A python code for mechanistic dataset generation.

git clone https://github.com/jfjoung/Mechanistic_dataset.git
cd Mechanistic_dataset
conda env create --file environment.yml
conda activate mechdata

In order to generate mechanistic dataset from the overall reactions, reaction SMILES and its reaction class name as in NameRXN are required. The reaction SMILES and its reaction class name should be separated by a space as shown below. It doesn't require atom mapping, a code will automatically assign all the atom mapping number.

BrCCCOC1=CC(OCO2)=C2C=C1.OC3C(CCN(C3)C(OC(C)(C)C)=O)C4=CC=C(O)C=C4.[OH-]>>OC1C(CCN(C1)C(OC(C)(C)C)=O)C2=CC=C(OCCCOC3=CC(OCO4)=C4C=C3)C=C2 Williamson ether synthesis

This section outlines the configuration parameters used in the script, detailing their purpose and functionality.

• num_combination: Limits the number of reactant combinations per template to prevent combinatorial explosion.
• uni_rxn: Adds a unimolecular version of a bimolecular reaction template while retaining the original.
• proton: Ensures proton balance by adjusting templates on-the-fly based on the presence of acids or bases.
• max_num_temp: Limits the number of possible templates when modifications like uni_rxn or proton are applied.
• stoichiometry: Allows duplicated reactants to be added if stoichiometry errors are detected.
• do_not_pruning: Controls whether to prune the reaction network to retain only those that lead to the final product.
• num_cycles: Sets a limit on the number of catalytic cycles to avoid long computation times during conversion to SMARTS.
• num_reaction_node: Limits the number of reactions in the reaction network to prevent excessive processing time.
• byproduct: Adds byproducts from previous reactions when converting to elementary reaction SMARTS.
• spectator: Includes all chemical species in the flask at the current step when converting to SMARTS.
• full: Determines whether to return overall reaction SMARTS instead of elementary reaction SMARTS.
• end: Adds reactions that generate a product from an already existing product.
• plain: Generates reaction SMARTS without atom mapping numbers.
• explicit_H: Explicitly represents hydrogens in the reaction SMARTS.
• reagent: Positions non-reacting species between '>>' in Reaction SMARTS (Reactants > Reagents > Products).
• remapping: Remaps atom mapping in SMARTS to start from 1.
• data: Path to the file containing overall reactions.
• save: Path to the file where results will be saved.
• debug: Logs reactions with errors for debugging purposes.
• all_info: Saves all reactions, including the reaction network.
• stat: Generates statistics for the reactions.
• rxn_class: Specifies a particular reaction class to extract elementary reactions from.
• process: Number of processors to use in multiprocessing.
• verbosity: Sets the logging level.

Mechanistic data can be generated with either of the two methods:

1. with bash script
   generate_mechanism_data.sh can be modified with the arguments.

bash run/generate_mechanism_data.sh

2. with python script

python ver2.py

If you want to interactively browse and visualize the reaction templates (reaction class, mechanism, SMARTS for each step),
you can open the Colab notebook below:

This notebook allows you to:

• Select a reaction class and mechanism from dropdown menus
• View allowed and excluded reaction SMARTS for each mechanism
• Visualize each elementary reaction template
• Choose how atom features (like H count or charge) are shown in the molecule view