SkinBug

SkinBug predicts the metabolism and biotransformation of molecules by the human skin microbiota to elucidate how cosmetics, medicines, and pollutants are processed across 19 distinct skin sites.

Key Features:

• Curated database: Contains 1,094,153 metabolic enzymes, reactions, and substrates derived from approximately 900 bacterial species across 19 skin sites.
• Computational methods: Employs machine learning algorithms, neural networks, and chemoinformatics to analyze and predict metabolic transformations.
• Predictive performance: Reports multiclass multilabel accuracy up to 82.4% and binary accuracy up to 90.0% for metabolic reaction predictions.
• Comprehensive reaction prediction: Identifies reaction centers, corresponding enzymes, implicated microbial species, and specific skin sites for any input molecule.

Scientific Applications:

• Exposome studies: Predicts microbiome-mediated transformations of cosmetics, medicines, and pollutants to inform skin exposome analyses.
• Microbiome research: Facilitates characterization of metabolic capabilities and functional potential of skin-associated bacterial species across sites.
• Dermatology: Identifies potential interactions between skincare products or therapeutics and the skin microbiome relevant to formulation and safety assessment.
• Skin cancer research: Elucidates microbiome-mediated metabolism of xenobiotics on skin that may influence skin cancer risk factors.

Methodology:

Prediction is based on a curated database of 1,094,153 metabolic enzymes, reactions, and substrates from ~900 bacterial species across 19 skin sites, analyzed using machine learning algorithms, neural networks, and chemoinformatics.