A Foundation Model Identifies Broad-Spectrum Antimicrobial Peptides against Drug-Resistant Bacterial Infection
Tingting Li, Xuanbai Ren, Xiaoli Luo, Zhuole Wang, Zhenlu Li, Xiaoyan Luo, Jun Shen, Yun Li, Dan Yuan, Ruth Nussinov, Xiangxiang Zeng, Junfeng Shi, Feixiong Cheng
Abstract
Development of potent and broad-spectrum antimicrobial peptides (AMPs) could help overcome the antimicrobial resistance crisis. We develop a peptide language-based deep generative framework (deepAMP) for identifying potent, broad-spectrum AMPs. Using deepAMP to reduce antimicrobial resistance and enhance the membrane-disrupting abilities of AMPs, we identify, synthesize, and experimentally test 18 T1-AMP (Tier 1) and 11 T2-AMP (Tier 2) candidates in a two-round design and by employing cross-optimization-validation. More than 90% of the designed AMPs show a better inhibition than penetratin in both Gram-positive (i.e., S. aureus) and Gram-negative bacteria (i.e., K. pneumoniae and P. aeruginosa). T2-9 shows the strongest antibacterial activity, comparable to FDA-approved antibiotics. We show that three AMPs (T1-2, T1-5 and T2-10) significantly reduce resistance to S. aureus compared to ciprofloxacin and are effective against skin wound infection in a female wound mouse model infected with P. aeruginosa. In summary, deepAMP expedites discovery of effective, broad-spectrum AMPs against drug-resistant bacteria. New approaches to develop antimicrobial agents are urgently needed. In this study, the authors develop a peptide language-based deep generative model to design broad-spectrum antimicrobial peptides against drug-resistant bacteria and validate promising candidates in a wound mouse model.