A novel approach to de novo protein structure prediction using knowledge based energy functions and experimental restraints

The focus of this work was to develop a method for rapid fitting of atomic resolution structural models into medium resolution electron density maps and Bayesian energy potentials for a de novo protein structure prediction method. The developed methods, BCL::EM-Fit, BCL::ScoreProtein and BCL::Fold, were benchmarked on large sets of proteins. All described work is implemented in the object oriented C++ software library termed "BioChemistry Library" (BCL), developed in Meiler Lab.;Chapter II is a reproduction of the first author paper "BCL::EM-Fit: Rapid fitting of atomic structures into density maps using geometric hashing and real space refinement" published in 2011 in Journal of Structural Biology [1]. Chapter III and Chapter IV are reproductions of co-first authored manuscripts titled "Knowledge based energy potentials for ranking protein models represented by idealized secondary structure elements" and " De novo prediction of complex and large protein topologies by assembly of secondary structure elements" respectively. Both of these manuscripts are currently in the process of being submitted to "PLoS Computational Biology" and are result of collaborative work with Mert Karakaº, another graduate student in the Meiler Lab.;The Bayesian energy potentials described in Chapter III and protein structure prediction framework described in Chapter IV serve as the basis for BCL::EM-Fold [2], a method for utilizing cryoEM density maps for protein structure prediction, as well as several other methods for which publications are currently under preparation. These other methods include but are not limited to protein structure prediction for membrane proteins, multimeric proteins, integration of NMR, MS and EPR restraints and loop building.;Chapter I provides an introduction with a limited overview of protein structure and experimental methods for protein structure elucidation. Additionally, computational protein energy evaluation through knowledge and physics based energy functions is introduced. Lastly, methods for protein-protein structure comparison are discussed. Chapter II describes BCL::EM-Fit, the algorithm for rapid fitting of atomic structures into electron density maps based on the image recognition algorithm known as "geometric hashing" employed for three dimensional problems. The chapter discusses how it improves time, accuracy and completeness compared to other methods. Chapter III concentrates on Bayesian energy potentials which are derived to evaluate protein structures focusing on the protein topology represented by the geometrical arrangement of secondary structure elements. This potential is used within BCL::Fold, a novel de novo protein structure prediction algorithm. Chapter IV focuses on the minimization framework, as well as the moves utilized in BCL::Fold and provides an excerpt of a benchmark of the method.