Structural Interaction Fingerprint (SIFt):  A Novel Method for Analyzing Three-Dimensional Protein−Ligand Binding Interactions

Representing and understanding the three-dimensional (3D) structural information of protein-ligand complexes is a critical step in the rational drug discovery process. Traditional analysis methods are proving inadequate and inefficient in dealing with the massive amount of structural information being generated from X-ray crystallography, NMR, and in silico approaches such as structure-based docking experiments. Here, we present SIFt (structural interaction fingerprint), a novel method for representing and analyzing 3D protein-ligand binding interactions. Key to this approach is the generation of an interaction fingerprint that translates 3D structural binding information from a protein-ligand complex into a one-dimensional binary string. Each fingerprint represents the "structural interaction profile" of the complex that can be used to organize, analyze, and visualize the rich amount of information encoded in ligand-receptor complexes and also to assist database mining. We have applied SIFt to tackle three common tasks in structure-based drug design. The first involved the analysis and organization of a typical set of results generated from a docking study. Using SIFt, docking poses with similar binding modes were identified, clustered, and subsequently compared with conventional scoring function information. A second application of SIFt was to analyze approximately 90 known X-ray crystal structures of protein kinase-inhibitor complexes obtained from the Protein Databank. Using SIFt, we were able to organize the structures and reveal striking similarities and diversity between their small molecule binding interactions. Finally, we have shown how SIFt can be used as an effective molecular filter during the virtual chemical library screening process to select molecules with desirable binding mode(s) and/or desirable interaction patterns with the protein target. In summary, SIFt shows promise to fully leverage the wealth of information being generated in rational drug design.