| Although much is known about the experimentally measurable properties of individual proteins, we do not have a solid understanding of the universal properties that control protein structure and stability. The amount of protein crystal structure data has grown exponentially in recent years, providing new datasets of the structures of soluble proteins, transmembrane proteins and protein-protein interactions. Yet much of our current understanding of protein structural properties is based off prior studies using smaller, less accurate datasets.;My work, building on prior protein structural studies, uses the new plethora of protein crystal structure data to develop a deeper understanding of the physical properties of proteins. I focus on using an atomistic model of proteins, involving steric interactions between atoms, to study the packing fraction of protein cores and the placement of amino acid side chains.;This dissertation presents three related computational studies of the properties of protein structures. In the first study, I present an analysis of the packing fraction of protein cores. I show that protein cores are composed of random close packed amino acids. I also show that the same packing fractions can be found for jammed packings of bumpy, non-axisymmetric particles. This result indicates that there is little room for movement of amino acids in the protein core.;I then take a more detailed look at protein cores to quantify the amount of side chain dihedral angle movement possible in the core. To do this, I apply a hard-sphere model with stereochemical constraints to rotate the side chain dihedral angles in each protein core and predict the collective side chain dihedral angle positions of sets of amino acids. I find that this model is able to predict 90% or more of the side chain dihedral angles in the core within 30° of the crystal structure values, showing that steric interactions dominate side chain placement in protein cores and that there is little room for rearrangement.;Finally, I expand both of these studies to two other protein types: protein-protein interfaces and transmembrane proteins. For both datasets, I investigate the packing fraction and side chain predictability of both core and solvent exposed amino acids. I find that soluble proteins, protein-protein interfaces and transmembrane proteins are all equally packed and that the side chain dihedral angles of amino acids with up to 30% solvent accessibility can be predicted using the hard-sphere plus stereochemical constraints model.;This thesis work challenges the commonly accepted properties of protein structure, in particular that proteins have a high packing fraction similar to crystalline packing of spheres and that transmembrane proteins are fundamentally different from soluble proteins. Future projects based on this work will apply these results to study protein mutations and to design new protein-protein interfaces. |
