Uncovering evolutionarily remote and highly potent antimicrobial peptides with protein language models

Area of Science:

• Microbiology
• Bioinformatics
• Drug Discovery

Background:

• Antibiotic resistance necessitates novel antimicrobial peptides (AMPs).
• Current AMP discovery methods miss evolutionarily distant candidates.
• Identifying remote AMPs is key for new clinical treatments.

Purpose of the Study:

• To develop a new computational method for identifying evolutionarily remote AMPs.
• To discover novel, potent AMPs from large-scale genomic data.
• To validate the therapeutic potential of newly discovered AMPs.

Main Methods:

• Developed HMD-AMP, a protein language model for AMP discovery.
• Applied HMD-AMP to mammalian host and gut microorganism genomes.
• Experimentally validated predicted AMPs for antibacterial activity and toxicity.

Main Results:

• HMD-AMP identified over 37 million potential AMPs.
• Experimental validation confirmed strong antibacterial activity in 74 of 91 sequences.
• 48 validated AMPs were evolutionarily remote from known AMPs.
• Four AMPs demonstrated broad-spectrum activity, low toxicity, and therapeutic efficacy in vivo.

Conclusions:

• HMD-AMP is an effective strategy for discovering novel, evolutionarily remote AMPs.
• The discovered AMPs show promise as clinical candidates against antibiotic resistance.
• This approach significantly expands the repertoire of known AMPs.