MaRIboES

MaRIboES models enzyme aspecificity to predict potential transformations of input compounds by specific enzymes, leveraging BRENDA enzyme activity data to expand known reaction space beyond databases such as KEGG.

Key Features:

• Enzyme Aspecificity Modeling: Predicts whether an input compound is likely to be transformed by a given enzyme, accounting for enzyme promiscuity.
• Structural and Stereochemistry Analysis: Employs structural similarity measures and stereochemical considerations to generalize enzymatic reactions.
• Molecular Fingerprints Utilization: Uses molecular fingerprints to enhance the accuracy of reaction predictions and identify novel compounds.
• Integration of BRENDA Data: Leverages enzyme activity data from the BRENDA database to ground and expand prediction rules.
• Cross-Validation Performance: Demonstrates predictive performance evaluated by leave-one-out cross-validation, recovering approximately 80% of known reactions.

Scientific Applications:

• Metabolic Network Completion: Predicts additional enzymatic reactions to fill gaps in reconstructed metabolic networks.
• Metabolic Engineering: Identifies potential enzyme–substrate interactions to guide design of engineered metabolic pathways.
• Mass Spectrometry Analysis: Aids interpretation of mass spectrometry data by suggesting novel metabolites formed through predicted enzymatic reactions.

Methodology:

Integrates structural similarity and stereochemistry measures with molecular fingerprints and uses BRENDA enzyme activity data for reaction generalization, with evaluation by leave-one-out cross-validation.