Screening antimicrobial peptides and probiotics using multiple deep learning and directed evolution strategies.

Area of Science:

• Biotechnology
• Computational Biology
• Microbiology

Background:

• Current antimicrobial peptide (AMP) prediction models exhibit limitations in accuracy and applicability, hindering industrial use.
• Developing accurate and efficient methods for AMP discovery and production is crucial for combating antimicrobial resistance.

Purpose of the Study:

• To develop and enhance an AMP prediction model for improved accuracy and industrial applicability.
• To optimize AMP production yields and identify novel microbial sources for AMP generation.

Main Methods:

• Implemented Comparing and Optimizing Multiple DEep Learning (COMDEL) algorithms for AMP prediction.
• Utilized high-throughput screening, phage-assisted evolution, and cell-free synthesis for AMP production.
• Employed multi-omics analysis and microdroplet sorting for probiotic screening and mutant selection.

Main Results:

• The COMDEL model achieved 94.8% accuracy in testing and 88% in experimental verification, outperforming existing models.
• Developed a cell-free AMP synthesis system achieving yields of 0.5-2.1 g/L within hours.
• Identified *Lactobacillus plantarum* as a key probiotic for AMP generation and screened mutants with doubled antimicrobial ability.

Conclusions:

• The enhanced COMDEL model and optimized production systems significantly advance AMP discovery and manufacturing.
• The identified *L. plantarum* mutants hold substantial potential for industrial-scale antimicrobial peptide generation.