Assessing, predicting and designing peptide ligands for proteins

Abstract

Up to 40% of the protein interactions in a cell are mediated by peptides and protein-peptide interactions play a vital role in a cell’s functioning. Peptide-mediated protein interactions have been suggested as a potential drug target in many cellular pathways and recently, peptide ligands have attracted a lot of attention as promising drug candidates. Therefore, knowing the structures of such interactions is very essential for their further characterization. In this study, we propose a knowledge-based method for predicting peptide ligands provided a query protein structure with a known binding site. The method first extracts a query structural motif from the binding site of the given protein. We have constructed a library of such structural motifs extracted from the protein structures present in the Protein Data Bank(PDB) against which the query is compared. After finding a structurally similar match from the database, the method extracts the neighbourhood information from the match to predict atoms that will be energetically stable in the query protein’s binding site. These predicted atoms will be used to suggest a potential peptide ligand for the given protein. Here, we have developed the framework for this method and performed a set of tests to validate the method’s ability to predict an energetically stable partner provided a set of neighbouring atoms. The method, when used to predict a known chemical group when subjected to deletion from a protein structure, was able to correctly predict it back approximately 81% of the time. Since the method focuses on the local packing of atoms in protein structure, it can also be used to predict protein structure stability and to identify missing atoms and residues in protein structures.