Baneyx, FrançoisCai, Yifeng2025-08-012025-08-012025Cai_washington_0250E_28041.pdfhttps://hdl.handle.net/1773/53442Thesis (Ph.D.)--University of Washington, 2025Controlling the stimuli-responsive assembly and disassembly of inorganic nanoparticles under mild conditions and with precision and predictability remains a key challenge in the design of next-generation nanomaterials. This dissertation describes experimental, theoretical, and machine learning approaches for programmable nanoparticle assembly using solid-binding proteins, a class of genetically encoded, high information content building blocks. In a first series of studies, bifunctional silica-binding proteins were used to mediate the reversible clustering of silica nanoparticles (SiNPs) between a fully dispersed state at pH 8.5 and large aggregates at pH 7.5. A multiscale simulation framework incorporating input from classical colloidal theory and atomistic molecular dynamics (MD) calculations into a rigid body simulation energetically benchmarked by scattering experiments successfully predicted assembly behavior and cluster size under various conditions of pH and electrolyte concentrations. Changes in the sequence and location of the silica-binding segments, and the introduction of steric forces using PEGylated monofunctional silica-binding proteins were further used to modulate SiNP assembly and achieve fine control over cluster size, polydispersity, and Förster Resonance Energy Transfer (FRET) between fluorescent protein blocks and dye-loaded SiNPs. A second set of studies extended the approach to thermoresponsive systems. Here, elastin-like polypeptides (ELPs) modified with a C-terminus cysteine (Cys) residue were used to decorate gold nanoparticles (AuNPs) and to demonstrate that the size of the clusters obtained upon liquid-liquid phase separation could be controlled through both the amount of free ELPs added to the solution and the diameter of the constituent AuNPs. The reversible nature of the assemblies was exploited for precision loading and release of small molecular cargos such as Nile Red and tetracycline. Additionally, we used Click chemistry to couple a Cys-modified ELP to an oligonucleotide and control the assembly-disassembly behavior of AuNPs over specific windows of temperatures. Finally, a dual variational autoencoder machine learning strategy was deployed to predict cluster size as a function of temperature from UV-visible spectra acquired by hyperspectral microscopy. Together, these studies lay the foundation for versatile strategies for nanoparticle assembly, combining genetic programmability, responsive behavior, and predictive modeling to advance the development of functional hybrid nanomaterials.application/pdfen-USnoneNanoparticle assemblySolid-binding proteinsStimuli-responsive systemNanotechnologyNanoscienceChemical engineeringChemical engineeringThesis