Machine Learning-Identified Potent Antimicrobial Peptides Against Multidrug-Resistant Bacteria and Skin Infections

Area of Science:

• Biochemistry
• Computational Biology
• Drug Discovery

Background:

• Antibiotic resistance is a global health crisis requiring new antimicrobial agents.
• Antimicrobial peptides (AMPs) offer a promising alternative to combat multidrug-resistant (MDR) pathogens.
• Machine learning (ML) accelerates AMP discovery, overcoming limitations of traditional methods.

Purpose of the Study:

• To apply ML for identifying novel AMPs effective against MDR bacteria and skin infections.
• To develop a catalogue of potential therapeutic peptides using the CalcAMP model.
• To validate the efficacy of novel peptide candidates against specific bacterial strains and infection models.

Main Methods:

• Utilized the ML-based CalcAMP model to predict antimicrobial activity of 16,384 peptide sequences.
• Generated a novel Guided Designed Smart Therapeutic (GDST) peptide catalogue.
• Tested GDST peptides and their retro-inverso (RI) variants against MDR bacteria and in skin infection models.

Main Results:

• GDST-038 and GDST-045, with RI variants, demonstrated potent activity against Acinetobacter baumannii and Staphylococcus aureus.
• Peptides rapidly depolarized bacterial membranes, showing broad-spectrum bactericidal effects against ESKAPE pathogens with minimal haemolysis.
• RI variants effectively reduced A. baumannii biofilms, while all GDST peptides significantly reduced S. aureus biofilms; efficacy was confirmed in a 3D skin model.

Conclusions:

• ML-driven screening successfully identified two novel candidate AMPs (GDST peptides).
• GDST peptides exhibit significant therapeutic potential for treating MDR bacterial infections.
• The study highlights the efficiency of ML in accelerating the discovery of effective antimicrobial therapeutics.