SPOT-Peptide: Template-Based Prediction of Peptide-Binding Proteins and Peptide-Binding Sites.

Abstract

Peptide-binding domains have been successfully targeted in therapeutic applications. However, many peptide-binding proteins (PBPs) remain uncharacterized. Computational prediction of peptide-domain interfaces is challenging due to short lengths, lack of well-defined structures, and limited conservation of peptide motifs. Here we present SPOT-peptide, a template-based protocol for the simultaneous prediction of peptide-binding domains and peptide binding sites independent of specific peptide composition. SPOT-peptide leverages the dogmatic relationship between protein structure and function to predict peptide-binding characteristics for an unknown target based on remote structural homologues. In a leave-homologue out benchmark evaluation, PBPs are discriminated with a Matthews correlation coefficient (MCC) of 0.420 and the correct binding sites are identified in 80% of the predicted PBPs. Furthermore, replacing the holo target structures with equivalent structures in the apo conformation only marginally diminished PBP recovery. The method is available as a web server at http://sparks-lab.org/tom/SPOT-peptide .