Protein-protein interactions play a variety of important roles in the living cell and the mapping of the protein interactome is currently pursued by many laboratories. High-throughput techniques such as the yeast two-hybrid system and mass-spectrometry are generating a large amount of interaction data for many model organisms. Further experimental characterization of these interactions is often time-consuming, and the detailed structural characterization of protein-protein complexes is challenging. For example, while the RCSB Protein Data Bank contains about 20,000 unique protein structures (as of 2007), the ProtCom database of protein complexes contains only a few hundred binary protein-protein complexes. Considering the challenges that NMR and X-ray crystallography face with characterizing protein-protein complexes, it is tempting to model protein-protein interactions computationally.
Methods for docking of small molecules to proteins are fairly well established. Methods for the docking of two proteins, on the other hand, are currently in the early stages of development. Current capabilities of different algorithms are regularly tested by developers, one example being via the CAPRI (Critical Assessment of PRediction of Interactions) community-wide experiment. The accurate docking of two proteins is difficult with the current computational methods because of the difficulty in accounting for the protein flexibility. Early protein docking methods treated the two interacting proteins as rigid bodies. Such approach is often successful when the protein structure does not change significantly. However, the proteins frequently undergo significant induced fit conformational changes upon interaction, and the rigid protein-protein docking algorithms fail to model such complexes.
Even though the detailed methodologies of protein-protein interactions may be difficult to understand, the basic principles that govern the protein-protein interactions are easy to grasp. The basic idea, complementarity of the surfaces of the two interaction partners, was anticipated in 1940's by Linus Pauling and Max Delbrück. In a simple view, a negatively charged surface patch from one protein is likely to interact with a positively charged surface patch of another protein due to favorable electrostatic interactions. Similarly, two hydrophobic patches are likely to interact with each other because the release of organized waters near the hydrophobic surfaces leads to the increase in system's entropy. The two proteins bind when such patches can adopt a geometry that allows a close contact between the two surfaces in the complex.
These simple ideas can be put into practical use through computer programs that can easily visualize the properties of protein surfaces and allow efficient 3D manipulation of two proteins on the computer. A number of computer programs capable of such tasks have been developed. One such program is PyMOL, developed by Dr. Warren DeLano in collaboration with many other scientists. PyMOL allows the highly customizable display and 3D manipulation of several molecular objects, partially through the use of a powerful command language. The calculation and display of electrostatic potential surfaces in PyMOL is handled via the ABPS plugin (APBS and the APBS plugin are works by Nathan Baker from Washington University and Mickael Lerner from University of Michigan, respectively). The present tutorial illustrates with select examples how PyMOL can be used to illustrates principles behind protein-protein interactions and provides workarounds for common issues that are encountered in such calculations.