Hosseinzadeh, Parisa,,Fear, Karly2022-07-122022-07-122022https://hdl.handle.net/1794/27304Critically-sized bone defects experience delayed regeneration correlated to deficient protein signaling at the injury site. Delivering the potent growth factor bone morphogenetic protein-2 (BMP-2) to the defect is a promising clinical intervention to augment fracture healing in such cases. The release kinetics of BMP-2 from a hydrogel scaffold play a crucial role in the success of therapeutic delivery in terms of improved healing outcome. For instance, rapid BMP-2 release causes suboptimal bone formation. Affinity interactions between BMP-2 and binder proteins offer a mechanism to fine-tune the spatial placement and timing of exogenous BMP-2 bioactivity over the course of bone healing. Experimental selection techniques may identify such binder proteins from large protein libraries. However, experimental pipelines fail to reveal structural details of the protein-protein interaction and explore only a small subset of the possible sequence landscape for binding proteins. The computational design route is therefore an attractive strategy to rationally engineer a BMP-2 delivery control mechanism. In order to slow BMP-2 release, I first sought to directionally modify experimentally selected binding proteins with computational modeling and mutagenesis. Next, I generated de novo protein binders hypothesized to retain BMP-2 via affinity interactions when conjugated to the hydrogel scaffold. I approached this challenge with well-established computational protein design methodologies to yield binding proteins with target surface site-specificity. I then developed a machine learning model to predict binding affinity for computationally modeled proteins. Finally, I sought to experimentally characterize my designed binders using yeast surface display and flow cytometry to validate in vitro that my computational pipeline generates functional designs, appropriate for controlled BMP-2 delivery.en-USCC BY-NC-ND 4.0Protein engineeringProtein designFracture regenerationBone healingMachine LearningThesis/Dissertation0000-0002-4189-9919