A Global Discovery of Antimicrobial Peptides in Deep-Sea Microbiomes Driven by an ESM-2 and Transformer-based Dual-Engine Framework

The global crisis of multidrug-resistant pathogens necessitates innovative antimicrobial peptide (AMP) discovery. Deep-sea microbiomes represent an underexplored resource for novel AMPs, but their mining is hindered by biases in current prediction methods, including sequence length imbalance, N-terminal methionine artifacts, and lack of microbial optimization. To overcome these, we developed XAMP, a dual-engine predictor integrating XAMP-E (based on ESM-2 for high-accuracy feature representation) and XAMP-T (a one-layer Transformer for accelerated screening). By training on debiased datasets, XAMP achieved a median AUC of 0.972, an approximately 10% improvement over state-of-the-art tools, with XAMP-T operating 5 to 40 times faster. Applying this pipeline to deep-sea metagenomes, we identified 2,355 promising AMP candidates. Experimental validation of six synthesized peptides against ESKAPE pathogens demonstrated potent, broad-spectrum activity, particularly against Gram-negative bacteria, which dominate deep-sea ecosystems and represent a major challenge in nosocomial infections. This study establishes a robust computational-experimental framework for discovering therapeutic candidates from extreme environments to combat antibiotic resistance crises.