8413 research outputs found
Sort by
A framework for generalizable neural networks for robust estimation of eyelids and pupils
No full textDeep Neural Networks (DNNs) have enabled recent advances in the accuracy and robustness of video-oculography. However, to make robust predictions most DNN models require extensive and diverse training data, which is costly to collect and label. In this work, we seek to improve the cost/benefit ratio of labeling to model performance. We develop pylids, a pupil- and eyelid-estimation DNN model based on DeepLabCut. We show that performance of pylids-based pupil estimation can be related to the distance of test data from the distribution of training data. Based on this principle we explore methods for efficient data selection for training our DNN. We show that guided sampling of new data points from the training data approaches state-of-the-art pupil and eyelid estimation with fewer training data points. We also demonstrate the benefit of using an efficient sampling method to select data augmentations for training DNNs. These sampling methods aim to minimize the time and effort required to label and train DNNs while promoting model generalization on new diverse datasets.National Science Foundatio
Delineating Glacial Features from 2-Dimensional Seismic Data (CHIRP) and development of a data processing workflow: Harrington Lake, Piscataquis County, Maine, U.S.A.
No full textGlacial features, including eskers, are preserved across Maine. Abundant lakes mask the glacial landscape throughout the state, obscuring spatial continuity of glacial features. Using 2-D CHIRP combined with lidar data, glacial features can be mapped below present-day lake surfaces to help improve the understanding of glacial and post-glacial landscape change.Harrington Lake is a 5.5 km-long lake located in central Maine, situated in a NW-SE trending valley and shallowing and draining to the northwest. The lake was dammed in 1930 with a 4.9 m-high dam. At present, the maximum lake depth is 42 m, suggesting the presence of a lake or a series of small ponds prior to damming. CHIRP reflection data were collected in a grid pattern covering much of the lake. Six NW-SE lines and 22 NE-SW lines were interpreted in combination with lidar data from the adjacent landscape. Points along each CHIRP line were selected to mark 1) depth to lake bottom, 2) thicknesses of Holocene sedimentary units, 3) thickness of Pleistocene glacio-lacustrine sediments, 4) thickness of the basal unit, and 5) depth to the top of the basal unit. Annotated CHIRP images were georeferenced using ArcGIS Pro and the Natural Neighbor technique was used to generate raster contour maps of the basin bathymetry and the various sediment thicknesses. In addition, drill core was recovered from the northern part of the lake following seismic imaging and used in conjunction with the thickness maps to interpret glacial geomorphology and post-glacial evolution of Harrington Lake.
A prominent esker oriented NW-SE is clearly expressed in the lidar data and extends into the northeastern part of the lake. Another small esker with a similar orientation occurs west of the prominent esker, and a third esker enters the lake along its eastern shoreline. Lake depth is shallowest over the eskers and deepest to the west and south of the eskers. Two Holocene lacustrine sedimentary units were measured and mapped along with a third sedimentary unit that is glacio-lacustrine. Total thickness of the three sedimentary units was mapped at 23 meters in the deepest parts of the lake. Post-glacial sedimentary units onlap the eskers and are absent in some areas, suggesting fluvial modification within the basin following glacial retreat. Based on initial core descriptions, the Holocene deposits consist of a post-glacial organic-rich mud unit which overlies a laminated glacial-lacustrine unit (likely varves) that is interbedded with coarser sediments in some areas. The basal unit is interpreted as glacial till, gravels associated with eskers, and/or bedrock. Basal unit thickness varies but is thinnest near the esker and along the western shore of the lake. The observations indicate that Harrington Lake now occupies a long valley that once contained a series of smaller ponds or marshes connected by a stream, as indicated by sedimentary unit thickness. The analyses of the lake bathymetry and sediment thicknesses reveal the position and geomorphology of an esker beneath Harrington Lake and provides a methodology for better understanding glacial geomorphology and deposits obscured by other lakes in Maine and elsewhere
Evaluating Websites: A Teacher's Guide
No full textA guide for instructions on teaching students about evaluating websites when conducting research
Utilizing Chemical Biology Approaches to Study Streptococcal Quorum Sensing
No full textBacteria have proven to be vital to the global ecosystem and our everyday lives, as they largely influence the delicate balance between health and disease. Of particular interest are Gram-positive bacteria known as streptococci, as these bacteria are incredibly common throughout the human microbiota. Many streptococci are considered commensals that contribute to beneficial health effects; however, some are also considered opportunistic pathogens, possessing natural competence for genetic transformation and the improved ability to acquire antimicrobial resistance. While it may seem logical to simply develop new antibiotics to treat infection, it is incredibly time-consuming, fiscally demanding, and difficult to discover novel antibiotic classes and develop new antibiotic derivatives. Moreover, the advent of antibiotic derivatives is closely followed by the emergence and spread of bacterial resistance, so continually developing new antibiotics has become ineffective. The impracticality of developing new antibiotic classes, coupled with the increased prominence of bacterial resistance demands the development of novel therapeutics that avoid selecting for antibiotic resistance amongst these bacteria. One such method is utilizing chemical biology tools to target nonessential pathways to hinder bacterial proliferation and pathogenesis. Here, the investigation of a cell-to-cell signaling mechanism known as quorum sensing (QS) is explored to further elucidate mechanisms to address and treat bacterial infection without selecting for antibiotic resistance. Generally, streptococci coordinate group behaviors and downstream phenotypes via the competence stimulating peptide (CSP)-mediated QS pathway known as the comABCDE regulon. Once exogenous CSP reaches a critical concentration threshold, bacterial populations are capable of acting synchronously to engage in specific, advantageous behaviors and phenotypes. This study aims to explore this pathway to gain insights about downstream phenotypic expression profiles, such as competence, biofilm formation, virulence factor production, and peroxide formation.
To this end, the competence regulon and downstream phenotypic expression profiles have been investigated in three species: Streptococcus sinensis, Streptococcus cristatus, and Streptococcus gordonii. In Chapter II, the role of the competence regulon in the pathogenesis of S. sinensis, an emerging human pathogen, was explored, with specific attention to transcriptomic and phenotypic effects following CSP exposure. In Chapters III and IV, the possibility of utilizing the competence regulon to enhance the biotherapeutic potential of two oral commensals, S. cristatus and S. gordonii, respectively, was investigated. Both species had previously demonstrated the ability to generate inhibitory hydrogen peroxide against a notorious oral pathogen, Streptococcus mutans, however studies here aimed to understand the connection between this phenotype and the competence regulon. Overall, this work largely explores the use of chemical biology tools to study streptococcal QS with the aim of providing foundational knowledge for future biotherapeutic approaches
An Evaluation of the Effect of Treatment Integrity Errors and Observation Conditions on the Accuracy and Reliability of Data Collection
No full textWithin the applied domain of behavior analysis (ABA), it is common practice to conduct measurement assessment (e.g., IOA). These measures are often collected as part of an ongoing evaluation of behavioral services and applied research. In other words, critical decisions related to the presumed success or failure of specific ABA interventions and procedures in applied practice and research are based on data that are collected during the course of service delivery and applied research (Vollmer et al., 2008). However, several factors in relation to staff-delivered consequences in the context of applied practice and research may impact the accuracy and reliability of collected data, including the treatment integrity of the services delivered along with the observation conditions of the data collection period. Data that are dependent on staff-delivered consequences, instead of the client behaviors, may result in making data-based decisions that can detrimentally impact the client and the success of the intervention (Vollmer et al., 2008). As such, it is crucial that data that are collected in the context of service delivery and research are both accurate and reliable in relation to the client’s/participant’s behaviors. The present research evaluated the effects of treatment integrity errors and observation conditions on the accuracy and reliability of data. Participants collected data on scripted videos of role-played client sessions, where we conducted two experiments manipulating: (a) treatment integrity and (b) observation conditions. Results showed that treatment integrity errors alone did not consistently impact data accuracy and data reliability, but that treatment integrity errors when paired with obscured observation conditions detrimentally impacted both the accuracy and reliability of collected data
The joint distribution of the maximum and duration of stochastic events driven by Pareto II observations
No full textWe study the joint distribution of X and N, where N has the geometric distribution and X is the maximum of the N independent and identically distributed Pareto II (Lomax) observations. The bivariate distribution is referred to as the geometric Marshall-Olkin Lomax distribution (GMOL). A related model for a geometric maximum of IID exponential observations was introduced by Kozubowski and Panorska (2008) and has proven useful in areas such as finance, hydrology and climate. However, the existence of heavy tails in environmental variables motivated this model. Our results for this research include derivations of the joint probability density function, cumulative distribution function, conditional and marginal distributions, conditional survival function, moment-generating function, Laplace transforms, and covariance matrix. We also address the problem of parameter estimation using the method of maximum likelihood. Estimation is empirically verified using a simulation study. We also present results of modeling precipitation and temperature data sets
Remaining Unperturbed by the Vibrational Response: Probing Vibrational Coupling, Relaxation, and Solvent Effects with 2D IR Spectroscopy
No full textUnderstanding molecular function first requires an understanding of their native structuresand fluctuations of their environments. One of the most powerful ways to probe the dynamic
behavior of molecular systems in solution is through the use of Two-Dimensional Infrared
(2D IR) spectroscopy. 2D IR is an ultrafast laser technique that is capable of probing the
vibrational properties and dynamics down to the picosecond or even femtosecond timescales.
This technique enables the analysis of vibrational couplings, energy redistribution pathways,
and solvation dynamics in real time. The early chapters of this dissertation will provide a semi-rigorous introduction to nonlinearresponse theory and explore newly developed models to more accurately capture the bilinear
coupling strengths between vibrational modes. These support experimental and
computational studies of vibrational energy flow, structural determination, and
solute-solvent interactions. A major focus of the research projects covered in this work will
be on Intramolecular Vibrational Energy Redistribution (IVR) on two key systems. The
first feature aromatic compounds that incorporate a heavy atom (selenium) as a way to block
energy transfer between an azido and cyano vibrational probe. Another involves a
terpyridine-aldehyde ligand that is relevant to spin relaxation in single-molecule magnets. In
both of these works, 2D IR is utilized to reveal vibrational coupling and energy transfer
pathways across the molecular scaffold. The last project that will be discussed is on the detection of localized solvation dynamicsthrough frequency fluctuations correlation functions (FFCFs), where the properties of a
cyanamide vibrational probe placed onto a deoxycytidine nucleoside are investigated in a
variety of viscous environments. This study not only reveals the relationship between the
FFCF decay of the NCN reported and solvent viscosity, but further correlates the decay time
directly with the nanoscale solvent dynamics via molecular dynamics (MD) simulations
Investigating The Role of Pancreatic eIF2-alpha kinase (PEK) In DLK/ Wallenda (Wnd)-mediated Neurodegeneration In Drosophila
No full textNeurons, the cells that are electrically excitable and form the basis of the nervous system, display an impressive range of morphologies and functions, which together drive processes from sensation and movement to cognition. These unique and mighty cells also have developed unique features such as dendrites as the recipient of signals, axons as wire for transmitting signals, and sophisticated protein localization and axonal transport for protein function and cellular homeostasis. However, neuronal homeostasis is constantly challenged by various intrinsic and extrinsic stressors, and results mislocalization and aggregation of these proteins occurs, subsequently resulting in functional and structural alterations in neurons, and leading to neuronal dysfunction and death, the characteristic features of neurodegenerative diseases. The highly conserved Unfolded Protein Response (UPR) kinase Pancreatic eIF-2α kinase (PEK) and the Dual Leucine Zipper Kinase (DLK) signaling kinase Wallenda (Wnd) are important regulators of neuronal health and neurodegeneration. Although these two (2) kinases are very important, their interaction in degenerative and regenerative responses are not fully understood. Therefore, understanding their cellular and molecular link is essential for fast-tracking the treatment of neurodegenerative disorders. Chapter 1 describes the mechanisms of protein homeostasis, including subcellular localization and axonal transport, essential for their survival. It highlights how the disruption of these fundamental processes leads to protein mislocalization and aggregation, which are central to neurodegenerative diseases.
Chapter 2 we investigate the novel regulatory and functional interaction between the ER stress kinase PEK and the axonal integrity regulator DLK/Wallenda (Wnd) in Drosophila. We demonstrate that overexpression of either PEK or DLK/Wallenda induces severe neurodegeneration and axonal arborization, and, critically, that PEK modulates Wnd's axonal enrichment, a defect rescued by PERK inhibition in neurodegenerative disease models. Our findings highlight PEK and DLK signaling are intertwined in degenerative and regenerative neuronal responses, revealing therapeutic potential for targeting this pathway in neurodegenerative conditions.
Finally, Chapter 3 discusses and addresses future directions, focusing on elucidating the precise molecular mechanisms by which PEK regulates Wnd axonal enrichment and influences its activity.
This thesis provides insights to current understanding of the interplay of PEK and DLK/ Wnd in neurons using Drosophila models
Perception of Warm and Cool Colors
No full textWarm and cool judgements of color perception are a fundamental aspect of visual experience. Very few studies have been conducted on this aspect of perception and only recently has this dimension of color perception gained traction in the visual sciences. In this work, I examine the nature of warm vs cool judgements, what dimension they inhabit in color processing, how we perceive them, and how warm-cool hues influence visual adaptation. In the first study, I examine how individuals categorize colors in cone-opponent space and how ratings of warm versus cool map onto asymmetries inherently found in perceptually uniform color spaces. In the second study, I explore the reason for this asymmetry and why the warm-cool dimension aligns well with it by using a visual search paradigm to test the salience characteristics of hues. In the third and final study, I test differences between the warm and the cool dimension using a threshold task to examine which side (warm or cool) dominates perception. Together, these studies should provide a picture of the nature of warm-cool judgements and what colors we are most adapted to given our visual diets
Ecological Perspectives on Cannabis sativa Phytochemistry: Biotic Interactions and Abiotic Stress in Managed and Natural Systems
No full textCannabis sativa is a chemically diverse and ecologically significant plant whose re-cent expansion across North American landscapes offers a unique opportunity to investi-gate the ecological roles of phytochemical diversity. While much attention has been given to cannabinoids for their commercial value, relatively little is known about how variation in phytochemistry influences arthropod interactions, or how these interactions differ across environmental conditions and cultivation systems. This dissertation takes an ecological lens to Cannabis phytochemistry, combining observational surveys, comparative analyses, and field experiments to explore how chemical traits structure arthropod communities across both natural and human-shaped systems. One of the first steps toward understanding any plant-insect system is to characterize the arthropod community. To answer this, I surveyed arthropods across 29 Cannabis sites in five western U.S. states, encompassing feral hemp, CBD- dominant hemp, and THC- dominant marijuana grown under diverse conditions. These field surveys—spanning backyard gardens, research plots, and large-scale farms—revealed rich and variable arthropod communities. A total of 5,817 arthropods representing 194 morphospecies from 13 insect orders were collected. Hemiptera, particularly aphids in the genus Phorodon, dominated in both abundance and site occupancy and were the most consistent insect associates across chemotypes and regions. Cultivated hemp supported the highest morphospecies richness, followed by cultivated marijuana and feral hemp. Most morphospecies occurred in only one location or chemotype–region combination, indicating high spatial turnover and little evidence for a core Cannabis-associated insect community. However, a few predators such as Orius sp. were broadly distributed and frequently co-occurred with aphids, suggesting potential roles in natural biological control.Building on these patterns, I turned to the role of selection. In Nebraska, where feral hemp has persisted in ditches and field margins over centuries, I compared six pairs of cul-tivated and feral Cannabis populations to assess how natural and artificial selection influ-ence plant chemistry and ecological dynamics. Cultivated plants—selected for high canna-binoid production—had greater cannabinoid diversity but reduced terpene diversity com-pared to their feral counterparts. These chemical shifts had ecological consequences: in cul-tivated fields, higher phytochemical diversity was associated with lower herbivore abun-dance, while in feral sites, this relationship was weaker. Natural enemies tracked herbivores in both systems, but predator-prey dynamics were stronger in feral fields. These findings suggest that selection for cannabinoid yield may weaken terpene-based defenses, altering not only the chemical phenotype of the plant but also its ecological relationships.
To disentangle the effects of environment from selection history, I conducted a fully factorial field experiment exposing clonal hemp plants to three water treatments over three growing seasons. Abiotic stressors altered both plant form and function: stressed plants were shorter, had reduced phytochemical diversity, and supported fewer arthropods. Can-nabidiol concentration was positively associated with arthropod richness, while high overall phytochemical diversity—particularly under stress—tended to suppress herbivore abun-dance. These results suggest that phytochemical diversity can serve multiple, and some-times opposing, ecological functions depending on context, and that environmental stress reshapes the ecological phenotype of Cannabis in ways that are not easily predicted by chemistry alone.
Across all three projects, a unifying pattern emerged: phytochemical composition plays a central role in shaping insect communities, but its effects are complex, context-dependent, and shaped by both human intention and environmental forces. Together, these findings offer new ecological perspectives on a plant that is both ancient and newly rele-vant, and demonstrate the value of blending natural history, field ecology, and metabolomics to understand how chemical traits mediate biotic interactions in emerging agroecosystems