Browsing by Author "Thyer, Ross"

Now showing 1 - 4 of 4
  • Thumbnail Image
    Item
    DLPacker: Deep learning for prediction of amino acid side chain conformations in proteins
    (Wiley, 2022) Misiura, Mikita; Shroff, Raghav; Thyer, Ross; Kolomeisky, Anatoly B.; Center for Theoretical Biological Physics
    Prediction of side chain conformations of amino acids in proteins (also termed “packing”) is an important and challenging part of protein structure prediction with many interesting applications in protein design. A variety of methods for packing have been developed but more accurate ones are still needed. Machine learning (ML) methods have recently become a powerful tool for solving various problems in diverse areas of science, including structural biology. In this study, we evaluate the potential of deep neural networks (DNNs) for prediction of amino acid side chain conformations. We formulate the problem as image-to-image transformation and train a U-net style DNN to solve the problem. We show that our method outperforms other physics-based methods by a significant margin: reconstruction RMSDs for most amino acids are about 20% smaller compared to SCWRL4 and Rosetta Packer with RMSDs for bulky hydrophobic amino acids Phe, Tyr, and Trp being up to 50% smaller.
  • Thumbnail Image
    Item
    Engineering Optogenetic Control of Bacterial Metabolism in Stationary Phase of Growth
    (2023-09-06) Lazar, John Tyler; Tabor, Jeffrey J; Thyer, Ross
    Genetically-encoded sensors are used to induce metabolite production in bacterial fermentations. However, these sensors are typically optimized for exponential growth phase rather than stationary phase where the majority of metabolite production occurs. In the first portion of this work, we find that our exponential phase-optimized green light-activated E. coli two-component gene regulatory system CcaSR is effectively non-functional in stationary phase. We show that the major causes of failure are stationary-phase specific mutation of the plasmid-borne biosynthetic pathway used to produce the required chromophore phycocyanobilin (PCB) and accumulation of the response regulator CcaR leading to very high leaky target gene expression. To address these problems, we move the PCB biosynthetic pathway into the chromosome and re-optimize expression of the component enzymes, and re-balance CcaR expression for stationary phase . The resulting CcaSRstat system exhibits low levels of leakiness and an 80-fold activation of target gene expression in stationary phase. Notably, our stationary phase-optimized CcaSRstat system is not functional in exponential phase, a feature that may have benefits for metabolic engineering and other applications. In the second portion of this work, we combine CcaSRstat with static and periodic illumination patterns to achieve high levels of production of several industrially-relevant phenylpropanoid metabolites such as p-coumaric acid and bisdemethoxycurcumin. We then proceed to demonstrate that our optimal light signals at the 0.5 mL light plate apparatus (LPA) volume scale to 25 mL optogenetic bioreactors. CcaSRstat is a useful tool for optimizing bacterial metabolite production and could be used to control bacterial behaviors in other non-growth environments such as the gastrointestinal tract or soil. Our work lays the foundation for increasing the exploration space of dynamic control of metabolic pathways while also providing valuable insight into design considerations of biosensors in the stationary phase of growth.
  • Thumbnail Image
    Item
    Mutational Profiling of the Adeno-Associated Virus Rep Protein for Gene Therapy Production
    (2024-10-24) Azim, Tasfia; Silberg, Jonathan; Thyer, Ross
    Adeno-associated virus (AAV), a federally approved gene therapy vector that is currently in clinical trials for hundreds of diseases, often presents suboptimal characteristics for therapeutic applications, such as suboptimal tissue specificity, limited cargo size, undesired immune responses, and costly manufacturing. To improve this gene therapy vector, AAV proteins are frequently studied via rational design and combinatorial engineering. While the latter approach increases the sequence space that can be explored, it also presents unique challenges such as genotype-phenotype mismatches, noise arising from mutational errors in cloning, and bias arising in amplicon preparation and sequencing. In AAV, Rep proteins mediate DNA packaging and virus assembly, suggesting that changes in Rep activity, expression, or DNA binding might affect genome packaging. However, these proteins are not as well-understood as the proteins that make up the virus shell. I sought to understand how mutations in the Rep protein affect activity by selecting a library of Rep mutants for their ability to produce virions. To do this, I designed large protein libraries to examine a broad sequence space, I designed a selection strategy that couples genotype-phenotype characteristics of the mutants, and I designed a single-stranded DNA isolation workflow that enabled me to sequence winning variants in deep sequencing. By sequencing the rep gene following the purification of viruses that package AAV genomes, I identified Rep mutants having non-synonymous mutations with a range of cellular activities. Surprisingly, synonymous mutations within the p19 promoter were enriched to the greatest extent, increasing in abundance by 102 to 104-fold. When the most highly enriched mutant was used to package a synthetic DNA cargo into the AAV capsid, the packaging efficiency could not be differentiated from native Rep. These findings suggest that these synonymous mutations enhance AAV genome packaging into capsids by affecting Rep-genome interactions. They also suggest that silent sequence changes in the DNA cargo packaged by Rep can be used to tune packaging DNA packaging efficiency. Additionally, I designed a sequential cloning method for developing barcoded chimeric protein libraries, which enables easier analysis of deep-sequenced these. This cloning strategy has been partially validated. Future work should be done to optimize this cloning method, as it would be applicable for chimeric library design for any protein engineering experiment. Lastly, this work outlines considerations in high-throughput protein engineering experimental design. I hope this section of my thesis enlightens those who wish to begin high-throughput protein experiments and learn from some of the critiques I have of my own work.
  • Thumbnail Image
    ItemEmbargo
    New Tools and Cultivation Protocols for Autotrophic and Diazotrophic Non-model Bacteria
    (2024-07-29) Alameri, Abdulaziz; Thyer, Ross; Verduzco, Rafael; Kampouri, Stavroula
    Anthropogenic climate change necessitates a societal shift toward sustainable manufacturing practices. The chemical and polymer industry, which accounts for a fifth of the global CO2 emissions, has been slow to adapt due to factors such as economic pressures and risk aversion. Biomanufacturing, which usually relies on microorganisms such as bacteria, offers a promising alternative by leveraging native biological pathways that can perform complex chemistry without the requirement for extremes of temperature and pressure, often needed in traditional chemical processes. Beyond the model bacterium E. coli, the diverse metabolisms of environmental bacteria can be harnessed to produce a wide range of chemicals, with some bacteria capable of directly fixing atmospheric CO2 (autotrophs) and N2 (diazotrophs) when provided with a source of reducing power. Biomolecular engineering and ‘domestication’ of these non-model bacteria is required to make sustainable biomanufacturing a reality. However, these efforts are hindered by a lack of genetic tools. Here, I present efforts to (i) address the challenges of introducing recombinant DNA into environmental Mycobacteriales, an industrially relevant yet largely genetically intractable clade of bacteria, (ii) investigate the metabolism of these bacteria under conditions of extreme nutrient limitation and isolate novel CO2 fixing species, and (iii) develop a modular DNA assembly framework for the model diazotroph Azotobacter vinelandii. From this work, we have developed a suite of new genetic tools which function in several Mycobacteriales genera, characterized a novel Gordonia species with potential autotrophic metabolism, and identified the presence of multiple restriction-modification systems as a dominant barrier to genetic manipulation and future biomolecular engineering efforts in these species.