Development and Applications of Computational Protein Design

Citation

Choi, Eun Jung (2006) Development and Applications of Computational Protein Design. Dissertation (Ph.D.), California Institute of Technology. doi:10.7907/ZNED-BV38. https://resolver.caltech.edu/CaltechETD:etd-06012006-103350

Abstract

Success in computational design of proteins requires a good understanding of the physical properties that determine the protein structure. However, computational protein design also provides us the opportunities to test and improve our current knowledge of protein structure. In our lab, the "protein design cycle" is used to improve the energy function and increase our knowledge of protein chemistry. This strategy utilizes a cyclic protocol where first, a modification that is predicted to improve the result is applied to the energy function. Next, proteins are simulated using this modified function. These molecules are then synthesized and analyzed to see if our simulation results correlate with the physical properties of the molecules. The new knowledge from the analysis is incorporated into the next design in the form of a new modification and the whole cycle starts again. Using this technique, an energy function highly successful in designing protein stability has been acquired in our lab.

A similar approach was used to determine whether the introduction of a backbone entropy term allows us to incorporate proline into the pool of amino acids we consider in designs. Using ORBIT, several proline mutants of protein G were simulated and synthesized. The correlation between the experimental results and computational results was analyzed. From this analysis, we learned that the entropic benefit of a proline mutation is usually small compared to the enthalpic disadvantages. However, including a backbone entropy term did increase the correlation between the rank order of the experimental stabilities and computational energies.

The ultimate test of our protein design protocol is its application to biological systems of interest. We have applied ORBIT to three different design problems, increasing the affinity of peptide-protein interaction of TRAF6, the elimination of disulfide bonds for biosensor design and enhancement of solubility and stability of an anti-htt antibody fragment. The results provided here show that protein design protocol is a fast and efficient way to manipulate and study biological systems.

Thesis Files