Abstract
Bacterial diseases cause high mortality and considerable losses in aquaculture. The rapid expansion of intensive aquaculture has further increased the risk of large-scale outbreaks. However, the emergence of drug-resistant bacteria, food safety concerns, and environmental regulations have severely limited the availability of antimicrobial. Compared to traditional antibiotics, antimicrobial peptides (AMPs) offer broad spectrum activity, physicochemical stability, and lower resistance development. However, their low natural yield and high extraction costs along with the time-consuming and expensive nature of traditional drug discovery, pose a challenge. In this study, we applied a machine-learning macro-model to predict AMPs from three macrogenomes in the water column of South American white shrimp aquaculture ponds. The AMP content per megabase in the traditional earthen pond (TC1) was 1.8 times higher than in the biofloc pond (ZA1) and 63% higher than in the elevated pond (ZP11). A total of 1033 potential AMPs were predicted, including 6 anionic linear peptides, 616 cationic linear peptides, and 411 cationic cysteine-containing peptides. After screening based on structural, and physio-chemical properties, we selected 10 candidate peptides. Using a rapid high-throughput cell-free protein expression system, we identified nine peptides with antimicrobial activity against aquatic pathogens. Three were further validated through chemical synthesis. The three antimicrobial peptides (K-5, K-58, K-61) showed some inhibitory effects on all four pathogenic bacteria. The MIC of K-5 against Vibrio alginolyticus was 25 μM, the cell viability of the three peptides was higher than 70% at low concentrations (≤12.5 μM), and the hemolysis rate of K-5 and K-58 was lower than 5% at 200 μM. This study highlights the benefits of machine learning in AMP discovery, demonstrates the potential of cell-free protein synthesis systems for peptide screening, and provides an efficient method for high-throughput AMP identification for aquatic applications.