As a protein’s sequence directly impacts its function, being able to understand and map protein sequence to function is integral to developing improved proteins with applications in various biotech fields such as medicine, agriculture, and energy. Machine learning algorithms can use protein mutation datasets to learn the relationship between sequence and function. The purpose of our project is to investigate the hypothesis that information encoded by convolutional neural networks (CNNs), a structure-based machine learning model, trained on protein sequences predominantly learn on secondary structural features such as alpha helices and beta sheets rather than on other regions of the protein that lack defined secondary structures. In order to test this hypothesis, we trained a CNN using the mutation data of several different proteins with the goal of finding a correlation between secondary structure fraction and model accuracy. Furthermore, in the future we will train on a protein that the CNN traditionally performed poorly on using data limited to protein mutations inside secondary structural features. If our hypothesis is upheld, the accuracy of the model will improve although fewer data are used to train the model overall. Moreover, an upheld hypothesis will help answer a frequently raised question in the literature as to why graphical CNNs (GCNs), or CNNs that can consider additional information about the tertiary structure of a protein, do not typically perform better than sequence-only CNNs: if the model does primarily learn on secondary structures, information on tertiary structure would not be expected to aid performance. The results of this project will enhance our understanding of how models learn sequence-function mapping and help produce improved models to better design new proteins with more desirable properties.

CRISPR-mediated gene activation (CRISPRa) is an emerging tool for simultaneous regulation of multiple genes in genetic circuit designs. In CRISPRa, a nuclease-deficient dCas9 protein is used to deliver a transcriptional activator domain (MCP-SoxS) upstream of genes of interest. A complementary guide RNA (gRNA) directs dCas9 to bind the DNA target at a designated position. Such circuits are vital in regulating metabolic engineering, providing a more eco-friendly, industrial biosynthesis process and applications to medical therapeutics. Mathematical modeling of guide RNA (gRNA) competition for dCas9 predicted that CRISPRa efficiency will be diminished as the number of gRNAs increases up to 15. To experimentally investigate the limitations and rules for multiplexed CRISPR systems, we designed CRISPRa circuits with an increasing number of gRNAs in P. putida. The circuits were encoded in plasmids constructed with a scalable gRNA insertion technique based on Golden Gate Assembly. Multi-gRNA circuit performance was evaluated by simultaneously regulating multiple fluorescent proteins. Our preliminary results indicated that an increasing number of gRNAs lead to decreasing CRISPRa expression levels on our fluorescent genes. A clear drop in activation ratio was evident when introducing the 4th gRNA to our circuit, falling below the theoretical limit and establishing the limit to our current CRISPRa circuit. We hypothesize that the decrease in CRISPRa efficiency is due to limitations of dCas9 or activator protein. Therefore, by incorporating gRNAs lacking the protein recruitment motif, it is possible to identify whether the limiting factor is dCas9 or activator to provide a better understanding for CRISPR circuit designs. Optimizing multi-gRNA CRISPR circuits pose a potential for transferring to other microorganisms, expanding metabolic engineering capabilities and chemical productions beneficial in a wide range of biosynthetic applications. 

 Microglia, the resident immune cells in the brain, have multiple functions including synaptic pruning to preserve resources, phagocytosis of apoptotic cells, and isolation and removal of foreign material. Depending on local environmental stimuli, microglia can change their shape between multiple states including highly branched, branched, or ameboid. To better understand microglia responses to changes in the brain environment, I investigated morphological shape features that include changes in area, circularity, and aspect ratio among other important features. I specifically focused on the microglial response to oxygen-glucose deprivation (OGD). Oxygen-glucose deprivation is a condition where the brain fails to receive the necessary oxygen and nutrients for growth and maintenance, resulting in higher levels of stress and cytotoxicity. Investigating the effects of OGD on microglia is part of a larger effort - developing a fluorescent imaging pipeline called microFIBER. Our goal for microFIBER is to create an unbiased, detailed, and replicable analysis pipeline for the robust characterization of microglia morphology. Images are from a previous investigation into effects of OGD on neonatal rat brains in the Nance Lab. We used SciKit-Image along with other Python packages to segment, label, and quantify the geometry of fluorescent-labeled microglia cells in the images. SciKit-Image’s module RegionProps was used to quantify shape features by drawing certain properties over the objects and then measuring those drawings. I then analyzed the response of microglia in non-treated, 1.5-hour OGD exposure, and 3-hour OGD exposure via data analysis in Python and Excel. I further divided these treatment groups into regional comparisons of the cortex, hippocampus, and thalamus. Results from statistical analysis supported differences between treatment groups and brain region, including statistically relevant differences in microglial circularity, area, and axes lengths. Differences in shape features could be used in the future as markers for diseased or distressed conditions for medical diagnosis.

Life depends on a series of well-orchestrated biochemical reactions facilitated by proteins, which are differentially transcribed and activated in response to changing conditions. Hydrogels, water-swollen polymeric biomaterials, have proven useful as synthetic platforms to probe and direct biological activities by enabling researchers to recapitulate many aspects of the native cell environment. Though current hydrogel protein patterning techniques are capable of driving specific cell fates in individual cells in time and space (i.e. 4D), the timescales for patterning place dramatic limits on the types of biological functions that can be controlled. Furthermore, current techniques rely on slowly diffusing bioactive proteins into materials prior to immobilization within gels, so complete temporal control of protein activation within hydrogels remains out of reach. To address these limitations, my project focuses on directly photoactivating proteins within hydrogels using cytocompatible light. We predict that the extent of protein activation can be controlled dose-dependently by varying light exposure duration and intensity. We intend to use this platform to direct stem cell migration, differentiation, and proliferation in 4D on physiologically relevant timescales, which has tremendous utility in stem cell biology and regenerative medicine.

Perovskite materials have promising power conversion efficiency for use in efficient and inexpensive solar panels but are handicapped by material degradation. The degradation of methylammonium lead tri-iodide (MAPbI₃) occurs in the presence of heat, illumination, hydration, and oxygen. The fastest degradation is via the reaction with oxygen and water under illumination (water accelerated photo-oxidation, WPO). The exact mechanism for degradation with oxygen under illumination (dry photo-oxidation, DPO) is unknown, but the most commonly proposed reaction includes net water production, which would allow transitioning to accelerated degradation by WPO even without water initially present. Previous degradation experiments in our group have been conducted with perovskite thin films at steady state with flowing gas delivering reactants and removing products. The degradation products in the gas phase are challenging to collect in sufficient concentrations to characterize. We develop a novel experimental strategy to determine the stoichiometry of MAPbI₃ degradation focusing on the possibility of water production. We simulate atmospheric perovskite degradation using MAPbI₃ crystals suspended in o-dichlorobenzene with saturated dissolved oxygen and water for easy and accurate characterization of degradation products. The liquid samples are collected for analysis by gas chromatography-mass spectrometry and ultraviolet-visible spectroscopy measurements following degradation with heat, illumination, oxygen, water, and combinations of the four. These measurements determine the types and amounts of components produced, allowing for conclusions on the overall stoichiometry of DPO and WPO isolated from other, parallel processes. Importantly, we find that water is not produced from DPO and net consumed by WPO. These findings allow for developing novel degradation mechanisms that will lend essential understanding in engineering methods to counteract degradation and move perovskite solar panels towards full-scale viability.

Many synthesized chemicals in current use, based on carbon-intermediates, are byproducts from the crude oil refining industry. Due to the industry's highly efficient infrastructures along with cost-effective methods, it has been hard to think of any chemicals other than petrochemicals. However, with a recently growing desire for alternative carbon sources, more and more research has started to target the biomasses, which are potentially richer sources full of diverse intermediates. Such biomass-derived intermediates convertible to various chemicals are defined as "bioprivileged" by previous work in the field (Shanks, Broadbelt). Still, there is no apparent method to identify the bioprivileged molecules; the existing experimental approach is slow because only a small number of structures may be evaluated at a time. Thus, we use de novo molecular design models trained on a variety of commercially available chemical compounds databases, such as ZINC and PubChem, to generate a massive set of potentially bioprivileged molecules having a set of desired functional moieties. We measure the likelihood of a molecule being bioprivileged by the R0 and R1 reactivity indices, a recursive quantification of the number of reactive moieties used to estimate the reactivity of corresponding species. The reactivity indices serve as a target for optimization within a de novo design framework and allow us to generate candidate molecules that are enriched with the desired functional groups. This method may serve as the foundation for future work on discovering bioprivileged molecules including retrofitting synthetic pathways from biowaste precursors, optimizing reaction kinetics/selectivity through catalyst design, and analyzing the physiochemical properties of downstream products with respect to a set of desired functionalities. It may also be used in more general contexts to generate a set of molecular structures that are enriched with any set of desired substructures.

The viability of redox-flow (RF) batteries has, in recent years, become an increasingly prevalent point of interest in the chemical research community. RF batteries make use of the reversible electrochemical conversion of active redox species as a form of long-term energy storage. Currently, the most practical versions of these batteries utilize a vanadium-based solution, which is both costly and difficult to manufacture on a large scale. To solve this issue, researchers explored the possibility of using organic-based solutions and natural solvents. However, most of these batteries are limited to specific classes of organic molecules. Through the development of a generalized predictive model, we will create an accurate method of predicting the redox potential of a wide assortment of organic molecules which can be used to improve downstream generative AI algorithms for molecular design. To create our predictive model, we construct a set of experimental and computational redox potentials, which train our model. After compiling a database of roughly 100 organic molecules, we use our model to find correlations between the molecules’ measured redox potential and additional properties, which are calculated using various cheminformatics packages. We expect to find an approximate correlation within an acceptable range of error, which our model can base its predictions on. The limitations of our predictive model stem from our small sample size—larger data sets directly correlate to more accurate results. The successful development of a predictive model with a bounded range of error largely improves our ability to accurately find candidate molecules with high redox potentials, molecules which could potentially be used in large-scale redox flow battery systems.

Hepatitis C (HCV) is a liver disease caused by the bloodborne HCV virus. When left untreated, HCV can lead to cirrhosis and liver failure. Recent developments in therapeutics present a cure for HCV; however, treatment must be received soon after infection to be effective. Thus, limited availability of HCV testing creates a barrier to treatment distribution as chronic HCV is identified through a detectable viral load. Current HCV testing involves polymerase chain reaction (PCR) testing of blood samples, requiring a central laboratory and technicians to run them. The delay between appointments, sample transportation, running PCR, and receiving results can lead to lost contact with patients, making it difficult to connect them with timely treatment. The goal of the project is to develop a rapid point-of-care assay for HCV nucleic acid testing that allows healthcare providers to diagnose chronic HCV in 30 minutes and immediately prescribe treatments. We designed and validated an isothermal nucleic acid amplification assay for detecting HCV RNA: a two-step process involving reverse transcription of HCV RNA into complementary DNA (cDNA) which is detected by recombinase polymerase amplification (RPA). RPA is an isothermal process held at 40℃ with a runtime of 15 minutes, where a fluorometer collects data from the reaction. We compared the results of our RPA detection assay to the PCR-HCV assay used by the UW Clinical Virology Lab. We tested RNA from all six major genotypes using serum samples from Harborview Liver Clinic, where we had a limit-of-detection of 25 copies per reaction. We were able to match the results of the RPA and PCR assays with 100% agreement. By developing a streamlined detection assay for HCV, we will contribute to HCV testing without the need for expensive machinery or trained technicians, increasing the testing availability to increase HCV treatment rate and decrease HCV prevalence.

Cells in the body grow inside the extracellular matrix (ECM), which is composed of a combination of carbohydrates and proteins, presenting chemical and mechanical cues to the cells inside. Nearly all cell types are sensitive to the mechanics of the ECM and respond to cues such as stress, strain, and curvature, which influence organism development and disease progression. Hydrogel biomaterials are water-swollen polymer networks that mimic the properties of the ECM in vitro, allowing researchers to study cellular behavior in a controlled environment. In this project, we aim to develop a hydrogel platform where strain on the material, generated by contractile cells embedded within it, can be activated externally by a researcher in order to induce curvature in an engineered tissue, which we will use to investigate the effects of mechanical cues on cells encapsulated inside the hydrogel. We have synthesized a peptide crosslinker that acts as a two-input Boolean AND gate, with one half degradable by cell-secreted enzymes and the other half degradable by sortase, a researcher-added enzyme. We predict that when a cell-adhesive hydrogel is made with this crosslinker, contractile cells will be unable to expand until the addition of sortase; after sortase degrades one arm of the cyclic AND-type crosslinker, they will be able to locally degrade the hydrogel, spread within the gel, and then contract to generate stress and strain. We intend to encapsulate immature cardiac stem cells partway through differentiation, predicting that curvature alone will trigger further specification of these cells into their mature subtypes. Understanding the mechanism by which mechanical cues affect development will help identify new therapeutic targets for diseases where tissue curvature is important, and it will also inform new stimuli to improve the similarity of tissue grown in vitro to native tissue.