Computational design of cyclic peptides to inhibit protein-peptide interactions.

Many protein-protein interactions result from the binding of one folded protein with one short peptide segment, such as complexes formed by SH3 or PDZ domains. These transient protein-peptide interactions are notably involved in cellular signaling pathways and generally have low affinities, which opens the possibility to design competitive inhibitors of these complexes. We present and assess here our computational approach, called Des3PI, to design de novo cyclic peptides with potential high affinity for protein surfaces involved in interactions with peptide segments. The results were not conclusive for two receptors, the αVβ3 integrin and the CXCR4 chemokine receptor, but were promising in the case of SH3 and PDZ domains: For the former, Des3PI was able to find at least one cyclic sequence with six hotspots that binds a SH3 domain with a better theoretical affinity to the known proline-rich RLP2 peptide. For the latter, Des3PI could identify at least four cyclic sequences with four or five hotspots that have lower binding free energies computed by the MM-PBSA method than the reference peptide GKAP.