Studying peptide structure-activity/property relationships (P-SA/PR) helps understanding how the structural variations of peptides influence their biological activities and other functional properties.¹ This knowledge accelerates the rational design and optimisation of peptide-based drugs, biomaterials, or diagnostic agents. Conventional computational studies often examine peptide structures from their primary sequences, essentially encoded from their proteinogenic residues, excluding peptide libraries with post-translational and synthetic modifications. The molecular fingerprint MAP4 emerges as a tool designed to mapping structural diversity of complex molecules.³

This study used structure-activity landscape modeling⁴ to perform P-SA/PR studies of an exemplary dataset of 223 antimicrobial peptides against methicillin-resistant Staphylococcus aureus (MRSA).⁵ The dataset contained peptides with canonical (200/89.7%) and non-canonical/modified amino acids (23/10.3 %). To this end, we employed the MAP4 fingerprint to represent the chemical structures of the peptides, study their relationship(s) with the antibacterial activity, and seek potential activity cliff(s). We identified critical residues and structural motifs that play a crucial role in the anti-MRSA activity of the peptides. The fingerprint-based similarity values correlated poorly with the sequence-based identity values (R² = 0.31), suggesting that the MAP4 similarity metrics complement the knowledge derived from sequence alignments, but do not replace them. About 31 % of the matched anti-MRSA peptides were considered activity cliffs. This is the first computational study to systematically explore the activity landscape of peptides with non-canonical residues, emphasizing the quantification of structural similarity.