Design of multimodal antibiotics using deep learning.

Area of Science:

• Biotechnology
• Infectious Diseases
• Artificial Intelligence

Background:

• Antimicrobial resistance (AMR) poses a significant global health threat, rendering many conventional antibiotics ineffective.
• Intracellular bacterial infections are challenging to treat as antibiotics struggle to reach pathogens within host cells.
• Designing molecules with both cell-penetrating and antimicrobial capabilities is a major hurdle in drug development.

Purpose of the Study:

• To develop a novel artificial intelligence (AI) model, ApexDuo, capable of generating peptides with dual cell-penetrating and antimicrobial properties.
• To address the limitations of conventional antibiotics against intracellular pathogens.

Main Methods:

• ApexDuo, a multimodal AI model, was employed to generate a library of 50 million peptide candidates.
• Selected peptides were characterized for their cell-penetrating and antimicrobial activities.
• The lead peptide, Turingcin-46, was tested against intracellular Staphylococcus aureus and Listeria monocytogenes.
• Efficacy of Turingcin-46 was evaluated in mouse models of skin abscess and peritonitis.

Main Results:

• ApexDuo successfully generated peptides with combined cell-penetrating and antimicrobial functions.
• The lead compound, Turingcin-46, demonstrated effective penetration into mammalian cells.
• Turingcin-46 significantly reduced intracellular Staphylococcus aureus and Listeria monocytogenes burdens.
• In vivo studies showed Turingcin-46 reduced bacterial loads by up to two orders of magnitude in mouse models.

Conclusions:

• The ApexDuo AI model provides a powerful platform for designing multifunctional peptide antibiotics.
• Turingcin-46 represents a promising therapeutic candidate for treating intracellular bacterial infections.
• This approach opens new avenues for developing next-generation antimicrobials to combat resistance.