Methods in Computational Protein Design

Abstract

In silico design of protein has generated enormous interest with the rapid advances in computational power. Biological systems are known for their complexity, and we have made a series of computational developments that allow us to perform computational protein design. In this work we present a methodology for the design and prediction of protein active sites. We begin by presenting SCREAM, a program developed to accurately position sidechains in proteins. We show how using an improved scoring function and placement algorithm allow us to achieve better accuracy in the placement and prediction of sidechains in proteins compared to other methods. We then describe the development of an accurate treatment for describing hydrogen bonding. This is done by refining the hydrogen bond term in the force field DREIDING. We also need to properly describe electrostatics effects in proteins, and to this end, we introduce neutralized residues for proteins. We found that this improves the variance in our predictions dramatically. Finally, having established the components described above, we describe a protein design methodology encompassing the above methods and tools. We show predictions we made and those having been verified by experiments.