Rational Design and Machine Learning-Guided Screening of Short α-Helical Antimicrobial Peptides Derived from Dendrocin-ZM1

Background

The increasing prevalence of multidrug-resistant bacterial pathogens has created an urgent need for new antimicrobial strategies with improved efficacy and safety profiles. Antimicrobial peptides (AMPs) are promising therapeutic candidates due to their membrane-targeting mechanisms and broad antimicrobial potential. This study aimed to computationally design and screen shortened analogs of Dendrocin-ZM1 to identify potential antimicrobial peptide candidates against Staphylococcus aureus. Materials and Methods: Dendrocin-ZM1, a 33-amino-acid α-helical antimicrobial peptide, was selected as the parental template for rational peptide redesign. A series of shortened and sequence-engineered analogs was generated by retaining selected cationic and hydrophobic residues associated with antimicrobial peptide activity. The designed peptides were evaluated based on physicochemical properties, including peptide length, net charge, hydrophobicity, and predicted α-helical structural characteristics. Antimicrobial potential was assessed using the CAMPR4 platform with Support Vector Machine, Random Forest, and Artificial Neural Network classifiers. Additional screening was performed using the DBAASP prediction platform, focusing on predicted activity against S. aureus. Potential toxicity and hemolytic effects were evaluated using ToxinPred and HemoPI2, respectively, and selected peptides were subjected to structural modeling to assess predicted α-helical conformations.