Advances in Protein Design: Conformational Switch, Multimeric, and Protein-DNA Design

The aim of protein design is to produce sequences that fold into a desired structure with improved or novel properties. Since the problem exhibits degeneracy, where many sequences can fold into the same structure, it is important to have design tools that can explore a large number of sequences. This thesis presents a series of computational protein design tools that expand the capabilities of quadratic assignment-like protein design methods to the design of conformational switches, multimeric systems, and protein-DNA binding. For the conformational switch design problem, an optimization model is introduced to design for sequences that change folds with a minimum number of mutations. Designed sequences are then computationally validated by a transition specificity metric that uses a detailed electrostatic energy function. This method is validated by an experimental test set and experimental results presented. Further, the detailed electrostatic energy function is shown to improve the accuracy of other validation metrics. For multimeric protein design, a molecular dynamics (MD) based procedure is presented for producing flexible templates for multimeric systems. These templates can be used in designing multimeric systems. The resulting sequences can be validated computationally using a multimeric fold specificity method and an MD-based approximate association affinity metric. This method was applied to the design of ultrasmall self-associating peptides, self-associating FG-repeat peptides, and CXCR4/CCR5 dual inhibitors. Experimental validation of the self-associating peptides and dual inhibitors are presented. For the protein-DNA design, a novel protein-DNA optimization model is introduced which accounts for both protein-protein and protein-DNA interactions. Resulting designed sequences are validated by fold specificity and a protein-DNA interaction energy metric. This method was applied to the design of a prototype foamy virus integrase for binding specificity and computational results presented. The integration of these methods into the automated Protein WISDOM framework is important to the wider academic community. The webtool is presented along with strategies for integrating the developed methods into the framework. An application of the protein design method to the design of methyltransferase inhibitors is presented. The methods introduced represent the expansion of the quadratic assignment-like protein design methods to a wider range of biologically relevant problems.