AMP-Atlas: Comprehensive Atlas of Antimicrobial Peptides to Combat Multidrug-resistant Bacteria

Abstract

The escalating threat of infections caused by drug-resistant bacteria poses a significant global health challenge, with projections estimating 10 million annual deaths by 2050. While the development of conventional antibiotics has stagnated since the late 1990s, antimicrobial peptides (AMPs), short amino acid sequences exhibiting potent antimicrobial activity, have emerged as a promising alternative, demonstrating efficacy even against drug-resistant bacteria. However, despite the identification of numerous AMPs, their translation into clinically approved therapeutics remains limited, highlighting the critical need for accelerated discovery methods that transcend traditional experimental screening. Here, we introduce AMP-Atlas, an AI system inspired by cutting-edge natural language processing, designed to accurately predict antimicrobial activity from peptide sequences alone. AMP-Atlas achieves state-of-the-art performance, outperforming existing methods in AMP identification. Furthermore, we leveraged AMP-Atlas to screen human indigenous bacterial flora species, revealing a vast reservoir of previously unexplored AMP candidates. Our findings underscore the transformative potential of AI-powered approaches to revolutionize AMP discovery and development, paving the way for innovative therapeutic strategies to combat the looming threat of drug-resistant infections.