COMPUTATIONAL DESIGN OF COMBINATORIAL PEPTIDE LIBRARY FOR MODULATING PROTEIN-PROTEIN INTERACTIONS

Abstract

Screening phage-displayed combinatorial peptide library is an effective approach for discovery of peptide modulators for protein-protein interactions. However, as peptide length increases, the chance of finding active peptides in a finite size library diminishes. To increase the likelihood of finding peptides that bind to a protein, we develop statistical potential for computational construction of biased combinatorial antibody-like peptide libraries. Based on the alpha shapes of antibody-antigen complexes, we developed an empirical pair potential for antigen-antibody interactions that depends on local packing. We validate this potential and show that it can successfully discriminate the native interface peptides from a simulated library of 10,000 random peptides for 34 antigen-antibody complexes. In addition, we show that it can successfully recognize the native binding surface patch among all possible surface patches taken from either the antibody or the antigen for seven antibody-antigen protein complexes contained in the CAPRI (Critical Assessment of Predicted Interactions) dataset. We then develop a Weighted Amino Acid Residue sequence Generator (WAARG) for design of biased peptide library. When compared with a random peptide library, WAARG libraries contain more native-like binding peptides at a significantly smaller size. Our method can be used to construct peptide library for screening of antibody variants with improved specificity and affinity to a target antigen. It can also be used for screening of antibody-like antagonist peptides modulating other protein-protein interactions.