Gregory Reck, PhD (School of Systems Biology, GMU)

Protein functionality in a broad spectrum of cellular processes is often dependent on protein-protein interactions. These interactions can be critical to enzyme performance and regulation, immune system response, signal transduction and a host of other complex cellular processes. While knowledge of many of the protein structures involved has dramatically improved, a more thorough understanding of the protein features contributing to protein-protein interactions could lead to improved techniques for predicting interaction sites, identifying nearest-native interface structures and analyzing protein behavior in reaction networks. This presentation describes an ongoing effort to apply Delaunay tessellation and associated four-body nearest neighbor statistical potentials to explore and compare these interface regions for several types of protein interactions.

Tessellation is a computational geometry tool that has been used in reported studies to isolate and characterize interface structures. Specific residues and atoms that are interacting in the interfacial region can be identified and examined. Delaunay Tessellation (DT) in particular decomposes internal protein structures into a set of irregular tetrahedra, each containing 4 residues at its vertices that are nearest neighbors of each other in the structure. DT can be applied to a set of protein structures to determine the frequency of specific residue quadruplets, which in turn can be used to develop a statistical potential function. Potential functions can be derived from representative non-redundant groups of proteins, or alternatively from sets of more specific structural components such as protein surfaces or interface regions. This effort applies potential functions derived via DT to study and compare interface regions representative of several types of protein-protein interactions, including obligatory, non-obligatory, homo and heterodimers, oligomers and complexes. Total potential values as well as potential contours can be derived for the interfaces and the associated monomer surfaces. These contours may provide opportunities for pattern recognition and machine learning tools to classify interfaces, identify nearest-native conformations and investigate features such as hot-spots.