How angusketo/pericyclictl works
Project type and purpose
• This is not a web app or microservice; it’s a data‐generation and processing pipeline built as a collection of Jupyter notebooks.
File structure
• Extract_Cope_and_Claisen.ipynb, Extract_Ene.ipynb, Extract_Nazarov.ipynb
– Each notebook is one pipeline “stage” dedicated to a reaction type.
• Show_Prediction_Results.ipynb
– A visualization or reporting notebook to display or validate the extracted data.
• nerfenv.yml
– The Conda environment specification listing RDKit (2020.09.1 or later), RXNMapper, and other Python dependencies.
• README.md
– High‐level overview, prerequisites (Reaxys CSV), external links (DA_DataExtraction, RXNMapper), notes on SMILES canonicalization changes.
README for additional insights
• Confirms that inputs are externally sourced CSVs from Reaxys.
• Emphasizes dependency on RDKit version and optional use of RXNMapper.
• Mentions that Diels–Alder is out of scope (linked externally).
a. Main components
1. External Data Source
• Reaxys database (subscription) → raw CSV export
2. Local Data Store
• Folder with downloaded .csv files
3. Conda Environment Container
• nerfenv.yml defining RDKit, RXNMapper, Python, etc.
4. Notebook Stages
• “Extract_Cope_and_Claisen”
• “Extract_Ene”
• “Extract_Nazarov”
5. Atom‐Mapping Service (Optional)
• RXNMapper API or local invocation
6. Output Store
• Processed CSVs, pickles, RDKit‐normalized SMILES
7. Reporting/Visualization
• “Show_Prediction_Results.ipynb”
b. Relationships and interactions
• Extraction notebooks output to the Output Store
• Show_Prediction_Results reads from Output Store and generates visualizations
c. Architectural patterns and principles
• Linear or branching data pipeline: raw → processed → visualized
• Environment as code: clear separation of runtime dependencies via nerfenv.yml
• Modular stages (one notebook per reaction type) to isolate concerns
d. Key technologies and frameworks
• Jupyter Notebooks (pipeline orchestration)
• RDKit for cheminformatics and SMILES canonicalization
• RXNMapper for atom mapping
• Conda (environment reproducibility)
Detail without overthinking
• Every notebook as a distinct component.
• Both external inputs (Reaxys) and outputs (processed data, visual reports).
• Do not merge notebooks—you want clear separation so future contributors can see where to add new reaction‐type stages.