BioTransformer

BioTransformer predicts metabolic transformations of small molecules across human tissues, the human gut microbiota, and environmental microbiota (soil and water) by integrating machine-learning and rule-based approaches to generate metabolite and enzyme predictions.

Key Features:

• Machine learning and rule-based integration: Integrates machine-learning techniques with rule-based systems to predict metabolic transformations.
• Biological contexts: Predicts metabolism in human tissues, the human gut microbiota, and environmental microbiota in soil and water.
• Input formats: Accepts molecular structures in SMILES and SDF formats.
• Predicted outputs: Produces predicted metabolites/transformation products and the enzymes anticipated to catalyze reactions.
• Sequential transformations: Supports customized sequential combinations of transformations and multiple iterations to simulate multi-step biotransformation events.
• Endogenous and xenobiotic coverage: Predicts transformation products for xenobiotics and endogenous metabolites including amino acids, peptides, carbohydrates, organic acids, and lipids.
• Databases and modules: Incorporates expanded databases, constraints, and prediction modules.
• Accuracy improvement: Demonstrates a reported 40–50% increase in prediction accuracy compared to earlier versions.
• Combinatorial control: Reduces combinatorial "explosions" while increasing metabolite coverage.

Scientific Applications:

• Human drug and xenobiotic metabolism: Predicts human tissue and gut microbiota-mediated metabolism of xenobiotics to support metabolic fate analysis.
• Endogenous metabolite transformation: Predicts biotransformation of endogenous metabolites such as amino acids, peptides, carbohydrates, organic acids, and lipids.
• Environmental biodegradation: Predicts transformations mediated by environmental microbiota in soil and water.
• Multi-step biotransformation simulation: Models sequential and iterative metabolic steps to represent complex human biotransformation events.
• Mechanistic interpretation: Provides predicted enzyme assignments to inform mechanistic hypotheses about metabolic reactions.

Methodology:

Integrates machine-learning techniques with rule-based systems and employs expanded databases, constraints, and prediction modules while supporting customized sequential combinations of transformations and multiple iterations.