Molecular de-extinction of ancient antimicrobial peptides enabled by machine learning

Area of Science:

• Biochemistry
• Genomics
• Drug Discovery

Background:

• Bioactive molecules from extinct organisms offer potential for drug discovery.
• Antimicrobial peptides (AMPs) are a promising class of therapeutics.
• Identifying novel AMPs is crucial due to rising antibiotic resistance.

Purpose of the Study:

• To develop a machine learning model for predicting cleavage sites in proteomes.
• To prospect for encrypted antimicrobial peptides within human proteins, including archaic forms.
• To establish molecular de-extinction via paleoproteomics as a framework for antibacterial drug discovery.

Main Methods:

• Introduction of the panCleave random forest model for proteome-wide cleavage site prediction.
• Evaluation of panCleave against protease-specific classifiers using human caspases.
• In vitro testing of identified protein fragments for antimicrobial activity, proteolysis resistance, and membrane permeabilization.
• In vivo testing of protein fragments in murine models for skin abscess and thigh infections against Acinetobacter baumannii.

Main Results:

• The panCleave model demonstrated superior performance in cleavage site prediction compared to existing classifiers.
• Antimicrobial activity was confirmed in vitro for both modern and archaic protein fragments.
• Lead peptides exhibited resistance to proteolysis and varied membrane permeabilization capabilities.
• Modern and archaic protein fragments showed anti-infective efficacy against Acinetobacter baumannii in preclinical models.

Conclusions:

• Machine learning-based encrypted peptide discovery can identify stable, non-toxic peptide antibiotics.
• Molecular de-extinction through paleoproteome mining is a viable framework for discovering novel antibacterial agents.
• This approach holds significant potential for addressing the challenge of antibiotic resistance.