Hierarchical machine learning model predicts antimicrobial peptide activity against Staphylococcus aureus

Area of Science:

• Computational Biology and Cheminformatics
• Infectious Diseases and Microbiology

Background:

• Staphylococcus aureus is a significant pathogen responsible for numerous infections.
• Antimicrobial peptides (AMPs) show promise as novel antibiotics against multi-drug-resistant bacteria like S. aureus.
• Existing AMP classification tools lack species-specific focus, limiting their practical application.

Purpose of the Study:

• To develop a hierarchical machine learning model for classifying antimicrobial peptides with activity against Staphylococcus aureus.
• To address the limitations of general AMP classifiers by creating a species-specific tool.

Main Methods:

• Utilized an up-to-date dataset for training and validation.
• Developed a two-level hierarchical machine learning model.
• First level: Classifies peptides as antimicrobial peptides (AMPs) or non-AMPs. Second level: Classifies AMPs based on activity against S. aureus.

Main Results:

• The hierarchical model demonstrated effectiveness in classifying S. aureus-specific AMPs.
• Physicochemical properties were identified as key features after selection.
• The final model achieved an F1-score of 0.80, recall of 0.86, balanced accuracy of 0.80, and specificity of 0.73 on the test set.

Conclusions:

• A hierarchical machine learning approach is superior for identifying species-specific antimicrobial peptides.
• This model provides a practical tool for screening peptide libraries for potential S. aureus-targeting agents.
• The findings highlight the importance of species-specific evaluation for antimicrobial peptide efficacy.