De Novo Design of Membrane-Targeting Antimicrobial Peptides Against Gram-Negative Bacteria Using a Generative...

Area of Science:

• Biotechnology
• Computational Biology
• Drug Discovery

Background:

• Antimicrobial resistance (AMR) in Gram-negative bacteria presents a critical global health threat, exacerbated by their outer membranes which impede conventional antibiotic effectiveness.
• Antimicrobial peptides (AMPs) offer a promising alternative to traditional antibiotics due to their unique mechanisms of action.
• Designing effective AMPs, particularly membrane-targeting antimicrobial peptides (MTAMPs) for Gram-negative bacteria, requires sophisticated approaches that consider multiple biological and chemical properties.

Purpose of the Study:

• To develop a novel, AI-driven generative framework for the de novo design of membrane-targeting antimicrobial peptides (MTAMPs) specifically against Gram-negative bacteria.
• To integrate sequence, physicochemical, and spatial structure (PCSS) descriptors into a conditional variational autoencoder (GenMTAMP) model for directed MTAMP generation.
• To establish a robust pipeline for screening, predicting, and experimentally validating designed MTAMPs with desired antimicrobial properties and safety profiles.

Main Methods:

• Development of a conditional variational autoencoder (GenMTAMP) model incorporating sequence, physicochemical, and spatial structure (PCSS) descriptors.
• Utilizing target PCSS descriptors as conditional constraints to guide the generative process for de novo MTAMP design.
• Employing subsequent identification (ClaAMP) and prediction (PreAMP) modules for candidate MTAMP screening and evaluation.
• Experimental validation of top-ranked MTAMP candidates, including assessment of antibacterial activity, cytotoxicity, and haemolytic activity.

Main Results:

• The GenMTAMP framework successfully generated novel MTAMP candidates with targeted properties.
• Two designed peptides, MTAMP003 and MTAMP004, demonstrated significant antibacterial activity against Gram-negative bacteria.
• These validated MTAMPs exhibited low cytotoxicity and haemolytic activity towards mammalian cells, indicating a favorable safety profile.
• Mechanism studies confirmed that the designed MTAMPs effectively disrupt the Gram-negative bacterial outer membrane with minimal impact on mammalian cell membranes.

Conclusions:

• The study presents a targeted and efficient generative AI framework for the de novo design of MTAMPs against Gram-negative bacteria.
• The developed framework enables the rational design of AMPs with specific functional properties, offering a generalizable approach for future drug discovery efforts.
• The findings highlight the potential of AI-guided design in addressing the challenge of antimicrobial resistance by creating novel therapeutic agents.