Design of Highly Potent Antibiofilm, Antimicrobial Peptides Using Explainable Artificial Intelligence

Area of Science:

• Computational chemistry and drug discovery
• Artificial intelligence in medicine
• Peptide therapeutics

Background:

• Antimicrobial peptides (AMPs) show promise against resistant bacteria and biofilms.
• Current AMP design methods require optimization for efficiency and cost-effectiveness.
• Rational de novo design is crucial for developing peptides with desired properties.

Purpose of the Study:

• To develop a computational pipeline for rational de novo design of AMPs and antibiofilm peptides.
• To leverage an explainable artificial intelligence (XAI) framework combining Wasserstein Autoencoder (WAE) and Generative Topographic Mapping (GTM).
• To computationally screen and identify potent antimicrobial and antibiofilm peptides.

Main Methods:

• Utilized a WAE to learn latent representations of peptide space.
• Employed GTM to guide the generation of novel AMPs via 2D latent space maps.
• Integrated machine learning models for screening generated peptides based on predicted activity.

Main Results:

• The pipeline successfully generated peptides with verified antimicrobial and antibiofilm activity.
• Experimental validation against methicillin-resistant Staphylococcus aureus (MRSA) yielded a 100% hit rate for targeting biofilms.
• The most potent antibiofilm peptide showed a significant improvement in IC50 compared to the reference peptide '1018'.

Conclusions:

• The developed XAI-based computational pipeline enables efficient rational de novo design of AMPs and antibiofilm peptides.
• The pipeline demonstrates high efficacy in identifying potent antibiofilm agents with improved activity.
• This approach is adaptable for optimizing additional peptide properties, facilitating the development of peptide-based therapeutics.