FlowER: Mechanistic Datasets for Chemical Reaction Prediction

FlowER datasets provide mechanistic data for retraining the FlowER model for chemical reaction prediction. The repository contains two datasets: one for full reproducibility of the original paper and a revised version with corrected templates. The data, released under an MIT license by author Tim Pinkhassik, is stored in ZIP files totaling 292.8 MB.

Use Cases

• Retraining the FlowER model based on the provided mechanistic datasets.
• Evaluating model performance on corrected mechanistic templates from the revised dataset.
• Studying chemical reaction mechanisms at the elementary step level using atom-mapped SMILES representations.
• Reproducing the results of the FlowER paper using the original mechanistic dataset.

Strengths

• Includes two distinct datasets: one for reproducibility and a revised version with corrected errors.
• Data is structured at the elementary step level with atom-mapped SMILES representations.
• Released under the permissive MIT license, facilitating reuse.

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

Tim Pinkhassik via figshare.

Collection Method

Mechanistic datasets generated for the FlowER model, likely derived from chemical reaction templates.

Freshness

Last updated 2026-05-29 22:29:37; freshness should be verified.