Machine learning for precision prediction of antimicrobial peptide activity and spectrum

Area of Science:

• Biochemistry
• Computational Biology
• Drug Discovery

Background:

• Antibiotic resistance is a growing global health crisis requiring novel treatments.
• Antimicrobial peptides (AMPs) show promise due to broad activity and low resistance development.
• Predicting AMP potency and spectrum is a key challenge for clinical use.

Purpose of the Study:

• To review how machine learning advances AMP research, focusing on prediction and design.
• To assess AMP databases for machine learning suitability.
• To explore peptide representation, de novo design, and prediction frameworks.

Main Methods:

• Review of machine learning applications in AMP research.
• Analysis of peptide representation learning techniques (sequence, structure, dynamics).
• Discussion of de novo AMP design and quantitative prediction models.

Main Results:

• Machine learning shifts AMP research from discovery to precision prediction and design.
• Limitations in current AMP databases hinder machine learning readiness.
• Progress in encoding peptide features for activity prediction and de novo design.

Conclusions:

• Machine learning is crucial for developing clinically viable AMPs.
• Future directions include integrated pipelines, interpretable models, and experimental validation.
• Precision-guided AMP design offers a path to combat antibiotic resistance.