School of Science Department of Chemistry 9 Developing Deep Learning Enabled Nucleic Acid Structural Generator Supervisor: SU Haibin / CHEM Student: LAW Yu Ming / CHEM Course: UROP 1100, Summer We incorporate π–stacking, Hoogsteen and other non-Watson–Crick contact information into our previous deep-learning nucleic-acid structure generator by replacing the current G4 matrix with a comprehensive residue–residue interaction map. A standard matrix was first built manually from high-resolution RNA G4 structures, then an automated Python pipeline was developed that uses geometric cut-offs to detect hydrogen bonds and π–stacking from PDB files. The automated model accurately identifies π–stacking but over-assigns Watson–Crick pairs and misses some non-canonical bonds. The result provide a new pathway for further improving the current model. Developing Deep Learning Enabled Nucleic Acid Structural Generator Supervisor: SU Haibin / CHEM Student: LE Quang Truong / CHEM-IRE Course: UROP 1000, Summer Metabolic engineering for enhanced biofuel production requires systematic understanding of cellular metabolism and identification of optimal engineering strategies. Constraint-based modeling provides a powerful framework for analyzing genome-scale metabolic networks and predicting the effects of genetic modifications on cellular phenotypes. In this study, we aim to fill the gaps in literature concerning whole cell simulation to enable biofuel production. We employed the COBRA Toolbox in MATLAB to perform comprehensive metabolic modeling including flux balance analysis (FBA), flux variability analysis (FVA), and gene essentiality analysis on the E. coli core metabolic model. Our methodology focuses on reproducing and validating results from multiple published studies by comparing model predictions with experimental data across diverse carbon sources and genetic perturbations. Through literature review and method reproduction, we evaluated various modeling approaches to identify the most effective strategies for whole cell simulation in biofuel production contexts. The validation framework enabled us to assess model accuracy using metrics and identify key sources of prediction discrepancies, including cofactor availability and geneprotein-reaction mapping issues. This work contributes to improving the reliability. Developing Deep Learning Enabled Nucleic Acid Structural Generator Supervisor: SU Haibin / CHEM Student: TRUONG Bao Ngoc / CHEM Course: UROP 1000, Summer Efficient utilization of xylose, a major component of lignocellulosic biomass, remains a challenge in yeastbased biomanufacturing. This study introduces a modular metabolic engineering strategy to enhance xylose assimilation in Saccharomyces cerevisiae by deregulating central carbon metabolism. Five distinct modules—promoter tuning, transcription factor manipulation, biosensor-guided regulation, heterologous enzyme expression, and mutant enzyme deployment—were integrated to redirect carbon flux toward acetyl-CoA. The engineered strains demonstrated a 4.7-fold increase in 3-hydroxypropionic acid (3HP) production, showcasing the potential of modular design in optimizing non-glucose feedstock conversion. This work provides a scalable framework for sustainable biochemical production using renewable carbon sources.
RkJQdWJsaXNoZXIy NDk5Njg=