Principles of protein-protein interaction: from binary complexes to cellular networks

Physical interactions between protein domains are fundamental to the workings of a cell. Akey challenge in biology in the post-genomic era is to predict functional protein interactionson a large scale. Experimental investigation into protein-protein interactions has been extensive, including recent large-scale screens using mass spectrometry; but the techniques are time consuming and labor-intensive. Computational research on protein interactions encompasses not only prediction, but also understanding the nature of the interactions and their three-dimensional structures. This knowledge is key to inferring protein function and to delineate protein interaction maps in complete genomes. I will talk about the results of extensive computational analyses carried out on protein-protein interfaces identified in large datasets of complexes obtained from the Protein Data Bank (PDB). The first approach was to analyze and classify interfaces in terms of their 3D architectures: I found that there are a limited set of binding motifs, and these are re-used in different combinations to construct the interface in functionally diverse protein complexes. I will outline the implications of this research for structure-based drug design of protein interfaces. Conserved architectural motifs provide similar interaction patterns in proteins with a completely different tertiary fold and function. Identification of these structural similarities will aid the design of drugs targeting protein interfaces. Second, I examined patterns of residue conservation in interface regions. By studying how interfaces have evolved, one can identify residues that are functionally important for the interaction and use this information for designing artificial interfaces. Conserved interface residues form clusters suggesting the necessity of critical residues acting in tandem. Interfaces generally consist of a central buried core and a peripheral rim region – the core is significantly more conserved in evolution than the rim pointing to its greater importance in binding. The core contains the majority of ‘hotspot’ residues that contribute significantly to the binding energy. I will discuss how energetic criteria can be combined with geometric information to computationally predict hotspots of binding. Further, I show how studying protein flexibility improves our basic understanding of protein-protein recognition. Currently, I am systematically characterizing conformational changes that occur upon binding by comparing known 3D structures of the unbound protein components and the corresponding protein complex. Finally I will discuss my future plans for applying all these results towards proteome-scale interaction prediction, applications such as computational docking, and, studying the modular design of cellular signal transduction networks.