As diseases like COVID-19 become endemic, it becomes more apparent that access to low-cost, user-conducted tests with high sensitivity and rapid results are necessary to help reduce the spread of disease and mitigate burden on healthcare and laboratory infrastructure. While paper-based colorimetric tests attempt to fill this gap, they have reduced sensitivity when compared to “gold-standard” tests such as RT-qPCR, which typically exhibit results with fluorescent reporters. The goal of this project is to detect fluorophore activity using a smartphone camera and flash, with zero modifications (or as minimal as possible). As many people have smartphones and the ability to take a picture, but fewer possess lab skills or access to a lab, we aim to develop a smartphone-based system with the highest sensitivity and lowest barrier for entry, that is capable of detecting a fluorescent output. We are experimenting with both biological and technological levers, including combinations of time-delay using FRET and long-lasting fluorophores. On the software side, we are looking into whether we can leverage a smartphone’s built-in bayer filter to better delineate between emission wavelengths of fluorophores, and using timed flashing and recording methods to detect the biological time delay. So far, we have collected preliminary data on colorimetric readout reactions and shown that the difference between reactant and negative control is apparent at fairly low concentrations (250 uM). We expect that with a simple filter setup, we will be able to excite and detect the fluorescent output from a fluorophore. Further research will aim to simply things setup further, to reduce the number of external modifications required for use. When coupled with a diagnostic test, these ideas could potentially bring any test that can be coupled to a fluorescent readout from the lab to the user, increasing accessibility and lowering the costs of such tests.

Deep learning methods for protein sequence and structure generation have shown remarkable success in many design scenarios when combined with structure prediction networks such as AlphaFold2. Despite this advance, many design challenges such as de novo binder design still haven’t been fully solved. Diffusion-based models have demonstrated considerable success in image and language generation yet their application in protein design has not yet been fully explored. Recently, the development of a protein diffusion model called RoseTTAFold Diffusion (RFdiffusion) has shown significant success in protein design and enabled us to explore the challenging problem of designing protein binders. Here I demonstrate utilization of RFdiffusion towards generation of de novo binders to disordered major histocompatibility complex (MHC) peptides. Specifically, we took an MHC peptide from KrasG12D and used RFdiffusion to generate a diverse range of structures that can bind this peptide. To optimize the sequence of these structures we used ProteinMPNN. We used AlphaFold2 to predict the structures of these optimized binders in complex with the peptide and saw promising interaction metrics. Further, structure prediction of the designs in complex with Kras wild type (WT) peptide resulted in lower AlphaFold2 confidence metrics of the interaction occurring. This is a promising preliminary result that RFdiffusion can generate fully de novo MHC-mimics, which can differentiate between neoantigens and WT peptide. Many cancers are caused by a single point mutation such as KrasG12D, thus, designing protein binders with point mutant specificity is exciting as it allows for targeting of disease causing proteins over healthy WT proteins. 

Photovoice is a research methodology that uses photography as a tool for participants to express their thoughts on a topic of interest. Our project is specified towards marginalized communities and to give them a voice regarding issues each group faces. The main purpose of photovoice is to create change in our community by analyzing photographs provided by the participants.

Recent publications of photovoice projects at universities have yet to focus on a number of groups and their experiences on campus. We have decided to create a photovoice project specifically tailored to this area of interest.

This project focuses on four groups in the University of Washington Bothell campus: Students of Color, Muslim students, LGBTQ students, and first-generation students. 15 students participated in this study, with around 2 to 4 participants in each group. Our goal is to learn how these groups experience college/university differently compared to students outside of these groups. We asked participants to answer prompts that discuss how they experienced college academically, socially, and emotionally. Thus, we analyzed their experiences and formatted them into differing “codes” and “themes”.

Most themes in the project overlapped between the groups. The three notable themes were (1) The disconnect between faculty and students, (2) Expectations of what students should have and have access to, and (3) Struggling in finding ways to express oneself. Two more themes are to be created from the data we collected.

The project is still under development, and we are expecting to analyze these themes more in-depth. The goal of this project is to highlight disparities in the universities’ process of matriculation. Moreover, we hope to use this research to revise the assumptions and rules regarding resources and general student welfare to create more accessible resources and a streamlined transition from primary to secondary education.

Identifying risk factors for alcohol misuse among young adults is a critical public health priority given high rates of heavy drinking and alcohol-related consequences observed in this population. The field of behavioral economics has provided a set of quantifiable metrics that measure individuals’ demand for alcohol, which are important predictors of alcohol use, consequences, and response to treatment. Previous literature has also found that one’s self-reported drinking motives (e.g., drinking to cope with negative affect, to conform to peers, etc.) have important associations with drinking outcomes. Despite this literature, little is known regarding how one’s drinking motives relate to one’s demand. The study aims to investigate how different drinking motives may be differentially related to alcohol demand and whether birth sex moderates these relationships. The current study recruited 220 young adults (18-25 year-olds) from Washington state who report drinking at least twice a week and at least one recent heavy drinking episode (4+/5+ drinks for females/males). Participants completed online assessments that included the alcohol purchase task, which asked how many drinks they would hypothetically purchase and consume at various prices ranging from free to $20. Participants were also asked to report their birth sex and drinking motives (social, coping-anxiety, coping-depression, enhancement, conformity). I will conduct regression analyses to test for associations between drinking motives and alcohol demand, and to examine whether these associations are moderated by sex while controlling for age and discretionary spending. I hypothesize (1) stronger positive associations between coping motives and demand relative to other drinking motives, and (2) this relationship to be stronger for males. Results will improve our understanding of the relationship between drinking motives and demand between sexes and inform interventions focused on reducing alcohol misuse through alternate coping strategies or reducing demand.

Despite heavy exposure to Mycobacterium tuberculosis (Mtb), the bacteria that causes Tuberculosis (TB), some individuals show no evidence of infection and by defining these resistance mechanisms, we may identify novel treatment strategies. Among Mtb resistant individuals, our lab identified the Rab5a protein as differentially expressed as compared to controls with Mtb infection. By regulating vesicle trafficking, Rab proteins modulate a variety of cellular pathways including inflammatory signaling, antigen presentation, and autophagy, likely playing a role in Mtb clearance. We hypothesized that loss of Rab5a would alter IFN-êžµ gene expression. Monocyte-like THP-1 cells were electroporated with siRNA targetting Rab5a and yielded 70-90% knockdown at 24 hours versus scrambled siRNA control. Cells were then stimulated with DNA ligands for four hours before RNA analysis. Loss of Rab5a resulted in lower levels of IFN-êžµ gene expression after stimulation with Sheared Calf Thymus DNA (p=0.002, 53.9% reduction), Poly(I:C) (p=0.01, 42.8% reduction), supercoiled plasmid (p=0.03, 45.3% reduction), and cGAMP (p=0.008, 45.7% reduction). We conclude that Rab5a expression is required for Type I IFN production through the DNA-sensing pathway. By characterizing the pathways by which Rab5a modulates the macrophage Mtb response, we may identify host targets to augment protective responses that may serve as adjuncts to current TB treatments and vaccines.

Smartphone detection of anemia using patient photos has the potential to provide a non-invasive method of measuring hemoglobin levels, introducing the possibility of increasing the accessibility and cost-effectiveness of current practices. While traditional methods of anemia detection require a complete blood count by a trained healthcare professional, smartphone detection instead relies on the user to take a high quality picture of their fingernails. However, it currently lacks the ability to provide feedback to the user on the quality of their image. For example, an overexposed image or one with low fingernail visibility can lead to inaccurate predictions of hemoglobin levels. We propose that machine learning classification methods can analyze these patient images to estimate the image quality and predict the effectiveness of smartphone detection of anemia for a given image. With various classical machine learning models, we demonstrate and compare the capabilities of each in classifying images of patients’ hands as being of “good” or “bad” quality (or on a more granular numerical scale) when given features of the images. Preliminary results show that a logistic regression model reaches 91.4% accuracy labeling images when compared to empirically assigned labels, and we expect iterative models to achieve improved performance. When completed, we would propose that this classifier could be used in the field to identify if patient image is of high enough quality to produce an accurate measurement of hemoglobin levels in real-time, providing feedback on the phone to adjust or correct the image-taking process.

Neutrinos are fundamental particles involved in many important universal processes; however, because they only interact via the weak force and gravity, reliably detecting neutrinos directly is notoriously difficult. A new strategy is to study neutrinos through interactions with enhanced cross-section, like coherent elastic neutrino-nucleus scattering (CEvNS), in which the neutrino interacts with the nucleus as a whole (coherent) while conserving kinetic energy (elastic). However, due to the low energy of nuclear recoil in CEvNS, not all nuclei can produce detectable recoil if the recoil energy is on the order of the noise fluctuations in other background radiative processes, as recoil energies become indiscernible. Sodium iodide (NaI) is a candidate for detectable recoil, and I am characterizing the background spectrum of NaI crystals to see if NaI has low enough rates of background processes to be used in CEvNS studies. I analyzed previously collected NaI background spectra to calibrate the event energies and to perform a waveform analysis to distinguish physics pulses from electronics noise. The resulting spectra are used to determine the background rates in the crystals. These routines were converted into scripts to automate the same analysis for future data. Measuring the energy of CEvNS nuclear recoil can help characterize neutrino-quark interactions, which the coherent nature of CEvNS amplifies, providing unprecedented sensitivity to searches for non-standard interactions between neutrinos and matter. Improved characterization of CEvNS also allows for novel checks of predictions made by the Standard Model of particle physics, and has broader applications in understanding supernovae (which produce large quantities of neutrinos) and searches of dark matter candidates that may interact with neutrinos.