Relative Risk (RR) is a highly interpretable parameter in epidemiology and biostatistics, based on both binary input and outcome. It is frequently used in vaccine development to measure the relative efficacy between two treatment groups.
Researchers are often tempted to use generalized linear models (GLMs) to estimate the logarithmic RR with respect to a set of baseline covariates. However, this approach has inherent flaws, as GLMs do not account for variation dependence in Relative Risk on its nuisance parameters. Richardson et al. have developed an unconstrained and variation-independent doubly robust nuisance model using the log Odds Product (OP).
To expand on this work, we will explore alternative nuisance models—both those developed by us and those from other researchers—and compare their computational robustness to that of the log Odds Product (OP).
Additionally, using the brm R package (which streamlines the methods proposed by Richardson et al.), we will analyze a dataset where Relative Risk serves as the target of inference and compare these results to those obtained using regression methods.

Due to considerations such as dose reduction, or physical limitations of the scanner, computed tomography (CT) images must sometimes be reconstructed from sparse-view or limited-angle sinogram data, resulting in a loss of image quality. In recent years, there has been a great deal of interest in using neural networks to improve image quality in these scenarios. In this work, we implement three neural network architectures – denoising convolutional neural network (DnCNN), U-net, and transformer – and apply them to sparse-view and limited angle problems in both a post-processing and iterative, “plug-and-play” reconstruction context. In post-processing, the neural network is applied to the final image to remove artifacts, while in the plug-and-play approach, it is incorporated into the algorithm that reconstructs the image from the sparse-view or limited-angle data. Based on standard image quality metrics, the post-processing approach with the U-net is found to give the best image quality. The plug-and-play approach, while not always providing the best image quality, is able to ensure fidelity with the sinogram data.

Neonatal hypoxic-ischemic encephalopathy (HIE) occurs when the brain receives insufficient oxygen and blood supply before or during childbirth. HIE is a leading cause of neonatal mortality and morbidity that may also affect later brain development, specifically gyrification - folding of the cerebral cortex creating gyri and sulci. The nonhuman primate (NHP) brain is gyrified, similar to humans, making NHPs a highly translatable model to examine brain development after injury, which has not been well-studied in HIE. In our nonhuman primate (NHP) model of neonatal HIE, we induced injury through in utero umbilical cord occlusion (UCO) for 20 minutes, mimicking the cause of HIE in humans. Twenty-two term-equivalent pigtailed macaques (Macaca nemestrina) underwent UCO and were randomized to no treatment (n = 11) or treatment with therapeutic hypothermia and erythropoietin (TH + Epo [5x1000 U/kg]; n = 11), while non-UCO animals served as controls (n = 7). All animals were delivered via cesarian section. Injury severity was determined by physiological parameters (Apgar score), lactate, and pH levels after resuscitation. To evaluate the impact of injury on gyrification, we will utilize magnetic resonance imaging (MRI) taken 6-months post-injury to measure the gyrification index (GI). GI will be calculated by taking brain’s inner-to-outer hemispheric ratio; the inner trace following the contours of the gyri and sulci, and the outer trace following the circumference of the cerebral cortex. We hypothesize that global and regional GI will be altered in animals exposed to UCO, corresponding with decreased brain volume and greater injury. We also hypothesize that treatment will mitigate some of these changes, leading to a GI closer to control. These results will help determine whether hypoxia-ischemia alters the trajectory of cortical development, as well as the association between injury severity, brain volume, and gyrification.

Traumatic brain injury (TBI) results from a blow to the skull that causes shearing forces in the brain. Elevating intracranial pressure (ICP) at the moment of impact may protect the brain from TBI by stiffening the brain tissue and decreasing shearing. When they expect an impact, humans naturally brace and perform a Valsalva maneuver (exhaling against a closed airway), which momentarily elevates ICP. In a ferret TBI model, we conducted abdominal compression using a blood pressure cuff to induce a Valsalva-like response (VLR) and determine whether VLR resulted in neuroprotection. The ferret model was chosen for its gyrified brain structure and white to grey matter ratio that closely resembles the human. TBI was induced using a CHIMERA (Closed-Head Impact Model of Engineered Rotational Acceleration) device, which is designed to deliver high-energy, controlled skull impacts. Initial work showed that the abdominal compression procedures increased ICP. The TBI study involved a total of 36 adult ferrets of both sexes randomized into three groups: (1) a sham control group exposed to isoflurane with a cuff but no compression, (2) a TBI group with a cuff but no compression, and (3) a TBI group with a cuff and abdominal compression. Baseline behavioral assessments (CatWalk, Novel Object Recognition, Swim Test, and Open Field) were conducted one week prior to injury. Post-injury behavioral testing, using the same assessments, was performed at 24–48 hours and 8 days post-TBI to evaluate functional outcomes. On day 8, ferrets were euthanized, and their brain tissue was collected and assessed for neuropathological outcomes. We hypothesize that abdominal compression will mitigate deleterious TBI outcomes. If these findings are supported, this intervention could improve the lives of those at risk of TBI and contribute to ongoing research in the field.

As the sensitivity and capabilities of modern synchrotron facilities continue to develop, so does the field of computational material sciences in an effort to meet the demand for analysis of new properties in various systems. 3d transition metals are of special interest due to their wide range of conductive and optical properties. Traditionally, local bonding environments are characterized in terms of group symmetries, but this has limitations in complex systems. Linearly polarized emission of x-rays from these 3d materials can provide information about local anisotropy, and valence-to-core (VtC) x-ray emission spectroscopy (XES) is especially sensitive to oxidation state, ligand environment, and bond length. The purpose of this project is to use local geometric and electronic information to formulate a measure of local anisotropy. This metric is evaluated against real-space Green's function calculations of linearly polarized XES, where we apply a supervised machine learning approach trained on this metric to predict differences in the polarized spectral shapes. Polarized spectroscopy techniques are critical for a wide range of applications including the development of microelectronics, nanostructure characterization, analyzing anisotropy within quantum dots, and studying the polarization sensitivity of non-linear optics. An accurate formulation of this continuous anisotropy parameter will provide researchers with quick and inexpensive computational insight. For the development of new functional materials, this metric can be used for searching databases efficiently, allowing researchers to select the candidates that will provide a more ideal signal of any polarization dependent properties.

Preterm birth is a leading cause of under-5 morbidity and mortality. No treatments exist to address the neurological complications of premature birth, which include loss of oligodendrocytes and activation of microglia, leading to white matter injury and inflammation, respectively. Our study explored repurposing azithromycin, an FDA-approved antibiotic with anti-inflammatory properties, to mitigate preterm brain injury caused by hypoxia-ischemia. We used a postnatal day (P)14 neonatal ferret model, equivalent to extremely preterm infants. We induced brain injury through a combination of inflammatory stimulus, bilateral carotid artery ligation, and oxygen fluctuations (hypoxia/hyperoxia). Ferrets were randomized into control, vehicle (saline)-treated, and azithromycin-treated groups. Littermate controls were not exposed to injury. Body weights and ex-vivo brain measurements (sulci and gyri widths) were recorded at P21, seven days after injury. Quantitative immunohistochemistry (qIHC) was performed to analyze microglia (Iba-1) and oligodendrocyte (Olig-2) density, and data were analyzed using Kruskal-Wallis tests. In our preliminary findings, post-surgical weights from the azithromycin-treated ferrets were similar to those of vehicle-treated animals. Azithromycin-treated ferrets also showed similar global microglia and oligodendrocyte staining compared to the vehicle group. The vehicle group had lower summed gyri measurements than controls (p=0.04), while azithromycin-treated ferrets had more similar gyri widths to controls (p=0.21). We will continue investigating microglial and oligodendrocyte density using qIHC across additional brain regions using pathology software (VisioPharm), including subregions of each gyrus (cortex, subcortical white matter, and coronal radiata), corpus callosum, hippocampus, and upper and lower thalamus. This will allow us to identify the brain regions most impacted by the injury and investigate if there are regional neuroprotective responses to azithromycin. By deepening our understanding of preterm brain injury and azithromycin-mediated neuroprotection, these findings could lay the groundwork for advancing azithromycin toward clinical trials, offering new hope for saving the lives of the tiniest neonates.

Traumatic brain injury (TBI), characterized by a physical impact to the skull, is a significant health concern among veterans, athletes, and the elderly, with over 200,000 TBI-related hospitalizations in 2020. TBI causes shearing forces and physical damage to the brain, resulting in increased risk of neurodegeneration and mental health problems. When they expect an impact, humans brace, exhaling against a closed airway in what is known as a Valsalva maneuver. This prevents venous return from the head, pressurizes the vascular network in the brain, and increases intracranial pressure (ICP) in a way that may protect the brain from TBI. We aim to mimic a Valsalva-like response (VLR) through external abdominal stimulation and measure corresponding ICP changes. First, we performed a 3mm-wide craniotomy in anesthetized ferrets and implanted a pressure transducer inside the brain to collect baseline pressure readings. After skull closure, VLR was performed both supine and upright (body at 45°), either physically (pVLR, 80-120mmHg by abdominal compression using a blood pressure cuff, n=4) or electrically (eVLR, bilateral 25-30mA stimulus of the rectus muscles, n=4). pVLR resulted in a 2-4mmHg increase in ICP over 2-5 sec. By comparison eVLR resulted in a larger and faster ICP increase - 3-7mmHg with an onset of 250-750ms. Consequently, we will utilize eVLR to modulate ICP in a TBI model to determine whether it is neuroprotective. Ferrets will be assigned to control or randomized to receive a TBI impact with either sham eVLR or eVLR. Animals will be subjected to baseline (pre-TBI), acute, and long-term behavioral testing. Additionally, we will perform brain cell specific histological staining. Results from behavioral testing and histology will inform us of the potential neuroprotective effects of eVLR against TBI and provide future direction towards translating the findings into a wearable device for at-risk individuals.

Age-related macular degeneration arises from irreversible photoreceptor loss. Photoreceptors, rods and cones, are specialized cells in the retina that allow light and color detection. My project investigates the role of retinoic acid (RA) on cone and cone-opsin development to understand the timeline of cone specification and development. RA, an endogenously synthesized vitamin A derivative present in the retina during development, drives rod photoreceptor differentiation, but its effect on cone development is still unknown. To understand RA’s role in opsin development, I use a retinosphere (RS) model, an in vitro system to culture human fetal retina. More specifically, I used RS from 70 to 90 days old (D70-D90) and cultured the RS until D100, when the cone-opsin onset occurs. I then fixed, cryosectioned, and immunostained the two conditions for S-opsin, M/L opsin, and NR2E3 (rod marker) and investigated changes in the density of cone opsin-positive cells between the two conditions using confocal microscopy. My findings showed that the condition containing exogenous RA had a decreased density of opsin-positive cells. To confirm that the observed effect is due to RA, I mimic the experiment by instead using WIN18446, an RA inhibitor. I then determined if RA's effects are dose-dependent. My results showed that increasing the concentration of exogenous RA amplified my previous findings. The next step is to understand the timeline of cone specification and development by using RS of a younger age, before cone-opsin onset. These results will allow my mentors and me to use our knowledge about RA to determine if inhibiting endogenous RA synthesis in the retina will play a role in developing therapeutics involving cone regeneration to aid in cone-related macular diseases and injuries.

The aim of this study is to identify possible differences in Anisakis spp., or “sushi worm”, infection intensities between three different species of wild Alaskan salmon, O. keta (chum), O. nerka (sockeye), and O. gorbuscha (pink), by examining canned salmon samples from three different canning regions from the 2024 season. Species were selected from three different canning regions: Kodiak, Prince William Sound, and Southeast Alaska. Distribution of these parasites among species and location have marine ecology (pinniped health and distribution), salmon biology (physiological and biochemical parasite defense, dietary preferences), and seafood safety implications (marketing, establishing safe food handling protocols) that make it important to establish a baseline dataset.

The retina is a layer of neurons on the back of the eye that sense light and relay visual information to the brain. Our goal is to understand the role of epigenetic repression in retinal cell development by focusing on the polycomb complex, a complex of many proteins that repress gene expression through deposition of the H3K27me3 mark on histones. The goal of my project is to learn how the polycomb complex influences retinal development by altering specific aspects of the complex’s activity and observing how these alterations influence cell fate, using two complementary model systems: fetal-derived retinospheres and stem cell-derived retinal organoids. To perturb different aspects of the polycomb complex, I have treated retinospheres with Gskj4, a UTX inhibitor, and BRM014, a BAF inhibitor. During development, UTX is responsible for removing H3K27me3 so genes that are silenced can be expressed. When I added Gsjk4 to 135-day old retinospheres, I observed that cell proliferation decreased, and more cells expressed the marker OTX2, indicating an upregulation of either bipolar or photoreceptor cell differentiation. These data indicate that H3K27me3 removal is critical for proper specification of retinal cell types. BRM014 inhibits BAF, an ATP-dependent chromatin remodeler that has been shown to be recruited by UTX to remove nucleosomes and initiate transcription. When I added BRM014 to day 135 retinospheres, I also observed an increase in the expression of OTX2, similarly indicating an upregulation of either bipolar or photoreceptor cell differentiation. From these experiments, we conclude that removal of H3K27me3 is necessary for proper retinal cell specification and development. A better understanding of epigenetic regulation during retinal development will allow us to develop therapies to regenerate damaged retina lost in blinding diseases and restore sight to patients.

Hypoxic-ischemic encephalopathy (HIE) is a leading cause of neonatal morbidity and mortality worldwide. The ferret provides a highly translational model to investigate HIE; the gyrified ferret brain has a similar grey-to-white matter ratio to humans, allowing for better assessment of white matter injury and impairment of cortical development compared to rodents. Our previous work has suggested that ferret brains also show greater resilience to hypoxia-ischemia (HI) than rats. Ferrets tolerate exposure to much longer and more significant HI, and 100-fold larger doses of inflammatory stimuli, than rats do. We seek to identify signatures of the ferret's protective mechanisms by comparing differentially regulated genetic pathways in the ferret versus the rat when exposed to identical insults. Whole-hemisphere organotypic brain slices were obtained from term-equivalent ferrets and rats and cultured for 72 hours. Slices were randomly assigned to control or oxygen-glucose deprivation (OGD), an in-vitro model of HIE. Cytotoxicity was assessed by lactate dehydrogenase (LDH) release, while global transcriptomics were analyzed via a 770-gene digital transcriptomics panel. Preliminary results show significantly lower LDH release in ferrets compared to rats, reaffirming the ferrets' resilience to OGD. We identified 90 differentially expressed genes in ferrets following OGD, and 11 genes in the rat. Ferrets upregulated CCL2 and LGALS, genes associated with inflammatory responses, and downregulated ADRB1 and NOS2, suggesting reduced oxidative stress. Rats downregulated KIR3DL1/2 and TGM1, which suppress natural killer cells and form the cell envelope, respectively. The experiment will be repeated with double the sample size and region-specific analysis of gene regulation. We hypothesize the ferret will display lower injury markers globally, which will be associated with regional differences in gene expression compared to the rat. We hope this will enable us to identify potential treatment targets for infants with HIE that can increase resilience and repair after injury. 

Hypoxic Ischemic Encephalopathy (HIE) is a brain injury caused by a lack of oxygen and blood flow in the peripartum period. Cardiac dysfunction occurs in up to 80% of infants with HIE and is associated with worse neurodevelopmental outcomes. The current standard of care for HIE is whole body therapeutic hypothermia (TH). The expected physiologic response to TH is a decrease in cardiac output by 10%, and heartrate (HR) by 10bpm, per 1-degree Celsius decrease in body temperature. However, neonates with cardiac dysfunction tend to have normal or elevated HR to compensate for decreased cardiac output. Therefore, normal or elevated HR during TH may indicate compromised cardiac function. We hypothesize that in neonates with HIE, HR trends during TH reflect cardiac function, and a sustained HR above 100bpm is indicative of cardiac dysfunction. Using echocardiograms performed within the first 2 days after birth in babies with HIE treated with TH at the Seattle Children's neonatal intensive care unit (NICU; n=19), we categorized neonates by cardiac function: normal, right ventricular (RV) dysfunction, and RV plus left ventricular (LV) dysfunction. We then extracted continuous HR data and compared median HR during TH across groups using linear regression during specific periods: 12-24h, 24-36h, and 36-48h after birth. Results showed that infants with RV+LV dysfunction had a higher HR than those with RV dysfunction only or normal function. Across all time periods, infants with any kind of cardiac dysfunction had an average HR above 100bpm, while those without dysfunction had average HRs less than 100bpm. Therefore, it appears that HR can be utilized as a proxy for cardiac dysfunction in neonates with HIE. Utilizing HR as screening biomarker for cardiac dysfunction may allow improve optimal resource utilization of echocardiograms as well as real-time, cost-effective monitoring and targeted treatment initiation. 

Planktonic protists (unicellular eukaryotes) play essential roles in open-ocean biogeochemical cycles and food webs, functioning as phototrophs, heterotrophs, or mixotrophs depending on the species. However, cultured representatives of protists from the Pacific Ocean are scarce, limiting our understanding of protists within the largest ocean on Earth. In this study, we analyze seven cultured protist strains isolated from the tropical Pacific Ocean from the upper ocean from 30 °N to 4 °S and from 120 to 140 °W, including seven haptophytes, five pelagophytes, and four dinoflagellates. We examine transcriptomes from laboratory cultures of these isolates. We construct a phylogenetic tree of the isolates based on single-copy marker genes to infer evolutionary relationships. We examine correlations between phylogenetic relatedness and the latitude and depth of isolation. An additional objective of this work is to resolve the species-/strain-level taxonomy of these isolates, enabling their integration into the Marine Functional Eukaryotic Reference Taxa database. This will improve our ability to characterize marine protist diversity and function in metagenomes and -transcriptomes.

Developing and validating new methods of enumerating species of concern is important for many conservation and management goals. Environmental DNA (eDNA) has shown potential to be a viable tool for obtaining non-invasive and cost-effective estimates of many organisms, including fishes in streams such as salmon. However, before eDNA can be used beyond an experimental basis, we need to understand how eDNA flows through small streams where salmon may spawn. This study aims to examine how sockeye salmon (Oncorhynchus nerka) eDNA is transported in small streams by collecting samples while ascending two morphologically unique streams. eDNA within each reach was analyzed against two measures: the abundance of salmon within each reach and cumulative abundance salmon above of the reach. Preliminary analysis suggests that eDNA is effectively transported to stream mouths when salmon are in high abundance. Moreover, eDNA does not accurately predict the abundance of salmon within individual reaches but corresponds more closely with the cumulative abundance of salmon above each reach, particularly when salmon are highly abundant. This closer alignment with cumulative salmon abundance is likely due to the cumulative nature of eDNA within streams.

In the face of global climate change, there is growing interest in growing seaweed and sinking it to depths to remove carbon dioxide. However, quantifying the carbon sequestration potential of such ventures is challenging. One key consideration is that rising seawater temperatures may increase the rate of kelp decomposition, thereby reducing the export of carbon-containing tissue to the seafloor. To assess whether blades of the bull kelp Nereocystis luetkeana decompose more rapidly in warmer water, twelve 35 mm-diameter tissue disks were allowed to decay at 10-12 °C (ambient temperature treatment) and another 12 tissue disks were allowed to decay at 17-19 °C (elevated temperature treatment). After 7 days, the mean change in disk mass for the ambient temperature treatment was compared to the mean change in mass for the elevated temperature treatment. Samples at elevated temperatures were visibly flimsier and more diaphanous, which was correlated with a significantly greater decrease in weight. In tandem with other studies, this finding suggests that brown algae may decompose more rapidly at elevated temperatures, which has important implications for how to maximize future macroalgal carbon sequestration as ocean temperatures rise.

Retinal cell degeneration is one of the leading causes of blindness and vision loss caused by retinal diseases and is irreversible in humans. However, regeneration of retinal cells occurs after injury in some non-mammalian vertebrates and mimicking these strategies in humans could evolve treatment options for the visually impaired. Previous research in the Reh lab discovered a way to generate new neurons by reprogramming Müller glia (MG), a support cell of the retina, through overexpression of the proneural Ascl1 transcription factor in the mouse retina. To stimulate reprogramming, we used a lentiviral construct with a glial specific promoter (HES1) to drive the expression of ASCL1. However, HES1 represses its own expression by binding specific DNA sequences called N boxes which regulate gene transcription and expression, thus creating a negative feedback loop. In order to limit the negative feedback loop, we designed two new constructs using the HES1 promoter with modifications to the N box sequences. While the current construct has a reprogramming efficiency of approximately 25 percent, the aim of my project is to use constructs with modified N boxes to increase the ratio of MG reprogramming into neurons and verify specificity of the new constructs to MG cells. My research with mouse MG has shown that constructs with N box modifications significantly increase Ascl1 expression as compared to the construct with no modifications. These results seem promising and if reproducible, I will proceed with applying this strategy to human MG by using an in vitro culture system of retinal organoids.