Computational Prioritisation of Fish-Derived Antimicrobial Peptides by Integrated Multi-Layer Immunogenicity and Structural Analysis.

Fish-derived antimicrobial peptides (AMPs) represent a promising but underexplored source of therapeutic candidates. However, their immunological safety for potential human application has not been systematically evaluated. Here, we applied a multi-stage computational approach to 154 fish AMPs retrieved from the Antimicrobial Peptide Database (APD6). Sequential safety filtering reduced the dataset to 23 AMP candidates, which were subsequently evaluated using a composite safety and immunogenicity risk score based on MHC class I binding, MHC class II binding, B-cell epitope exposure, and residual hemolytic risk. B-cell risk was conditioned on MHC class II activation to account for the role of CD4⁺ T-cell help in class-switched antibody responses. Eight AMPs were classified as Predicted Low Risk based on their relative composite safety and immunogenicity risk. Both this classification and the rank order of the three highest-ranked candidates remained stable across the tested sensitivity scenarios. The eight Predicted Low Risk candidates were further evaluated for conformational variability, secondary structure, structural rigidity, and membrane interaction. NRC-01 showed the lowest composite safety and immunogenicity risk score (0.213), indicating the lowest predicted risk among the evaluated candidates. NRC-18 achieved the highest final multi-criteria score (0.764), combining a relatively low composite safety and immunogenicity risk score (0.296), strong predicted membrane interaction (ΔG = -5.2 kcal/mol), high helical content (69.0%), and the highest dominant cluster population among the top three candidates (57.86%). Overall, the combined computational analyses identified a small group of fish-derived AMPs with favourable predicted safety, immunogenicity, and structural characteristics that warrant further experimental investigation.