Characterization of protein interfaces
(Edelsbrunner, Rudolph; Ban)

Protein-protein interactions, which form the basis for most cellular processes, are poorly understood in regard to specificity and energetics and rank among the most important yet least understood biochemical phenomena. Prior work in this field consists primarily of
1. visualizing the surfaces of individual proteins at their interaction sites,
2. statistical summaries of amino acids involved in protein-protein interactions, and
3. cataloging of results from mutagenesis experiments designed to probe the importance of specific residues in protein-protein interactions.

We are pursuing computational approaches to understanding the protein-protein recognition problem by developing novel tools to characterize protein-protein interaction sites both qualitatively and quantitatively. Using tools from computational geometry including weighted Voronoi diagrams, weighted Delaunay triangulations, and topological persistence, we have rigorously defined an algorithm and implemented software to describe the molecular interface between two or more complexed proteins. This molecular interface can be thought of as a wrinkled sheet of paper and can be visualized independently of the contributing proteins. With this representation, we aim to provide a novel scaffold for mapping both global and local measures of interest onto the interface. Such measures include electrostatics, hydrophobicity, and wrinkledness. We are currently implementing these mapping tools. Additionally, in our algorithm we have defined a seal function that allows us to assign a hierarchy of interaction surfaces.

This hierarchy appears to distinguish sites of importance that correlate with so-called hot-spot residues in the limited data set we have studied thus far. We are in the process of characterizing more protein-protein interfaces using this algorithm to test the universality of this hierarchy and its physical basis.