The electrochemical oxidation of methane is a promising reaction that can provide an alternative to methane flaring for gas drilling sites. However, there is still little known about the ideal conditions for this reaction, making it unrealistic for industrial use. This study aims to further the understanding of the methane oxidation reaction (MOR) by analyzing the products under different potentials. A titanium cylinder was plated with a cobalt catalyst and then rotated in a potassium carbonate solution while methane flowed into the solution. Samples were taken from the cell every 20 minutes for 2 hours and analyzed using NMR. It was found that methanol and acetate were predominantly formed at low potentials. Methanol averaged over 1.6 μM at potentials of 0.6 to 1.0 V and acetate averaged 25 μM at potentials of 0.6 and 0.8 V. To add further support, methanol and formate only had a substantial faradaic efficiency at 0.7 V of 7.9% and 7.6% respectively. Formate, a compound not detected in previous MOR studies, was observed when acetate persisted at high concentrations above 40 μM, suggesting that there is a reaction occurring that relates the two. With this knowledge, a baseline for comparison can be set for experiments with other catalysts. The discovery of formate will encourage further studies of the mechanism when acetate is at high concentrations. In future work, additional catalysts will be analyzed using this setup to determine what conditions are the best for producing useful products.

When trying to understand molecular variation in biological systems, the use of numerical models provides key insight into cellular dynamics and can serve to predict future states. Traditionally, this has been done through mathematical modeling; however, the advent of machine learning has enabled more powerful computational tools. To integrate the interpretability of explicit modeling with the robustness of machine learning, a hybrid method, CellBox, has been proposed. However, the generalizability and scalability of this model are not yet optimized. To overcome these issues, we developed a novel approach, FactorBox, which simplifies the CellBox model and increases fitting efficiency through a combination of iterative solving and optimization functions. Here, we apply FactorBox to data from the NIH LINCS project, wherein thousands of genes were knocked down and their transcriptional effects measured, to quantify how similar the resulting network models are across cell lines. We find few shared interactions between cell lines, signaling a lack of consistency in the model. We then formulate a cross-validation scheme to determine the generalizability of the model's predictions. Using this, we discuss potential adjustments to be made, such as prioritization of the iterative solving process over the optimization method. With this analysis, we will be able to refine FactorBox to ensure that it produces accurate predictions and can be used to map gene regulatory interactions in a variety of biological settings.

Adipose tissue plays a major role in maintaining whole-body glucose or energy homeostasis. Based on the rate of energy metabolism, adipose tissue is classified into white adipose tissue (WAT) that predominantly stores energy and brown adipose tissue (BAT) capable of oxidizing fatty acids to produce heat through a process called thermogenesis. In addition, some adipocytes capable of oxidizing fatty acids commonly referred to as beige adipocytes are also found within WAT. Increasing the metabolic capacity of such beige adipocytes can improve whole-body glucose homeostasis. Previous studies from our lab have shown that the loss of mitochondrial pyruvate carrier (Mpc1) in brown adipocytes increased fatty acid oxidation capacity. Whether Mpc1 inhibition will also promote fatty acid oxidation in the white adipocytes remains to be tested. We aim to genetically ablate Mpc1 in both brown and white adipocytes in vivo and study the impact on adipose metabolism and whole-body energy homeostasis. Our preliminary studies show that loss of Mpc1 in brown and white adipocytes leads to the formation of small “beige-like” adipocytes exclusively in the subcutaneous/inguinal white adipose tissue (iWAT). Additionally, improved glucose homeostasis in the adipose-specific Mpc1 knockout mice when compared with control mice has been observed. Our ongoing studies will elucidate the molecular mechanisms behind thermogenesis in the iWAT and the role of peripheral organs contributing to improved glucose metabolism in adipose-specific Mpc1 knockout mice. The outcomes will help us understand the mitochondrial metabolic intricacies in the adipocytes contributing to metabolic health.

Modern-day x-ray light sources use relativistic electron beams to produce photons. Quadrupoles and undulators are magnetic devices that form an integral part of these light sources. Current quadrupole and undulator technology at the micro to millimeter scale use electromagnets and are fabricated on silicon wafers. The objective of this research project is to explore the usage of printed circuit boards (PCBs) in quadrupoles and undulators and to test their limitations. PCBs could potentially make the current fabrication process of these devices easier and cheaper. However, PCB-based devices cannot reach the same current densities as silicon-based ones due to heating; the thermal conductivity of FR4, the PCB material, is significantly lower than that of silicon. Four designs (1 quadrupole, 3 undulators) with 1 to 2 trace layers were created in DipTrace and manufactured. COMSOL simulations showed that the four designs could run at current densities from 2.1×108 to 2.7×108 A/m2 before the FR4 reaches its glass transition temperature (130°C). Experimental data obtained with a thermal camera agreed reasonably well with the simulated data. The simulations and experiments both demonstrated that wider copper traces and more trace layers reduce temperature rise for the same total current but increase temperature rise for the same current density. Initial magnetic simulations of the quadrupole show that the quadrupole design can produce a magnetic gradient up to 34 T/m at the maximum experimental current density and 100 μm gap size. Future works include conducting real-world magnetic tests with the PCB designs.

Quiescence is the reversible arrest of cell proliferation that is important across biological processes like stem cell maintenance and cancer cell dormancy. We and others have shown that the transition to quiescence leads to specific upregulation of the H4K20me3 mark and widespread changes in gene expression and chromatin organization. Previous studies have shown the methyltransferase responsible for H4K20 trimethylation (Suv4-20h2) may be involved in the recruitment of other proteins that regulate genome organization. Here we investigated the relationship between chromatin architecture and the H4K20me3 mark with quiescence. For this, we used a quiescence cell culture model based on human dermal fibroblasts, and characterized the quiescent cells based on low expression of proliferation markers (like PCNA) and low levels of DNA and RNA content using flow cytometry. To probe chromatin organization, we used dCas9-mediated fluorescence in-situ hybridization (CasFISH) that utilizes a catalytically inactive Cas9 protein to bind to specific genomic loci. This revealed that the number of detected telomere puncta in quiescent cells is lower than proliferating cells. Furthermore, the Suv4-20h2 knockout cells had highly reduced levels of H4K20me3 and a different number of detected telomeric and centromeric puncta than WT cells, suggesting changes in chromatin organization upon depletion of H4K20me3. Preliminary cellular analysis using immunofluorescence microscopy appears to show that the distribution of H4K20me3 changes to become more nuclear with quiescence. These findings when taken together suggest that H4K20me3 may play a role in the changes associated with chromatin compaction and genome organization with quiescence.

Visual processing is an integral piece of robotic autonomy, as cameras attached to the robot collect and process pictures, allowing robots to sense the world around them. Robots collect information from the processed image and use it to make decisions. The goal of this project is to use colored blob detection and AprilTag detection to control low-cost robotic swarms, or cohesive groups of robots. I used an openMV camera, which is a small microprocessor with a camera attached, to accomplish this. To recognize colored blobs, I used thresholds to group pixels by color, which is a method that reduces issues with lighting changes. I programmed the camera to effectively recognize multiple different colored blobs at the same time and determine the distance between the camera and the object if the object’s dimensions are known. The error associated with this calculation is about 20 millimeters from a range of less than a meter. AprilTags are recognized by the unique pattern of the tag, and I determined that the range of detection is dependent on the tag size. I used the AprilTag recognition capability to determine the relative position of the robot. The calculation of the distance between the camera and the tag has an average error of about 40 millimeters at a range of 3 meters, which is a relatively insignificant amount. We then applied these capabilities to robot swarm autonomy, using them to recognize a target object, calculate distances between robots, and perform other swarm behavior.

The combined increase of intermittent renewable power sources and natural disaster related power outages have placed increasing pressure on load balancing for electric power utilities. Currently, many back-up or on-demand power sources have undesirably high emissions that counteract the benefits of renewables penetration. This project aims to develop a camless internal combustion engine with advanced high-speed sensors to monitor emission levels and performance and enable fuel flexibility to produce low emissions and high efficiency over a wide range of loads. Specifically, the sensors will be implemented to monitor temperature, NOx, and CO concentrations. The temperature and concentration measurements will be used as inputs for electric-hydraulic valve actuation system for real time optimization. In addition to less emissions, it is expected that this adaptable engine will demonstrate gains in efficiency over a variable load profile. It also will enable the use of new types of fuels, including renewable and carbon free fuels.

Synthetic Notch (SynNotch) is an engineered receptor that recognizes specific surface-bound ligands on neighboring cells to initiate gene transcription. Through genetic modifications, SynNotch circuits allow the user to define which surface-bound ligand on a neighboring cell will activate the gene circuit as well as its target genes. Cells presenting ligands on their surface have been primarily deployed to study the capabilities of SynNotch activation; however, interactions between SynNotch cells and neighboring cells present a challenge to multiplex and control with precision. Here, we explore the activation of SynNotch by presenting the ligand conjugated to a microparticle, mimicking the interaction between SynNotch cells and ligand-presenting cells. To characterize SynNotch activation, we developed ligand-conjugated particles of different sizes and ligand concentrations. We quantified its synthetic gene activation using an engineered fibroblast fluorescent reporter cell line and compared the percentage and extent of activation to the traditional SynNotch activation method. Our experiments demonstrate that activation of SynNotch circuits is dependent on the particle-to-cell surface area and ligand concentration on the particle. Also, we show that particle presentation of ligand yields a higher SynNotch activation response compared to ligand-presenting cells. These results reveal not only a new method of presenting ligands to SynNotch with greater modularity, but also provide insights on the concentration of ligand required to robustly activate gene transcription using SynNotch.

Convolutional neural networks are the state-of-the-art algorithm for object classification. Due to the various types of objects that are processed and to facilitate training, typical convolutional neural networks (CNNs) require data preprocessing like zero padding or 3D to 2D space projections and do not work with point cloud data. Light Detection and Ranging (LiDAR) is one of the main technologies used in self-driving cars and terrain Page 2 of 2 mapping. Since LIDAR uses time of flight from laser beams to create a 3D map of the area, the generated data is a point cloud. In order to solve these problems, here we present an implementation of CNNs using a modified PointNet architecture. PointNet architecture is directly capable of taking a point cloud and running it on the classification algorithm, which is much more efficient than transforming the data before being fed to the network. In this study, we optimize the said convolutional neural network based on PointNet architecture. We train the model using LiDAR data taken in Westwood and tune its parameters accordingly to achieve close to state-of-the-art performance. As of now, in preliminary testing, the model achieves an 89.82% training accuracy. The goal is to further achieve a model that can be able to map external environments to aid driver-safety and autonomous navigation.

Idiopathic pulmonary fibrosis (IPF) is a life threatening disease which affects about 100,000 people in the United States. Treatment of IPF and other fibrosis is limited due to a lack of viable models for pulmonary fibrosis (PF) that can reproduce its characteristic spontaneous initiation and continuous progression. Recently, an in vitro model called the induced fibrosis activation (iFA) model has identified specific cell types and profiled proteins and RNAs that possibly play a pivotal role in spontaneous and progressive development of PF. The iFA model is developed by increased stiffness of the extracellular matrix (ECM), which modulates the expression of these proteins and RNAs via various stresses, including endoplasmic reticulum (ER) stress. ER stress regulates gene expression and protein folding which could upregulate ECM components including collagens and aggravate PF. Previously, ER stress has been proven to induce myofibroblast outgrowth with collagen-1 (Col-1) deposition in PF-diseased lungs; however, it has not been clarified yet how ER stress potentiates collagen-1 overexpression in PF. Thus, in this study, we aim to reveal the role of ER stress in collagen-1 biosynthesis using the iFA model. We hypothesize that ER stress regulates collagen-1 expression in fibroblasts and drives activation of fibroblasts to myofibroblasts. By gaining a better understanding of regulatory effects of ER stress on Col-1 expression in pulmonary fibroblasts, we could pave the way to ameliorate fibrogenesis with restoring collagen homeostasis in PF-diseased lungs.