FURTHERING MULTISCALE MEMBRANE PROTEIN PREDICTION AND DESIGN APPROACHES
MetadataShow full item record
Over the past decade, there have been many advances in developing computational tools toward sub-Angstrom biomolecular structure prediction accuracy. A remaining challenge is capturing helix hinge dynamics within membrane proteins. Modeling these dynamics is challenging because their location and qualities are determined by a fine balance of intermolecular interactions with neighboring helices and the surrounding lipid bilayer. In this work, I aimed to enable sampling of kinked helices through the use of a kinked peptide fragment library. By first developing a classification method for kinked helices, I generated a kinked helix library. Exploration of this library revealed diverse helical representations which depended on the kink degree, resulting from the number of backbone hydrogen bonds present. I expect this library to allow for insertion of kinked protein fragments from the Protein Databank into membrane proteins. This library has the potential to significantly improve the accuracy of membrane protein structure prediction and enable de novo design of membrane proteins that contain flexible hinges. Protein-protein interface prediction and design methods provide insight into protein function and guide protein engineering. For membrane proteins, this task is especially difficult because they reside in a heterogeneous lipid bilayer. In this work, I develop a multiscale modeling approach to dock membrane-anchored proteins. CYP76AD1 and NCP1 are redox enzymes which interact to produce potent small molecules. This system is challenging to model due to its many complexities: (i) membrane-anchored proteins, (ii) 600-700 residue proteins, and (iii) small molecules. I used a combination of molecular dynamics, global docking and local docking to predict the protein-protein interface region. Through experimental validation, I determined an important residue involved within the interface. Furthermore, I applied the change in binding energy calculations to guide structural predictions. This multiscale approach has the potential to predict interface regions between large membrane-anchored proteins which have posed a challenge in the past.