Search

Eye tracking has proven to be a valuable tool for a number of disciplines such as imaging science, vision science, psychology, and neurology. They have used eye-tracking to study the human visual system and to understand how humans attend to their environment for different tasks. Analyzing the gaze data obtained from eye-trackers begins with visualizing the gaze point. By monitoring gaze we can answer questions including, ‘where was the person looking?’ and ‘what image region or object was the person looking at?’ Traditional methods aimed at answering such questions from gaze data: required frame-by-frame analysis of videos, manually annotating gaze. In this dissertation, I describe a new automated tool that aids and simplifies analysis of gaze data from one or more observers performing complex tasks in natural environments. I implemented a number of analysis pipelines that process raw gaze video, build a model of the environment, calculate 3D gaze from 2D gaze video, summarize each observer’s 3D gaze interaction with the environment, and compute the observer’s 3D location in the environment. The final pipeline includes an advanced visualizer that allows one to play and pause multi-observer interaction data like a video.

The process of mapping neuroanatomy at multiple scales is defined as neurocartography and has the ultimate goal of revealing a complete wiring diagram of synaptic connections. Neurocartography is an important pursuit because brain function is intricately linked to neuroanatomy, similar to how the function of a protein depends on its structure. Here, we developed new light microscopy approaches for neurocartography at both microscale and nanoscale levels to map single neuron morphologies and synaptic connectivity respectively. Spatially sparse labeling of neurons has been necessary when studying neuron morphologies to minimize overlap and avoid ambiguities during reconstructions. We developed two strategies to overcome this limitation and achieve spatially dense labeling. First, we used a multicolor genetic labeling (Brainbow) approach to stochastically express fluorescent proteins in a spatially dense population of neurons, facilitating reconstruction of single neuron morphologies. We extend the Brainbow viral toolbox by 1) introducing 12 fluorescent proteins in the form of 6 new viruses, 2) using a membrane and cytoplasmic dual labeling strategy, and 3) adopting the AAV.PHP.eB capsid to systemically induce expression and better control color diversity. Second, we used expansion microscopy to increase confocal imaging resolution by physically magnifying brain samples. We developed a multi-round immunostaining Expansion Microscopy (miriEx) protocol that enables multiplexed protein detection at multiple imaging resolutions. We then combined Brainbow with miriEx to simultaneously map morphology, molecular markers, and connectivity in the same brain section. We define the derivation of these properties from hyperspectral fluorescent channels as spectral connectomics, a light microscopy based approach towards mapping neuroanatomy and connectivity with molecular specificity. We applied our multimodal profiling strategy to directly link inhibitory neuron cell types with their network morphologies. Furthermore, we showed that correlative Brainbow and endogenous synaptic machinery immunostaining can be used to define putative synaptic connections between spectrally unique neurons, as well as map putative inhibitory and excitatory inputs. We envision that spectral connectomics can be applied routinely in neurobiology labs to gain insights into normal and pathophysiological neuroanatomy across multiple animals and time points. We hope that the light microscopy approaches developed in this dissertation will facilitate the extraction of new biological insights from neurocartography.

Altering plant species to make crops for human use is a centuries old practice. Through traditional breeding, most of the extant crop varieties used today have been optimized to increase yield and growth potential. Despite advancements over time, current breeding is limited in its throughput. In order to address increasingly problematic issues such as climate change, a multitude of technologies have been developed to improve the speed and ease of this process. Technologies such as CRISPR/Cas based gene editing and innovative plant transformation methods promise to greatly democratize the ability to generate new cultivars of crops. Plant gene editing is typically performed by delivering gene editing reagents (e.g. Cas9 and sgRNAs) to explants in culture. Edited cells are induced to differentiate into whole plants by exposure to various hormone regimes. This process serves as a major bottleneck in current plant biological studies since creating edited plants through tissue culture has many drawbacks: the process is often inefficient, requires considerable time, works with limited species and genotypes, and causes unintended changes to the genome and epigenome. Developing methods to circumvent these issues by allowing for simpler delivery and less intensive regeneration across a wide number of species would obviate this bottleneck. Looking first to optimize editing reagent delivery, a simple and scalable reagent delivery method was required. To this end, improvements were made to existing Agrobacterium tumefaciens based co-culture methods, yielding a streamlined fast treated co-culture (Fast-TrACC) reagent delivery method for rapid reagent testing. The Fast-TrACC procedure combines treatment of Agrobacterium to increase T-DNA transfer and a luciferase reporter to allow for high-throughput screening after co-culture. Application of this method to deliver reporters, gene editing components and test expression elements in a number of dicot species has successfully demonstrated the broad usefulness of this delivery method. Next looking at the major bottleneck of regeneration, ways to overcome the tedious and time consuming nature of current regeneration practices were necessary. Advances in ectopic expression of plant developmental regulators have demonstrated the feasibility of inducing somatic cells to form meristems. A high-throughput method was developed to optimize combinations of developmental regulators for induction of meristems on dicot seedlings. After analysis of several different developmental regulator combinations, the combinations of ZmWUS2 & AtSTM as well as ZmWUS2 & ipt were found to be most effective in generating de novo shoots. When these developmental regulators are co-delivered with gene editing reagents, induced meristems produce fertile plants that transmit gene edits to progeny. The de novo induction of gene edited meristems sidesteps the need for tissue culture, promising to overcome a current bottleneck in plant gene editing. These methods were designed and established to function effectively in the model plant, Nicotiana benthamiana, but application in a crop species, such as tomato, is necessary to realize the technology’s full potential. Classical editing with tissue culture in tomatoes has been used to alter traits such as locule number and branching. Despite this ability to generate desired mutations, the key bottlenecks still exists in regenerating whole plants after editing has occurred. This limitation comes from the combination of tedious explant manipulation and subsequent regeneration. Based upon the inferences made from N. benthamiana, ectopic overexpression of ZmWUS2 & ipt was shown to promote re-differentiation of somatic tissues into meristems. By combining DRs with gene editing reagents, we have demonstrated edited tomato lines can be quickly established, thus circumventing the need for tissue culture. Using these tools, new avenues of engineered across plant species become more accessible, thus allowing for improvements in accelerated breeding, de novo domestication and reengineering metabolism in a simpler fashion than what was previously been possible.

This thesis outlines the development of several tools for the prediction of protein molecular functions. The limited data available for training represents a challenge for developing accurate, generalizable models. To combat this challenge, we established several molecularfunction predictors based on sequenceand structurebased inference from weakly annotated data. The core of the approach is driven by a geometric alignment between 3D structures as a proxy for functional relatedness. We also exploited a large dataset annotated for a related task to inform a model via transfer learning. In Chapter 2, SPOTLigand 2 was established for predicting proteinligand interactions. The method employs weakly annotated ligandbinding sequences to improve virtual screening performance by 93% (top 1% enrichment factor) when compared to a baseline using complex structures only. In Chapter 3, SPOTpeptide was developed for identifying peptidebinding domains and peptide binding sites. The method is an implementation of a structurebased homology modelling pipeline augmented by local measures of interface complementarity. This is the first method devoted specifically to the identification of peptidebinding domains at a genomescale and outperformed a simple sequencebased baseline by 30% according to Matthews correlation coefficient (MCC). Binding site MCC was also improved by 20% compared with the next best method from the literature. In Chapter 4, the SPalignbased structure homology framework was further validated in a prospective study involving the characterization of Bacillus subtilis YesU as a carbohydratebinding protein. In Chapter 5, we developed a deep learning model called SPOTMoRF for the identification of short peptide segments that undergo a disordertoorder transition when binding a functional partner. The deep learning model was facilitated by a transfer learning framework which allowed an MCC improvement of 40% compared with a model trained directly from random initialization. These tools are made available as web services from http://sparks-lab.org/ and should facilitate the annotation of protein function at various levels of resolution.

The study of human vision can include our interaction with objects. These studies include behavior modeling, understanding visual attention and motor guidance, and enhancing user experiences. But all these studies have one thing in common. To analyze the data in detail, researchers typically have to analyze video data frame by frame. Real world interaction data often comprises of data from both eye and hand. Analyzing such data frame by frame can get very tedious and time-consuming. A calibrated scene video from an eye-tracker captured at 120 Hz for 3 minutes has over 21,000 frames to be analyzed. Automating the process is crucial to allow interaction research to proceed. Research in object recognition over the last decade now allows eye-movement data to be analyzed automatically to determine what a subject is looking at and for how long. I will describe my research in which I developed a pipeline to help researchers analyze interaction data including eye and hand. Inspired by a semi-automated pipeline for analyzing eye tracking data, I have created a pipeline for analyzing hand grasp along with gaze. Putting both pipelines together can help researchers analyze interaction data. The hand-grasp pipeline detects skin to locate the hands, then determines what object (if any) the hand is over, and where the thumbs/fingers occluded that object. I also compare identification with recognition throughout the pipeline. The current pipeline operates on independent frames; future work will extend the pipeline to take advantage of the dynamics of natural interactions.

Sexual assault prosecution incorporates forensic expert detection of AGI to support the forcible nature of suspected rape. Understanding how researchers are currently describing, detecting and measuring AGI after SA would help healthcare providers, SA researchers, law enforcement and the justice system communicate more consistently. An extensive literature review was performed which exposed the inconsistences researchers are currently using to detect and describe AGI after SA. Next, the Vulvar Acute Genital Injury Score (VAGIS) tool was developed using a modified e-Delphi method in a panel of 15 SA forensic experts. After a set of 53 digital images were judged for quality, a panel of 14 forensic examiners validated the images for use with the VAGIS tool. To further develop the VAGIS tool, a group of SA forensic examiners were tested for consistency at scoring the digital image set on the VAGIS. This is a call to all national forensic examiner organizations to come together with the common purpose of standardizing the language around acute genital injuries after SA. Standardizing communication in this area could improve services to individuals who suffer from sexual assault and increase confidence that the justice system designed to protect and defend victims is serving at its highest capacity.

For the first objective, predicted impacts of selected Mississippi River Basin Initiative (MRBI) conservation practices (CPs) on sediment and nutrient loss were assessed. The study area was L’Anguille River Watershed (LRW), a priority focus watershed of the MRBI program and the simulated CPs were filter strip, critical area planting, grade stabilization structure, irrigation land leveling, irrigation pipeline, irrigation water management, and nutrient management. The Soil and Water Assessment Tool (SWAT) model was calibrated (1998 – 2005) and validated (2006 – 2012) for flow, sediment, total phosphorus (TP), and nitrate-nitrogen (NO3-N) at the Colt site (503 sq. km. drainage area) and for total flow, surface flow, and base flow at the Palestine site (784 sq. km. drainage area). The statistical results for the calibration and validation were found to be satisfactory or better except a few root mean square error – standard deviations ratio (RSR) values for the calibration period. The SWAT model results were predicted from 2013 – 2017 for assessing the performance of CPs. Out of the CPs used in the LRW, critical area planting was the most effective in reducing the predicted nutrient (58% TP and 16% total nitrogen (TN)) and sediment (80%) loads, followed by filter strip, irrigation land leveling, grade stabilization structure, irrigation pipeline, irrigation water management, and nutrient management. Results such as these could help inform watershed planners and policy makers in selecting appropriate CPs that will most effectively bring about desired nutrient and sediment load reductions. For the second objective, a CP tool was developed with Python programming language for integrating a user-defined target area utilizing either a single or multiple selection criteria with the SWAT model. The tool uses open source packages such as Geospatial Data Abstraction Library (GDAL) and Matplotlib. The tool is standalone and was designed in such a way that it simulates CPs at the lowest simulation level (hydrological response unit) of the SWAT model by building a new targeting procedure for SWAT applications and decision-making. The tool automates the process for simulating CPs on a target area and analyzing differences between the baseline and CP scenario. The tool was evaluated for the Cache River Watershed (CRW). A target area was selected in the CRW and irrigation land leveling CP was simulated. A 22% decrease in sediment losses, 20% decrease in TP losses, and 12% decrease in TN losses were predicted. The tool provides a quick approach to address the water quality impacts on a specific target area. For the third objective, the Python-based CP tool developed for objective 2 was further updated to simulate CPs at user-defined locations using an interactive simulation approach. The tool allows the user to select the target area with mouse-clicks in a user-friendly and interactive environment. The LRW located in northeastern Arkansas was used as the test area. A target area was selected interactively in LRW and filter strip and irrigation land leveling CPs were simulated. A 70% decrease in sediment losses, 68% decrease in TP losses, and 47% decrease in TN losses were predicted.

OBJECTIVES: To explore the issues around health related quality of life (HRQoL) of children living with HIV (CLH) through qualitative inquiries and to formulate and validate a culturally acceptable HRQoL instrument.METHODOLOGY: The study was conducted in three districts of West Bengal, India. During the qualitative phase, we conducted 34 in-depth interviews (IDI) with caregivers of CLH and 4 focus group discussions (FGD) with 8 to 15 year old CLH. We identified items affecting HRQoL through qualitative analysis to create 2 HIV-targeted scales – ‘symptoms’ and ‘discrimination’. These 2 scales were combined with 4 generic core scales of the ‘Pediatric Quality of Life Inventory (PedsQL)’ to formulate the ‘Quality of life (health-related) of children living with HIV/AIDS in India (QOL-CHAI)’ instrument. In order to assess applicability and detect problems with understanding of intended meaning of items, cognitive interviews were conducted with 10 children (5 each from 8 - 12 years and 13 – 15 year age categories). In the quantitative phase we recruited 199 CLH and 194 children who were exposed but not HIV infected (HIV-affected) to evaluate psychometric properties of the QOL-CHAI instrument and to compare HRQoL between CLH and HIV-affected children.RESULTS: IDIs with caregivers identified several barriers and facilitators related to pediatric disclosure of HIV status. Further, various perceptions and experiences associated with caregiving for CLH and issues such as discrimination, adherence to medication, grievances about available health care provisions were also revealed through qualitative inquiries. The QOL-CHAI showed acceptable psychometric properties with Cronbach’s α for the scales ranging between 0.69 and 0.85. Factor structure of the QOL-CHAI was found to be roughly in agreement with the domain specific categorizations. In terms of convergent validity, symptom scale score showed significant negative correlation with CD4 cell count (Pearson’s coefficient=-0.23, p

Engineered cellular systems are a promising avenue for production of a wide range of useful products including renewable fuels, commodity and specialty chemicals, industrial enzymes, and pharmaceuticals. Achieving this breadth of biological products is facilitated by the diversity of organisms found in nature. Using biological and engineering principles, this diversity can be harnessed to make efficient and renewable bio-based products. Such advancements rely upon our ability to modify host genetics and metabolism. This work focuses on the development of new biotechnological tools which enable cellular engineering, and the implementation of these tools in eukaryotic systems. Mammalian cell engineering has important implications in protein therapeutics and gene therapy. One major limitation, however, is the ability to predictably control gene expression. We address this challenge by examining critical aspects of gene expression in human cells. First, we evaluate the impact of selection markers, a common mammalian expression element, on cell line development. In doing so, we determine that Zeocin is the best selection agent for human cells. Next, we identify loci across the genome that support high level expression of recombinant DNA and demonstrate their advantage for stable integration. Finally, we optimize a Cre recombinase based methodology that enables efficient retargeting of genomic loci. Collectively, this work augments the current genetic toolbox for human cell lines. Beyond basic gene expression, there is interest in understanding global interactions within the cell and how they relate to phenomena including gene regulation, expression and disease states. Although our tools are not yet sufficient to study these phenomena in many hosts, methods can be developed in lower eukaryotes and then adapted for more complex hosts later. We demonstrated two methods in S. cerevisiae that utilize a systems-level approach to understand complex phenotypes. First, we developed condition-specific codon optimization that utilizes systems biology information to optimize gene sequence in a condition-specific manner. Additionally, we developed a Graded Dominant Mutant Approach which can be used to dissect multifunctional proteins, understand epigenetic factors, and quantitatively determine protein-DNA interactions. Both can be implemented in many cellular hosts and expand our ability to engineer complex phenotypes in eukaryotic cell systems.