Relevant Features of Polypharmacologic Human-Target Antimicrobials Discovered by Machine-Learning Techniques

Area of Science:

• Microbiology
• Computational Biology
• Pharmacology

Background:

• Complex diseases like diabetes and hypertension are linked to the human microbiome.
• Polypharmacologic human-targeted antimicrobials (polyHAM) offer potential therapeutic strategies.
• Previous work utilized machine learning to identify polyHAM from FDA-approved drugs.

Purpose of the Study:

• To identify polyHAM with broad-spectrum antibiotic activity.
• To determine if topological or chemical features are more informative for classifying antimicrobial activity.
• To develop a machine learning model capable of identifying antimicrobial human metabolites.

Main Methods:

• A heterologous machine-learning approach was trained using broad-spectrum antimicrobials.
• The model was tested on human metabolites, classified as antimicrobial or non-antimicrobial via text mining.
• Topological features were found to be more informative than chemical features for classification.

Main Results:

• PolyHAM are more prevalent among antimicrobials with broad-spectrum antibiotic activity.
• A heterologous machine learning model successfully classified human metabolites with antimicrobial activity.
• Topological features were identified as key predictors of antimicrobial classification.

Conclusions:

• Machine learning can identify antimicrobial properties of human metabolites.
• These findings support the development of dietary interventions to control the human microbiome.
• Targeting human metabolites offers a novel approach to managing microbiome-associated diseases.