| Popis: |
Protein-protein interactions (PPI) play essential roles in virtually all biological processes. While modern structural determination methods such as x-ray crystallography and NMR provide valuable information that can aid in our understanding of these complexes, they often produce many putative conformations that require further refinement to determine the biological structure. Here, we outline two methods that may aid in the refinement of these potential conformations. Using a manually curated database of both biologically functional and biologically non-functional PPI (FLIPdb), we explore the relationship between binding free energy as determined by computational alanine scanning (CAS) and small translational perturbations of the interacting protein subunits. In short, the interacting protein subunits are translated along a grid that is coplanar with their dividing plane and the interface is subjected to CAS at each point along the grid. The coordinates of the translation are then used in combination with the binding free energy to create an energy landscape that is further analyzed for various features. Analysis of these features by SVM has allowed us to achieve discrimination of biologically functional/biologically non-functional interfaces with an accuracy of seventy-six percent, and suggests that biologically functional interfaces are more sensitive to perturbation than biologically non-functional interfaces. In addition to translational perturbation sensitivity, we analyze the size and distribution of energetic hotspots present in each PPI. Using solvent accessible surface area (SASA) as measured by NACCESS, and binding free energy as determined by CAS, residues are clustered into hotspots using density based clustering. The size and energy density of these hotspots is then analyzed. Analysis of these features by SVM allows us to discriminate between biologically functional and biologically non-functional PPI with an accuracy of approximately eighty percent. |
