SynEPD: a curated mechanistic data resource with electron-pushing annotations and hierarchical taxonomy for polar organic reactions

<title>Abstract</title> <p>Electron-pushing diagrams describe how valence electron pairs move during organic reactions, yet curated datasets that connect atom-mapped transformations with explicit electron flow remain scarce. We present SynEPD, a data resource for polar organic reaction mechanisms that links each record to a hierarchical POLAR taxonomy, an Imaginary Transition State graph, a net reaction center (RC), a mechanistic center (MC), and structured electron-pushing diagram (EPD) annotations. The SynEPD dataset contains 1,926 mass- and charge-balanced reactions, 939 active taxonomy nodes, 1,521 distinct RC templates, 1,540 distinct MC templates, 2,277 molecular structures, and 8,123 electron-pushing arrows. Arrows are encoded as ordered two-electron movements between atom-map indices using an eight-type donor-to-acceptor vocabulary. A crosswalk links 189 POLAR taxa to the Name Reaction Ontology (RXNO) through 198 typed assertions. The data product comprises a relational SQLite database, source records, taxonomy and ontology crosswalk files, a checksummed release manifest, and rebuilding scripts. Deterministic validation verifies reaction balance, atom mapping, explicit mechanistic hydrogens, database integrity, and executable EPD replay. All 1,926 records pass these checks. By separating net transformation topology, reusable EPD-aware topology, and reaction-specific event order, SynEPD supports mechanism retrieval, RC/MC-template mining, ontology-linked comparison, and development of predictive models for electron-pushing mechanisms.</p>