Characterizing Binding Properties of Protein Interaction Domain

Abstract

A large proportion of protein-protein interactions are mediated by families of peptide-binding domains. Comprehensive characterization of each of these domains is critical for understanding the mechanisms and networks of protein interaction at the domain level. A systematic experimental strategy was developed for efficient binding property characterization of peptide-binding domains based on high throughput validation screening of a specialized candidate ligand library using yeast two-hybrid system. As for simple adaptor protein without any other known functional domains, the potential functions of the complex were predicted by functional annotations from a MILANO literature search and subcellular localizations. The ligands were considered more likely to be functionally associated if they had similar patterns of functions or closely related functions. For some functionally associated ligand pairs, interaction with one ligand was found to be influenced by another ligand in a yeast three-hybrid system. Ideally protein-protein interactions should be studied with high throughput computational approaches first to screen protein sequence databases to direct the validating experiments toward the most promising peptides. An integrated machine learning systems was built using three novel coding methods and used to screen the Swiss-Prot and GenBank protein databases for potential ligands of SH3 and PDZ domains. A large fraction of predictions has already been experimentally confirmed by other independent research groups, indicating a satisfying generalization capability for future applications in identifying protein interactions. S2/July 2008/Special Issue