Discovery of AMPs from random peptides via deep learning-based model and biological activity validation

Area of Science:

• Biotechnology
• Computational Biology
• Drug Discovery

Background:

• Antimicrobial peptides (AMPs) are crucial for combating drug-resistant pathogens.
• Discovering novel AMPs is challenging and costly.
• Artificial intelligence (AI) offers potential to accelerate AMP discovery.

Purpose of the Study:

• To develop and apply deep learning-based multi-discriminator models for identifying novel antimicrobial peptides (AMPs).
• To screen a large library of random peptides for potential AMPs.
• To validate the efficacy and safety of identified AMP candidates.

Main Methods:

• Construction of three deep learning multi-discriminator models.
• Screening of 30,000 random peptides.
• Antimicrobial and hemolytic activity assays for candidate peptides.
• Mechanistic studies on bactericidal effects.
• In vivo efficacy testing in a mouse wound infection model.

Main Results:

• Twelve novel antimicrobial peptides (AMPs) were successfully screened using the deep learning models.
• Three candidate peptides (P2, P11, P12) demonstrated potent antimicrobial activity and low hemolytic activity.
• Peptides exert bactericidal effects via membrane disruption, reducing resistance potential.
• Peptide 12 (P12) showed significant efficacy in a mouse model of Staphylococcus aureus wound infection with minimal organ toxicity.

Conclusions:

• Deep learning-based multi-discriminator models are effective for identifying novel antimicrobial peptides (AMPs) from random peptide libraries.
• Identified AMPs exhibit promising clinical potential for treating bacterial infections, including those caused by drug-resistant strains.
• The developed AI approach enhances the efficiency and reduces the cost of antimicrobial peptide discovery.