Major Histocompatibility Class I (MHC) Molecules serve as a window into the cell, whereby T-cells can use their T-cell receptor to recognize foreign peptides presented on the MHC of a cell and induce apoptosis. Unfortunately, for diseases such as cancer, cancerous mutations may only be a single residue different from the native protein, resulting in similar mutant and wild-type peptide-MHCs. This small difference often results in negative selection of T-cell receptors that recognize mutant peptide-MHCs, leading to an absence of T-cells that can target cancer cells. This issue motivated us to use de novo protein design to generate binders with a high level of specificity between mutant and native peptide-MHCs. Using deep learning based protein design methods such as RFdiffusion and ProteinMPNN, we have generated promising in silico designs against a range of target peptides and MHC alleles. After selecting our top designs, we tested them using yeast surface display against our target MHC molecules with peptides loaded from KRas, PIK3CA, MAGE, and TP53 mutants. We observed binding events to all targets; some designs also had specificity to their respective mutant peptide-MHC over the wild-type peptide-MHC. After further work we have shown that the same design scaffold can bind to multiple peptide-MHC targets after slight redesign, similar to native T-cell receptors, holding promise that we could easily and quickly repurpose these scaffolds for new targets. Following these results we will incorporate our binders into T-cells as chimeric receptors and test for the ability of our binders to activate T-cell signaling and cell killing. This method of targeting peptide-MHC molecules is promising as a novel and rapid way to target cancer.

Convection in porous media within an inclined layer is relevant to a wide range of geophysical and engineering applications, e.g., in understanding large-scale convection in a geothermal reservoir. Previous work found that stable stationary localized convective structures are present at moderate Rayleigh numbers and a sufficiently large inclination angle when the boundary conditions are symmetric with respect to the layer midplane. In this project, I study the dynamics of traveling localized structures in inclined porous medium convection in the presence of asymmetric temperature boundary conditions. I conducted direct numerical simulations (DNS) of the fluid equations and found that one- and two-pulse structures exhibit a quadratic relationship between the travel speed of the structure and the symmetry breaking control parameter in the boundary conditions, while three- to five-pulse domain-filling structures display a linear relationship. With further simulations and increasing domain size, we discovered that, for sufficiently strong symmetry breaking, adjacent pulses repel each other while traveling and so tend to spread out, eventually becoming equidistant in the finite domain. The repulsion is sensitive to the travel speed (and thus to the asymmetric boundary conditions) and the domain size. I show that these interactions are associated with the spatial eigenvalues of the base flow that are responsible for the leading and trailing tails of the 1D along-slope temperature profile of the localized structures. These eigenvalues are complex implying that the tails oscillate while decaying exponentially. We employ the computed spatial eigenvalues to predict the tail profiles of traveling pulses and show that these successfully match observations from DNS. This comprehensive analysis enhances our understanding of the stability and bifurcations in the dynamics of traveling localized structures in inclined porous medium convection, offering valuable insights for geophysical and engineering applications.

VLA-4 is a surface protein of immune cells that plays an important role in their extravasation into tissues during an immune response. In multiple sclerosis (MS), pathogenic T cells enter the brain and attack nerve cells by using VLA-4 to bind VCAM-1, a cell adhesion molecule on endothelial cells that line blood vessels. Current MS treatments rely on antibodies that bind VLA-4 and block interaction with VCAM-1, preventing a pathogenic immune response. However, antibodies are expensive to manufacture, and their binding cannot be easily regulated to control drug-induced side effects. Aptamers are single-stranded DNA or RNA molecules that fold into sequence-defined structures capable of binding targets with affinities and specificities comparable to antibodies. Being chemically synthesized, they are much cheaper to manufacture and offer no batch-to-batch differences. Unlike antibodies, their binding in vivo is rapidly reversible, which could alleviate some side effects of disease treatments. However, aptamers have limitations in vivo – degradation by nucleases in serum, and rapid clearance into urine. This project designs and assesses modifications to a novel VLA-4 binding aptamer to improve in vivo function, with the goal of developing an alternative for MS treatment. We designed various modifications to the aptamer backbone to prevent nuclease degradation and conjugated the aptamer to a polymer to increase size and reduce clearance. To assess aptamer functionality, an in vitro model of T cell adhesion is used. VCAM-1 coated plates are used to simulate endothelial cells, and VLA-4+ T cells are incubated in the plates to allow adhesion in the presence of modified versions of the aptamer and serum. VLA-4 inhibition by our aptamer designs is assessed by characterizing the extent of cell adhesion inhibition. Successfully designing a modification that significantly improves the in vivo function of aptamers will have broad implications for their clinical translation to in vivo use.

Stimulator of Interferon Genes (STING) signaling contributes to tumor immunity. However, treatments targeting the STING pathway are limited by route of administration, insufficient STING activation, and off-target toxicity. We introduce poly-STING, a copolymerized, mannosylated variant of the diABZI STING agonist-3 known to activate the cGAS-STING signaling pathway, promoting the release of type-1 interferons and pro-inflammatory cytokines leading to tumor immunogenicity. The STING agonist-3 is a non-nucleotide molecule that activates the STING pathway, but it has poor solubility, which limits its usage in-vivo. The developed poly-STING platform improves the drug's solubility, is designed to target immune cells, and provides enzyme-triggered drug release upon delivery, which has been shown to induce improved therapeutic efficacy compared to the free drug. The Pun and Stayton labs seek to investigate modalities for optimization of the cGAS-STING pathway activation and characterize the mechanism of action. Specifically, my project will evaluate STING activation by observing macrophage repolarization from type M2, as the mannose from the poly-STING binds to the CD206 receptors on M2 macrophages. This activates the STING pathway, repolarizing the macrophage to pro-inflammatory type M1. To test effects in vitro, I will culture bone marrow-derived M2 macrophages with various formulations of poly-STING, and repolarization will be measured through flow cytometry and RT-qPCR to quantify expression of macrophage markers. We expect to find higher M1 activity in macrophages treated with poly-STING as opposed to the free drug. Next, I evaluate the therapeutic efficacy of the STING formulations through an in-vivo tumor reduction study using murine models of breast cancer and melanoma, expecting to find longer survival of mice treated with poly-STING. The culmination of this project will result in a polymer-based STING agonist delivery platform that solves the solubility and bioavailability issues associated with the STING-3 agonist, with enhanced efficacy and decreased toxicity after systemic administration.

Large language models (LMs) are prone to generate diverse factually incorrect statements, which are widely called hallucinations. Current approaches predominantly focus on coarse-grained automatic hallucination detection or editing, overlooking nuanced error levels. In this project, we propose a novel task—automatic fine-grained hallucination detection—and present a comprehensive taxonomy encompassing six hierarchically defined types of hallucination. To facilitate evaluation, we introduce a new benchmark that includes fine-grained human judgments on two LM outputs across various domains. To run this evaluation, I directly managed the collection of around 400 total human annotations which were analyzed to better understand the hallucinations present in LM outputs. My analysis using this benchmark reveals that ChatGPT and Llama2-Chat exhibit hallucinations in 60% and 75% of their outputs, respectively. A majority of these hallucinations fall into categories that have been underexplored in previous work. As an initial step to address this, I trained FAVA, a retrieval-augmented LM by carefully designing synthetic data generations to detect and correct fine-grained hallucinations. I set up the synthetic data generation pipeline to train FAVA which consists of prompting ChatGPT to noise a passage and insert errors one by one. The noisy passage is then post processed into our training erroneous input and edited output pairs. On our benchmark, our automatic and human evaluations show that FAVA significantly outperforms ChatGPT on fine-grained hallucination detection by a large margin though a large room for future improvement still exists. FAVA’s suggested edits also improve the factuality of LM-generated text, resulting in 5-10% FActScore improvements. These results further demonstrate the strong capabilities of FAVA in detecting factual errors in LM outputs.

The project aims to explore and experiment, through critical and creative geovisualization practice and pedagogy, innovative ways to embody, imagine, and represent the more nuanced social and spatial understanding of the Capitol Hill Organized Protest (CHOP). Specifically, we plan to implement qualitative forms of representation and analysis into the CHOP digital geographical archive as new data and layers. It involves analyzing individual graffiti, the messages, and patterns of speech, memories shared in the interactive map, as well as data collected from the field visit and interviews/conversations with the participants of the CHOP through the transformation of various modes of data and analysis that value representation in different ways. It draws on disruptive approaches from black digital geographies and critical and socially-relevant GIS to challenge how meaning is made from digital data. Yet, the project’s objective is not just to map these data and representation in/of the CHOP, but the opposite—remaking and even re-imagining through a process of humanistic sense of place-making. We engaged in multiple conversations with both direct and indirect stakeholders, fostering a comprehensive understanding of the diverse perspectives surrounding the CHOP. Direct stakeholders included individuals who had lived and experienced similar conflicts in Seattle’s history with the WTO protests, which provided invaluable insights that allowed for direct comparisons to the circumstances within the CHOP. Indirect stakeholders include individuals who contributed greatly by offering their perspectives on approaching potential challenges of the project. We put forward critical and creative geovisual processes that can draw connections between local community-based participatory and policy work, and how these diverse evolutions of critical GIS and geovisualization tell a more complex and nuanced story about the CHOP by relating to social, political, and cultural practices and as a way of engaged knowledge production.

Though Seattle’s reputation is for economic and technological prosperity, some Seattle high schools do not nurture students with the academic and professional rigor needed to partake in their local economy. This project examines the relationship between poverty and income in the City of Seattle and secondary education outcomes in Seattle Public Schools (SPS). The project puts forward the concept and practice of community mapping and GIS (Geographic Information Systems) for representation and visualization. Conducting a mixed-methods research using both qualitative and secondary research analyses, we uncover education narratives and outcomes in SPS while considering socioeconomic and political influences. SPS and state government online databases provided relevant information such as school graduation rate, college-readiness index, census tract poverty percentage, average annual income, and income inequality. We also conducted an online survey on SPS counselors, inquiring about their perception and input behind local socioeconomic contributions to the SPS curriculum and school life. In visualizing our research data, we utilized ArcGIS Pro, sectioning Seattle by SPS school zones and census tract, organizing our data by poverty percentage and income level through bivariate analysis, and juxtaposing them with education outcomes in SPS. Altogether, employing a social community investigation in conjunction with our mapping process allows for a deeper geospatial analysis that can be used as a tool to highlight socio-economic issues. In light of the socio-economic problems in Seattle, SPS counselors infer that other variables such as rich community sentiment, domestic and scholastic support, extracurricular involvement, and an overall meaningful life balance can contribute to successful education. Though our work establishes the relationship between high poverty percentage, low income, and poor education outcomes, we found that specific education metrics such as graduation rates do not necessarily entail success beyond secondary education. We proposed to use other metrics, such as the college readiness index, to infer education outcomes more accurately.

Traditionally known for its toxicity, hydrogen sulfide (H₂S) also possesses physiological roles as an endogenous gaseous signaling molecule in multiple biological processes. Previous research has demonstrated changes in levels of H₂S-producing enzymes during oxidative stress, hypoxia, and inflammation in various tissues including the liver and heart. H₂S protects cells from cytotoxicity in part by promoting the synthesis of glutathione, neutralizing reactive oxygen species, and inhibiting apoptosis signaling pathways. Thiol methyltransferases TMT1A and TMT1B can methylate endogenous H₂S to methanethiol. TMT1B has been shown to have a potential role in mediating the toxic effects of methanethiol. Gene silencing of TMT1B was found to significantly alleviate the observed cytotoxicity induced by methanethiol in human bronchial epithelial cells (16HBE). Methanethiol may induce harm to human respiratory tract cells, and understanding the mechanisms involved, including the role of TMT1B, could potentially lead to insights for mitigating these harmful effects. Doxorubicin (Dox) is a widely used chemotherapy drug for the treatment of various cancers but can induce oxidative stress in cells. In my preliminary experiments, I assessed the cell viability of HepG2 liver cells that were supplemented with various concentrations of NaSH (H₂S donor) and sodium methanethiolate (NaSMe, MeSH donor), followed by Dox treatment. Supplementing the cells with H₂S significantly increased cell viability in the presence of doxorubicin, while the methanethiol had no effect. The goal of my project is to identify H₂S-dependent protective pathways during cellular stress in HepG2 versus cardiomyocytes. My preliminary data indicates that both H₂S and its metabolite, methanethiol, may alter cellular responses following treatment of exogenous compounds that induce cellular stress such as CoCl₂, hydrogen peroxide and Dox. My goal is to pinpoint genes altered during the stress response. This understanding of H₂S-dependent pathways may pave the way for designing novel therapeutics that maintain or enhance H₂S levels.

The aim of this study was to analyze the three-dimensional deformational changes in the tongue base during natural chewing in a minipig model. Eight 7-8-month-old Yucatan minipigs were used in this study. Under anesthesia, eight 2mm ultrasonic piezoelectric crystals were implanted in the tongue base forming a cubic-shaped configuration, representing the right/left dorsal (RDL-LDL) and ventral (RVL-LVL) lengths, anterior/posterior dorsal (ADW-PDW) and ventral (AVW-PVW) widths, right/left anterior (RAT-LAT) and posterior (RPT-LPT) thicknesses. After the minipig was awakened after anesthesia withdrawal, unrestrained feeding was offered for 10-15 minutes. The amplitudes and onsets for each dimensional change of the crystal-circumscribed region were measured from the start of the jaw opening phase of chewing, and up to 21 concessive chewing cycles were measured. The phases of jaw opening-closing/power stroke during chewing were determined from the simultaneous electromyographic recordings. The duration measurements represented the chewing cycle lengths. The ADW was set up as the reference dimension for the onset calculation (zero point) due to its most stable nature during chewing. All measured dimensions showed either increased (peak/elongation) or decreased (valley/shortening) signals from the baseline. Overall, lengths (RDL-LDL and RVL-LVL) are either shortened or elongated depending upon a given chewing side. The widths (ADW-AVW and PDW-PVW) increased, and the RPT-LAT thickness increased while RAT-LPT thickness reduced during jaw opening of chewing. RDL showed the largest shortening (-42.28% valley-amplitude) while LDL showed the largest elongation (21.15% peak-amplitude, p ≤ 0.05). Earlier onsets occurred in shortening and later onsets in elongation (1.41% to 10.53%) in relation to the reference. Last, the duration of the chewing cycle was 0.45-0.58 seconds. The findings of this study showed a specific kinematic pattern of the tongue base in chewing. This will contribute to a better understanding of the biomechanics of the oropharyngeal function.

Acute pancreatitis is an abrupt and painful inflammation of the pancreas that is often associated with serious complications and high mortality rates. Current diagnostic and therapeutic options remain limited, underscoring the need for novel strategies based on the underlying mechanisms of the disease. Our previous research suggests a key role for activin A, a member of the Transforming Growth Factor β (TGF-β) superfamily, in modulating inflammatory responses. Notably, activin A appears to trigger the c-Jun N-terminal kinase (JNK) pathway, a crucial member in macrophage activation during inflammatory diseases. This study aims to explore the role of activin A and the pJNK signaling pathway in macrophages during the pathogenesis of acute pancreatitis. Our investigation involves cell culture experiments using the RAW264.7 macrophage cell line, qPCR and Western blot analysis to identify protein alterations in activin-stimulated macrophages, and transwell migration assays to examine macrophage’s migratory ability toward chemoattractant after activin stimulation. We predict that activin A-stimulated macrophages would have increased JNK activation, increasing macrophage infiltration in inflamed tissues and polarization toward the pro-inflammatory subtype, potentially exacerbating the severe inflammatory response of acute pancreatitis. Our study is expected to reveal novel pathogenic mechanisms of acute pancreatitis and pave the way for the development of more effective diagnostic and therapeutic strategies for this disease.

Open-top light sheet (OTLS) microscopy enables the volumetric imaging of large tissue specimens for research and potential clinical assays. No destructive sectioning is required, allowing the tissue to be used for standard downstream assays (e.g. H&E and molecular analyses) after the 3D pathology process is completed. While our typical 3D pathology workflow uses many of the same reagents as standard formalin-fixed paraffin-embedded (FFPE) histology, including xylene and ethanol along with a food-grade cinnamon oil (ethyl cinnamate), we would like to show that our processes do not negatively impact the quality of molecular biomarkers in valuable archived clinical specimens (FFPE). In previous research, we qualitatively demonstrated that tissue morphology and immunohistochemistry markers were unchanged before and after our 3D pathology workflow. Here, we aim to quantitatively assess the effects of our processing methods on FFPE breast carcinoma tissues using standard ER and HER2 immunohistochemistry (IHC) and HER2 fluorescent in situ hybridization (FISH) analyses, as well as nucleic acid integrity metrics (RIN scores, RNA bulk yield, signal quality). We hypothesize that the molecular characteristics of our processed specimens are statistically equivalent to those of unprocessed specimens. To demonstrate this, I have obtained two adjacent 3-mm diameter punch biopsies from 26 archived (FFPE) breast specimens. For each specimen, one sample will undergo our standard 3D pathology workflow. I will then submit both samples to pathology labs for quantitative comparison of standard clinical biomarkers (e.g. HER2 and ER expression) and nucleic acid integrity metrics. We will demonstrate that our lab’s 3D pathology protocols do not negatively impact tissues in terms of molecular characteristics, which will be important for clinicians to allow our nondestructive 3D pathology methods to be performed on valuable archived tissue specimens, and for our methods to more-easily translate into standard clinical practice.

Ultrasound elastography (USE) is an emerging technique used to measure in vivo tissue elasticity, or stiffness. The present study applies USE to examine tissue property of the tongue in a minipig animal model. The long-term goal of this research is to examine the association between tongue stiffness in relation to obesity and obstructive sleep apnea (OSA). Eight Yucatan minipigs, 7-8 months-old, were used for this study. USE data were collected for the caudal tongue from the submandibular region with sagittal orientation and for the left masseter muscle when the pig was under sedated sleep in the lateral position. The USE data are collected as video clips, and image frames are extracted and then analyzed with custom-developed software that converts image pixel colors to stiffness values. Multiple video clips are recorded at each site for each study. For each video clip the USE values in extracted ROIs are averaged over 10-20 frames, producing an average circular elastography ROI. The USE pixel values are then averaged over the video clips acquired in each region of the tongue and the left neck muscles. I have processed the tongue video clips to extract blocks of USE measurements that will be compiled to summarize the tissue stiffness and USE data quality for each animal. This procedure will also be applied to the masseter muscle images to produce reference tissue stiffness values for each animal. These data will provide a set of normal baseline stiffness values that can be compared to animals with obesity and OSA in ongoing research studies.

Spinal cord injury (SCI) is a prevalent human trauma that greatly reduces an affected individual’s quality of life. Natural healing post-SCI results in glial and fibrotic scarring, which are common bodily reactions to central nervous system (CNS) injury, but components within these scars unfortunately inhibit axonal regeneration and the sprouting of injured neurons. Interestingly, mammals of the genus Acomys have evolved mechanisms to overcome these deficits and regenerate CNS cells to full functionality post-SCI. Since adult mammalian neurons do have an intrinsic capacity to regenerate, we reasoned that neuron-extrinsic factors are the likely culprit for failed cell regeneration in SCI patients. Thus, the overall aim of our project is to characterize neuron-extrinsic factors such as immune cells and alterations in extracellular matrix (ECM) molecules that give the species Acomys cahirinus their regenerative capacity. We hypothesized that Acomys will exhibit greater axonal regeneration and less fibrotic scarring than the common mouse model, Mus musculus, after SCI. To study this, we produced a clinically relevant contusion SCI in both Mus and Acomys. We then examined major inflammatory cells known to be activated post-SCI. Standard immunohistochemistry targeting IBA1 (a microglia marker) and GFAP (an astrocyte marker) was used to detect: 1) microglia, the first-responders to injury in the CNS and 2) astrocytes, a major component of the glial scar. My analysis showed that more microglia, but less astrocytes, were activated in Acomys than Mus. This suggested that after SCI, Acomys activated less astrocytes but recruited more reparative immune cells compared to Mus. Next, we will examine the distribution of fibroblasts and collagen, important ECM components that compose fibrotic scars, in post-SCI Acomys and Mus tissue. Understanding how neuron-extrinsic factors respond to SCI in Acomys will help us further identify and define cellular targets for the development of novel therapeutics to treat human SCI.

 Alzheimer’s Disease (AD) has well-known brain alterations such as Tau protein build-up, beta-amyloid plaques, and neuronal cell death, yet the role of the brain’s microvasculature on the progression of this neurological condition has not been fully uncovered. My research examines changes in the microvasculature density, length, and function during normal aging using a well-established aging model in Brown Norway rats. This study contributes to the pantheon of previous microvascular research and forwards the field toward understanding AD development from another perspective. My hypothesis is that the density and length of these microvasculature are decreased in areas associated with learning and memory (i.e., the hippocampus and parietal cortex) before the development of AD symptoms and worsen as the disease progresses. To test this hypothesis, first in normal aging, 3 experimental groups of Brown Norway rats are employed: (n=6) young rats at 5-6 months, (n=6) middle-aged rats at 15 months, and (n=6) old rats at 20-24 months. Sagittal slices of the right hemisphere were fluorescently marked for their microvasculature, astrocytes, and cellular nuclei. The ImageJ analytical program was used to compartmentalize the areas of the dentate gyrus, CA1, CA2, and CA3 along with the parietal cortex into 900 x 900-pixel boxes for examination. The preliminary results show that the density of microvasculature within the 3 age groups were consistent while the distribution of the vessel lengths had more variability. The two leading postulates are increased tortuosity with increased age and/or rarefaction, where the microvasculature experience shortening with increased age. Further analysis is needed to examine this distribution among the 3 age groups.

Neurogenic bladder is a common condition associated with traumatic spinal cord injuries (SCIs), which results in inhibited detrusor function, bladder-sphincter dyssynergia, and scarring of the bladder walls and muscle. Care for neurogenic bladder is aimed at reducing abnormally high pressures, which left untreated lead to hypertrophy and tissue fibrosis of the bladder wall, as well as upper urinary tract complications. Current treatments target the neurotransmitter release of acetylcholine, utilizing anticholinergic drugs to counter the overactive bladder. However, these drugs can have negative/deleterious side effects, and have broad symptoms influencing unintended targets around the body. Targeted chemodenervation that uses Botulinum toxin (Botox) to interfere with nerve conduction is often reserved as a second line of defense to treat neurogenic bladder. Unfortunately, this late provision concedes irreversible damage to the detrusor muscle. We hypothesize that administering early chemodenervation can prevent the development of neurogenic bladder, improve bladder compliance post SCI, and increase the overall quality of life of SCI affected patients. Using a rat model, the Khaing lab administers a controlled contusion injury to the T8/T9 vertebrae, simulating a spinal cord injury in humans. The recovery of the rats is tracked with behavioral observations, cystometry data collection, and histological stains. My team and I are working to determine the effective therapeutic time post SCI for Botox injections, and the optimal doses for treatment. Our results thus far support that acute chemodenervation with Botox reduces bladder overactivity, and bladder wall thickness.

As we receive spatial and temporal information, our brain develops sequential patterns to store events, giving us the ability to learn and store relationships. For learning and memory, these rapidly-encoded sequences are reactivated as “replay” sequences after experiencing the original trajectory, as often observed in the hippocampus. This brings up the question: what biological mechanisms enable us to build, encode, and trigger these relationships and replays? The goal of this project is to model sequential replay in spiking neural networks to explore and understand various biological mechanisms that produce the acquisition of such sequences. We are using NEST Simulator, a spiking neural network simulator software, to model large-scale neural networks. Then, we explore how changing dynamics such as non-random structure of the network and interactions between excitatory and inhibitory cells can contribute to sequence generation, as well as the salience and speed of such sequences. We have observed the significance of interplay between particular parameters, such as the widths of spatial Gaussian distributions for neuron connection strengths, by analyzing generated spiking raster plots. Recent work has also suggested an important role of long-term potentiation of intrinsic excitability in sequential replays, which we are integrating with the aforementioned dynamics by building a unique synapse model within the simulation software. This is a novel method to introduce excitability in a network, which is important to determine how changing excitability through potentiation, rather than plasticity, facilitates network formation and propagation. This research is significant because it highlights the components of neural networks that could be crucial to quickly generating and maintaining sequences for learning and memory, therefore helping us understand the brain’s mechanisms for storing spatiotemporal relationships.