Bioculture Journal
Not a member yet
1309 research outputs found
Sort by
Where and When Does Streamflow Regulation Significantly Affect Climate Change Outcomes in the Columbia River Basin?
No full textThesis (Master's)--University of Washington, 2021The Columbia River basin is a large transboundary basin located in the Pacific Northwest, straddling the US-Canadian border. The basin spans seven US states and one Canadian province, encompassing a diverse range of hydroclimates. Strong seasonality and complex topography, combined with the prominent role of snow in the hydrologic cycle, give rise to spatially heterogeneous climate impacts on unregulated streamflow. The basin’s water resources are economically critical for the region, and regulation across the domain is extensive. This study investigates where and when regulation significantly affects projected changes in streamflow due to climate change by comparing climate outcomes across 80-member ensembles of unregulated and regulated streamflow projections at 75 sites across the Columbia River basin. Unregulated daily streamflow projections are taken from an existing dataset of climate projections. Regulated streamflow projections were modeled by the US Army Corps of Engineers and the US Bureau of Reclamation by hydro-regulation models that simulate system operations based on current and historical water demands. Regulation dampens large shifts in winter and summer streamflow volumes and cool-season high flow extremes. Results for changes in warm-season high flow extremes and dry-season low flow extremes are spatially variable. At historically snow dominant headwater reservoirs, regulation amplifies the change in warm-season high-flow extremes, but these effects generally diminish downstream where, in some cases, dampening effects occur. Regulation dampens dry-season low flow changes in headwater tributaries where regulation is large, but elsewhere regulation has little effect on changes in dry-season low flows
Associate-Degree-Plan scheduling and Recommendation system for Virtual Academic Advisor system
No full textThesis (Master's)--University of Washington, 2021Community college students come from diverse backgrounds and experience levels. They begin their education path pursuing a degree in a major of their choice. Most students aim to get transferred to certain universities, an academic path that demands to fulfill specific requirements, which makes students eligible for the transfer. Academic advisors at community colleges help students in creating academic plans trying their best to incorporate students’ interests, life constraints, and background. Being a heavily manual process that demands experience and familiarity with the process, there is a clear need to automate this process.The Virtual Academic Advisor (VAA) system aims to address the problem of automating academic plan creation for community colleges. The VAA is a research project paired with the development of an interactive software system that supports creating and displaying academic plans based on the needs and preferences of students. Work previously done by various students, focused on automated recommendation of core courses for targeted majors. However, no research or development has been done to incorporate selection of elective-course choices when generating an academic plan, nor a clear strategy on how to integrate elective-recommendation with the VAA system has been outlined. Incorporating electives opens up a whole new research aspect of automated scheduling. Furthermore, elective-course selection is crucial for scheduling associate degrees plans. Associate degrees are offered by community colleges and students can earn such a degree before/without getting transferred to a university.
In this thesis, we incorporate the logic and functionality of scheduling elective courses along with the core courses to generate associate degree schedules for the intended major and university of the student. We gather and collect the necessary data for the elective courses and test our scheduler for the associate degree schedules. This project also addresses the research and implementation necessary to generate alternative-schedule recommendations and its integration with the VAA system using APIs. This will assist students in exploring alternate academic paths
High Linearity Full Duplex System Implemented with Novel Impedance Matching Network
No full textThesis (Master's)--University of Washington, 2021This thesis provides a design of a full-duplex (FD) communication system that hurdles in the form of the linearity and bandwidth (BW) by utilizing a feedforward canceler for self-interference (SI) cancellation purpose and a tunable impedance matching network to widen the operational frequency. This work is implemented with the TSMC 45nm CMOS technology. With the aid of a highly-linear impedance matching network, the IIP3 can reach +60dBm with a VSWR of 1.5:1. Additionally, the overall cancellation depth from TX to RX is larger than 60dB
Computational Discovery of Novel Secondary Structures from Non-Canonical Amino Acids
No full textThesis (Ph.D.)--University of Washington, 2021Protein secondary structures are a fundamental component of biological macromolecules, which are responsible for the myriad molecular processes of life. However, these biological protein macromolecules are not evolved to be amenable for rational molecular engineering due to the conformational polymorphism of alpha amino acid polymers. Therefore, developing new protein secondary structures that exhibit exceptional stability and consequently programmability could accelerate the efforts to engineer and, more importantly, utilize designed macromolecules. In this thesis, I describe a computational method to identify new secondary structures which are composed of non-canonical amino acids. Further I describe the subsequent experimental validation of these secondary structures to expand the known secondary structures for future molecular engineering efforts. Here I considered a pool of 135 non-canonical amino acid building blocks to create combinations of di-peptide repeat units, because this theoretical space has only been sparsely investigated previously. In total, the combinations of these building blocks yielded over 15,000 unique sequences which were computationally evaluated for their propensity to form a stable helical structure. Due to the lack of experimental data on the conformational properties of these residues, I developed an exhaustive and adaptive resolution computational search method to efficiently sample the enormous space of potential conformations. This method enabled the computational evaluation of the entire molecular conformational ensemble. Using this method, I identified 10 novel secondary structures which were expected to occupy a single low energy state. Experimental evidence suggests that these molecules are well-folded, engineerable helical polypeptides. Moreover, a select secondary structure was characterized as a polymer to explore the potential for these molecules as new classes of helical polymers for future materials applications
Effects of HAART and Time on the Beta Diversity of Breast Milk Microbiome in HIV-infected Postpartum Women
No full textThesis (Master's)--University of Washington, 2021Introduction - Highly active antiretroviral therapy (HAART) has been used in HIV-infected pregnant women to suppress viral replication and reduce perinatal transmission. However, the influence of HAART on the breast milk microbiome remains largely unknown. In addition, analysis of unbalanced longitudinal studies of β-diversity data is limited by a lack of appropriate statistical methods. The purpose of this study is to investigate the effects of HAART and time on β-diversity of breast milk microbiome from HIV-infected women during the first month postpartum using a novel statistical analysis method. Methods - HIV-infected pregnant women in Nairobi, Kenya in two separate studies were randomized to receive either HAART (treated group) during pregnancy to 6 months postpartum or short course zidovudine (control group) up to delivery. Breast milk samples were collected from 25 treated and 24 control women every week during the first month postpartum. These samples were subjected to 16S ribosomal sequencing. Microbial community analysis (PERMANOVA with restricted permutation for β-diversity) was used to determine the effects of HAART and time on the breast milk microbiome. Results - PERMANOVA analysis revealed statistically significant changes in breast milk microbiome β-diversity when comparing postpartum week 1 to week 4 (p < 0.01). In contrast, no obvious difference was detected between the treated and control groups. Conclusion - PERMANOVA analysis with restricted permutation was used to evaluate the effects of time and treatment in mixed models on an unbalanced longitudinal dataset. During the first month postpartum, the β-diversity of the breast milk microbiome changed significantly in HIV-infected women from both arms of the trials. In contrast, HAART treatment had minimal effects on the β-diversity
Provable and Control-Theoretic Methods for Deep Object Pose Estimation
No full textThesis (Ph.D.)--University of Washington, 2021In this dissertation, we consider the task of object pose estimation using deep neural networks. We draw our motivation from the fact that neural networks have shown to be successful at the task of pose estimation, but are poorly theoretically understood and lack meaningful performance guarantees. As a result, our aim in this dissertation is to analyze pose estimation neural networks by developing provable performance guarantees, as well as connecting pose estimation to control theory. We take four different approaches in our analysis. First, we consider object pose estimation from the standpoint of observability in control theory, using the observability Gramian as our main tool for analysis. Next, we explore the idea of estimating the pose of a dynamic object by applying an unscented filter to pose estimates from a neural network. Next, we derive analytical bounds on the local Lipschitz constants of neural networks with ReLU activations. Finally, we consider the task of developing sensitivity bounds for pose estimation neural networks, and construct a pose estimation network with provable bounds for both the rotation and position estimates
Temporal Variability in Marine Ecosystems with Implications for Biological Monitoring
No full textThesis (Ph.D.)--University of Washington, 2021Marine environments are changing, and further changes are expected in response to climate change, industry development (e.g. oil and gas explorations and marine renewable energy), pollution, and overfishing. There is an urgent need to understand effects of these stressors on marine ecosystems and to adopt effective management measures that minimize detrimental effects. Accomplishing this goal requires a comprehensive understanding of “natural” temporal patterns of biological components and underlying processes. High-latitude environments and marine renewable energy development sites have been particularly understudied due to sampling challenges (e.g. presence of sea ice, and high currents). This lack of baseline information required to measure biological responses to environmental change has increased the difficulty to document impacts in these areas and to predict effects of further changes in the ecosystems. Chapter 1 reviews temporal variability in marine ecosystems. Chapters 2 and 3 evaluate high resolution, stationary acoustic data from the Chukchi Ecosystem Observatory (CEO) and concurrent measurements from a large set of environmental sensors to characterize temporal variability in the abundance and behavior of fish and zooplankton in the Chukchi Sea. Chapter 4 quantifies the spatial area that is represented by acoustic point source measurements to define the spatial scope of CEO observations and to inform cost-effective monitoring design at high latitudes. Chapter 5 compares temporal variability in biological characteristics at sites selected for wave and tidal energy industry development to assess the potential for applying standard methods and analytic tools for biological monitoring. Chapter 6 provides a synthesis of results and implications for biological monitoring. This comprehensive characterization of fish and zooplankton dynamics in the Chukchi Sea and at sites selected for marine renewable energy development increases our ability to detect and predict biological responses to environmental change, ensure the collection of representative samples, and assist in the design of standard strategies for biological monitoring at a range of aquatic ecosystems
MORPHOLOGICAL AND MOLECULAR DIFFERENTIATION OF ULVA SPP. (ULVOPHYCEAE, CHLOROPHYTA) AND FUCUS SPP. (PHAEOPHYCEAE, OCHROPHYTA) OF THE SAN JUAN ISLANDS, WA, US
No full textMarine macroalgae are foundation species that play a critical ecological role in coastal
communities as primary producers in the ecosystem. Both Ulva and Fucus genera are vital in
intertidal communities serving a food source and shelter for other organisms. Previous studies were
limited, focusing only on morphological characteristics of these algal genera. This project aimed
to identify the diversity of Ulva and Fucus species using an integrated approach of morphological
and molecular analysis in the San Juan Islands, WA, to better understand defining characteristics
of species and overall biodiversity. Ulva (Ulvophyceae) and Fucus (Phaeophyceae) specimens
were collected from the lower, mid, and upper intertidal zones; each representative having different
macroscopic morphological characteristics and collected in varying tidal zones. The tufA and COI-
5P loci were amplified for Ulva and Fucus specimens, respectively, then sequenced. Our study
indicates that morphological assessment of these genera alone is not definitive. Molecular-based
classification of proper Ulva and Fucus species identification are important to understand the
biodiversity within coastal ecosystems
Robust Submodular Partitioning and Linear Models of Deep ReLU Networks
No full textThesis (Ph.D.)--University of Washington, 2021Machine learning models, especially deep neural networks, have achieved great success in numerous real-world tasks. As we achieve better performance with larger models, one major challenge emerges that the costs of training machine learning systems become expensive and even prohibitive. Also, the deep learning model works as a block box in many applications with little interpretation of its behaviors. In this dissertation, we investigate two problems: 1) partitioning of training data into diverse and representative blocks for gradient computation to get improved efficiency and performance for machine learning models and 2) decomposition of ReLU deep neural networks as a collection of linear models for data points and we utilize the linear models to better understand and improve the network performance. For the {\bf{first part}} of the thesis, we first investigate the problem of partitioning the training dataset into multiple blocks which are equally diverse. The theoretical abstraction of the problem is denoted as robust submodular partitioning. In robust submodular partitioning, we aim to allocate a set of items into blocks, so that the evaluation of the minimum block according to a submodular function is maximized. Robust submodular partitioning promotes the diversity of every block in the partition. It has many applications in training machine learning models, e.g., partitioning data into blocks for distributed training so that the gradients computed for every block are consistent. We study the robust submodular partitioning problem and give an efficient Min-Block Greedy algorithm with a guarantee. We further study an extension of the robust submodular partition problem with an additional constraint (e.g., cardinality, multiple matroids, or knapsack) on every block. For example, when partitioning data for distributed training, we can add a constraint that the number of samples of each class is the same in each partition block, making the partitioned data balanced. We present two classes of algorithms, i.e., Min-Block Greedy based algorithms ( bound), and Round-Robin Greedy based algorithms (constant bound) and show that under various constraints, they still have good approximation bounds. We further investigate the robust submodular partitioning problem under cardinality constraint and apply it to generate high-quality mini-batches for stochastic gradient methods. With computational hardware (e.g., GPUs) getting dramatically faster over time, sampling a mini-batch of data points uniformly at random becomes less practical, as randomly accessing data points from disk can be slow, leading to a bottleneck for modern machine learning systems. In practice, datasets are typically written to disk according to an arbitrarily generated sequence of indices. This makes sequential access of this chosen order possible with low overhead compared to random access. On the other hand, there is a chance that the sequence is poor for training, and since it is fixed over multiple iterations of training, performance can suffer. We prove better bounds of the Min-Block Greedy algorithm for this case and greatly reduce the memory/computation costs by applying hierarchical partitioning. We compare our deterministically generated mini-batch sequences to randomly generated sequences and show that the deterministic sequences significantly beat the mean and worst performance of random sequences, and often outperform the best of the random sequences. For the {\bf{second part}} of the thesis, we focus on understanding and improving the ReLU deep network through its decomposition as a linear model for every data point. A ReLU deep network (or more generally for deep networks with piecewise linear activation functions) is essentially a piecewise linear model. Therefore, the model is locally linear around every data point, and the linear model weights are equal to the gradient of the network output with respect to its input data point. Based on this observation, we first introduce the Extended Data Jacobian Matrix (EDJM) as an architecture-independent tool to analyze neural networks at the manifold of interest. For ReLU networks, the EDJM is essentially a collection of linear models for all data points, represented as a matrix. The spectrum of the EDJM is found to be highly correlated with the complexity of the learned functions. After studying the effect of dropout, ensembles, and model distillation using EDJM, we propose a novel spectral regularization method that improves network performance.However, we note that such a regularization method has greatly increased computational costs, limiting its practical usage. Next, we show an efficient regularization method Jumpout, an improved version of dropout, based on linear models of ReLU networks. We discuss three novel insights about dropout for DNNs with ReLUs: 1) dropout encourages each local linear piece of a DNN to be trained on data points from nearby regions; 2) the same dropout rate results in different (effective) deactivation rates for layers with different portions of ReLU deactivated neurons; and 3) the rescaling factor of dropout causes a normalization inconsistency between training and test when used together with batch normalization. The above leads to three simple but nontrivial modifications resulting in our method “Jumpout.” Jumpout significantly improves the performance of different neural nets on multiple datasets, while introducing negligible additional memory and computation costs. Finally, we aim to explain the network behavior based on the linear model for every data point, particularly based on the bias term of the linear model. The gradient of a deep neural network (DNN) w.r.t. the input provides information that can be used to explain the output prediction in terms of the input features and has been widely studied to assist in interpreting DNNs. In a linear model (i.e., ), the gradient corresponds to the weights w. The bias b, however, is usually overlooked in attribution methods. We observe that since the bias in a DNN also has a non-negligible contribution to the correctness of predictions, it can also play a significant role in understanding DNN behavior. We propose a backpropagation-type algorithm “bias back-propagation (BBp)” that starts at the output layer and iteratively attributes the bias of each layer to its input nodes as well as combining the resulting bias term of the previous layer. Together with the backpropagation of the gradient generating w, we can fully recover the locally linear model . In experiments, we show that BBp can generate complementary and highly interpretable explanations
Coordination and regulation of bystander memory T cells
No full textThesis (Ph.D.)--University of Washington, 2021Memory T cells (Tmem) are best understood as highly specific killers, using their T cell receptor (TCR) to identify and rapidly eliminate infected or transformed cells. Although T¬mem are best understood through the lens of TCR-mediated activation, it is not the lynchpin of their function. Tmem can also respond to pro-inflammatory cytokines in the absence of activating TCR signals, termed bystander activation. Although this phenomenon was described over 20 years ago, it was thought to be of little biologic significance until it was shown that bystander activated Tmem could kill using TCR-independent mechanisms. These killing programs were best understood during systemic and/or chronic inflammation, leading many to classify bystander-mediated killing as an “immunologic accident” resulting from astoundingly high levels of inflammation. The full gamut of circumstances in which bystander activation can occur, as well as the signals governing it, remain unclear. Here we address basic questions in bystander T cell biology: 1) in what biological contexts does bystander activation occur, 2) how can limited inflammation beget bystander activation, 3) what fate changes accompany bystander activation, and 4) do mechanisms that attenuate/control bystander activation exist? Using both clinical vaccine samples and mouse models, we demonstrate that bystander activation is not a niche phenomenon but can also arise during dose- and anatomically-restricted inflammation. This hinges on CXCR3-mediated recruitment of Tmem to sites of early immune activation, at where they become bystander activated by pro-inflammatory cytokines and contribute to localized target killing. While induction of cytotoxicity is the best-known consequence of bystander activation, we demonstrate that other fates, like tissue retention, can be elicited by inflammation alone. Using a multi-omics approach, we demonstrate that homeostatic cytokine networks exist in healthy human placental tissues, which sufficiently bystander activate Tmem without precipitating cytotoxicity and tissue pathology. We show this to result from cooperation between anti-inflammatory cytokines and metabolites, which restrain bystander-mediated cytotoxicity without forfeiting other activation-induced programs of tissue retention and surveillance. Together, we show that bystander activation is an intentional program oft used to maximally leverage Tmem in the absence of TCR agonism and identify mechanisms to better wield or restrain this population therapeutically