From AI-Driven Sequence Generation to Molecular Simulation: A Comprehensive Framework for Antimicrobial Peptide...

Area of Science:

• Computational chemistry and drug discovery
• Antimicrobial peptide research
• Bioinformatics and machine learning applications

Background:

• Bacterial resistance to existing antibiotics is a growing global health threat.
• Antimicrobial peptides (AMPs) show promise but face challenges in clinical translation.
• Deep learning offers new avenues for designing novel AMPs.

Purpose of the Study:

• To develop and validate an integrated computational framework for systematic AMP design and screening.
• To identify novel AMP candidates with potential antibacterial activity.
• To establish a cost-effective strategy for AMP discovery.

Main Methods:

• Utilized a character-string-based generative adversarial network (GAN) to generate candidate AMP sequences.
• Employed the PGAT-ABPp discriminative network and physicochemical analysis for preliminary screening.
• Conducted molecular dynamics simulations to assess membrane interaction and synthesized promising candidates for in vitro testing.

Main Results:

• Generated 50 candidate AMP sequences, identifying 9 potential candidates after initial screening.
• Molecular simulations indicated water pore formation in bacterial membranes for two selected peptides.
• Synthesized peptides demonstrated in vitro efficacy against Gram-negative (E. coli) and Gram-positive (S. aureus) bacteria.

Conclusions:

• The integrated computational framework successfully discovered two novel, clinically relevant antimicrobial peptides.
• The identified AMPs exhibit broad-spectrum activity against key bacterial pathogens.
• This study validates a cost-effective and broadly applicable computational strategy for accelerating AMP discovery.