FlowER: Mechanistic Chemical Reaction Datasets at the Elementary Step Level

Mechanistic datasets for chemical reactions are provided at the elementary step level. The archive contains train, validation, and test splits, each line formatted as atom-mapped SMILES reactants and products with an overall reaction index. Mun Hong Fong released this revised dataset on May 29, 2026.

Use Cases

• Training reaction prediction models based on atom-mapped SMILES sequences.
• Validating mechanistic reasoning in AI chemistry models based on elementary step data.
• Testing model generalization on corrected reaction templates mentioned in the description.

Strengths

• Dataset is split into train, validation, and test sets, facilitating standard ML workflows.
• Data is formatted with atom-mapped SMILES, which is a precise representation for chemical informatics.
• A 292.8 MB archive suggests a medium-scale collection of mechanistic data.

Limitations

• Row count is unknown, which may limit suitability assessment.
• Column-level documentation is absent; field semantics must be inferred after download.
• Description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

Mun Hong Fong

Freshness

Last updated 2026-05-29 22:16:30; freshness should be verified.