivation of the SiC substrate and provided evidence of the air sensitivity of
the bismuthene film.
The second part of this thesis is about STM studies on two cleavable materials composed of
heavy elements, BiTeI and UTe3. Chapter 4 is adapted from this manuscript:
Ruizhe Kang, Jian-Feng Ge, Yang He, Zhihuai Zhu, Daniel T. Larson, Mohammed
Saghir, Jason D. Hoffman, Geetha Balakrishnan, Jennifer E. Hoffman. ”Nanoscale
variation of the Rashba energy in BiTeI.” arXiv.2402.18779
The strong spin-orbit coupling (SOC) leads to a huge Rashba effect in BiTeI. In this chapter,
we observed ring-like charging states on the iodine surface of BiTeI, which could be used as a probe
of the local electric field. We extracted the local Rashba energies by fitting the van Hove singularities observed in our scanning tunneling spectroscopy. We discovered that the Rashba energies have
nanoscale variations, which positively correlate with the local electric field probed by the charging
ring states.
Chapter 5 reports the first-ever STM measurement on UTe3, where we measured Kondo resonance. In this chapter, we demonstrate how the Kondo holes affect the local Kondo resonance. We
discovered that the Kondo holes in UTe3 will reduce the local q factors and shift the Kondo resonance energies towards the valance band. However, the hybridization factor Γ shows a very weak
correlation with the Kondo hole locations, indicating that the Kondo holes could induce some hybridization disorder. This manuscript is in preparation.Engineering and Applied Sciences - Applied Physic
Deduction-Projection Estimators for Understanding Neural Networks
No full textWe introduce Deduction-Projection Estimators: a family of methods for measuring properties of neural networks inspired by the notion of a "deductive heuristic estimator" introduced in Christiano et al. (2022). Unlike traditional techniques used in machine learning, a DPE produces its estimate by mechanistically tracking how activations are processed throughout a neural network. This allows us to understand how a model behaves over an entire input distribution without having to generalize from observed behavior on a finite number of sampled inputs.
The first half of this thesis deals with the philosophy and theory of deductive estimation: how it differs from inductive reasoning, examples of deductive estimators in mathematics and machine learning, and why we might expect concise deductive explanations to exist and be tractable to find in the first place. We also discuss how efficient deductive estimators might be used to scalably control the worst-case behaviors of AI systems.
In the second half, we empirically evaluate DPEs against traditional machine learning methods. First, we introduce the problem of low probability estimation: given a transformer language model and a formally specified input distribution, can we estimate the probability that it generates a particular output, even when this probability is too small to detect with naive sampling? We develop activation extrapolation methods based on simplified DPEs, which empirically outperform naive sampling. However, the highest-performing estimators leverage importance sampling, which can be thought of as a generalization of adversarial training.
Finally, we explore how our techniques can be used to optimize small neural network classifiers to achieve maximal accuracy on an algorithmic task. Our experimental results show that DPEs often outperform cross-entropy loss as an optimization target by providing a stronger gradient signal towards maximizing accuracy. We discuss the limitations of our empirical settings and list future lines of work.Computer Scienc
Understanding the Forces that Shape Evolution
No full textEvolution by natural selection is shaped by many forces, such as clonal interference, demographic factors, and epistasis. Understanding and inferring these factors is not only central for our ability to make predictions about populations’ evolutionary futures, but also give insight into biological processes. In this thesis, I describe work on two factors that shape evolution. First, in Chapter 1, I describe a new statistic for distinguishing Kingman coalescent processes from the more complicated multiple-
merger coalescents. In Chapter 2, I describe an experiment to uncover pleiotropic patterns in the genotype-to-phenotype map in budding yeast. Finally, in Chapter 3, I discuss the technical details of genotype-phenotype mapping and a biological interpretation of the algorithm.Physic
The Association Between Metabolic Syndrome and Oral Diseases Among US Adults
No full textThe objective was to assess the association between oral diseases and metabolic syndrome (MetS), examining both the individual and combined effects of MetS five components. Also, to validate the NHANES diagnostic criteria and Overjet artificial intelligent (AI) platform with the gold standard for periodontal disease diagnosis.
First, a systematic review and meta-analysis were conducted by searching PubMed, Embase, and Web of Science for studies published between 1990 and 2023 that examined the association between Dental caries and MetS in adults. Two independent authors selected and analyzed articles, assessed risk of bias, and evaluated the overall evidence certainty. Meta-analyses were performed to estimate pooled odds ratios (ORs), or mean differences (MDs), and corresponding 95% confidence intervals (CIs) for decayed teeth (DT) and DMFT (Decayed, Missing, and Filled Teeth).
Second, data from NHANES 2011-2018 were used to assess the association between MetS and oral diseases (dental caries and periodontal disease). The sample included adults over 29 years who completed laboratory and clinical assessments for MetS and oral diseases. Logistic regression models (OR and 95% CI) were used to assess associations between MetS and both untreated caries and periodontal disease. Negative binomial regression (mean ratio [MR] and 95% CI) was conducted to examine the association between MetS and DMFT score.
Third, clinical and radiographic records of patients aged over 29 years were utilized to validate the NHANES diagnostic criteria (based on clinical measures alone) and the Overjet AI software (using full-mouth radiographs) for detecting periodontal disease. Sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were calculated to assess the accuracy of both clinical measures and the Overjet AI software, using a gold standard defined as a combination of clinical and radiographic analysis conducted by a periodontist.
The meta-analysis (nine studies with 59,075 participants) revealed that there was a positive statistically insignificant association between DT and MetS (OR: 1.17, 95% CI: 0.87–1.58; MD: 0.22, 95% CI: -0.08–0.52). However, a significant positive association was found between MetS and DMFT (OR: 1.28, 95% CI: 1.08–1.51). Results from NHANES showed that participants with MetS were more likely to have untreated caries and high DMFT mean score by 34% and 10%, respectively. While there was no association between MetS and periodontal disease. Low HDL was the most significant MetS component associated with dental caries, while insulin resistance was the strongest component linked to periodontal disease. Having all five MetS components was associated with a 17% higher likelihood of a high DMFT score. Results from the validity study showed that detecting periodontal disease using clinical measures achieved 96% sensitivity and 99% specificity. While Overjet AI software achieved 100% sensitivity and 89% specificity.
In conclusion, the study revealed a positive association between MetS and dental caries, while there was no association between MetS and periodontal disease. Future prospective cohort studies could provide a better understanding of these associations. The validity study demonstrated that both clinical measures and Overjet AI software achieved high accuracy in detecting periodontal disease.Dental Public Healt
Negotiating Sovereignties: Competing Governance and Planning Logics in the Indigenous Municipality of Hueyapan, Mexico
No full textIn this thesis, I investigate governance and planning in the Indigenous municipality of Hueyapan, Morelos, Mexico. As an Indigenous municipality, Hueyapan is at once an Indigenous community that self-governs based on customary practice, as well as a municipality that is subordinated to state and federal law. I find that these conflicting aspirations create dynamic negotiations of sovereignties, resulting in a compromise between customary practice and state law. I investigate how this conflict manifests in relation to the Municipal Development Plan required of all municipalities, which, in the case of Hueyapan, must be adapted to customary practice. I emphasize the importance of well-established Indigenous governance to effectively assert Indigenous customs in state-led planning. I conclude by speculating on future directions for sovereignty negotiations at larger ecological scales, beyond municipal boundaries.Department of Urban Planning and Desig
By Lawful Ways and Means: The New-York Manumission Society’s Efforts in the Early Republic
No full textThis thesis examines the activities and impact of the New-York Manumission Society from its founding in 1785 through the passage of New York State's gradual manumission act in 1799. The emphasis on this period, as opposed to the entire life of the Society through the mid-1800s, is that it provides an analysis of the Society’s multi-pronged strategy to achieve the main goal of the organization, which was an act of gradual manumission for New York State.
The Society worked through four main pillars of activity, which were preventing the kidnapping of free Black people, establishing and operating the African Free School, correspondence and communication on both a national and international level, and influencing state legislation. Led by noteworthy New York citizens and using a decentralized structure composed of members with numerous professional backgrounds, the Society was able to show their genuine concern for the plight of slaves and those illegally held in bondage in the city. The Society’s methodical activity towards weakening the institution of slavery helped define the standard of a responsible citizen in the new republic with respect to ending an abhorrent institution.
This thesis presents, through an analysis of records related to the Society and its members, that its measured and incremental approach to promoting manumission, while imperfect and one not always welcomed by Black or white New Yorkers, was ultimately successful in contributing towards establishing the conditions necessary for the state’s 1799 gradual manumission act.Extension Studie