EHreact

EHreact extracts and scores enzymatic reaction templates from extensive reaction databases using a Python implementation of extended Hasse diagrams to support data-driven computer-aided synthesis planning and prediction of enzyme–substrate activity.

Key Features:

• Data-driven extraction: Leverages large reaction datasets to automatically derive reaction rules without expert-curated templates, addressing substrate promiscuity.
• Hasse-diagram organization: Organizes extracted reaction rules into extended Hasse diagram tree structures based on common substructures identified in imaginary transition states.
• Template scoring and specificity: Scores reaction templates at enzyme-specific levels of specificity and determines the optimal number of atoms to include per rule to reflect substrate promiscuity.
• Heuristic prediction: Infers enzyme reactive-site information from known reactions and their organization within the template tree to predict activity on new substrates.
• Extraction modes: Outputs single or multiple reaction rules from each Hasse diagram.
• Implementation: Implemented as a Python package for computational extraction and scoring of enzymatic reaction templates.

Scientific Applications:

• Computer-aided synthesis planning: Provides reaction templates and specificity scores for data-driven planning of organic and biocatalyzed syntheses.
• Biocatalytic pathway exploration: Enables exploration of new biocatalytic pathways by generating and scoring reaction rules for novel substrates.
• Enzyme substrate-scope analysis: Assists in defining and comparing enzyme substrate scopes to guide enzyme selection or optimization.
• Automated rule generation: Reduces reliance on manually curated datasets by extracting reaction templates from extensive databases.

Methodology:

Extracts reaction templates from reaction databases; constructs extended Hasse diagrams based on common substructures in imaginary transition states; scores templates to select appropriate atom inclusion levels per enzyme; heuristically predicts enzyme activity by inferring reactive-site information from known reactions and the template tree; implemented in Python and applied to large datasets.