Derify/ReactionSmiles

---
license: cc-by-4.0
tags:
- chemistry
- molecules
- reactions
- smiles
- cheminformatics
size_categories:
- 1M<n<10M
task_categories:
- text-generation
- fill-mask
pretty_name: Reaction SMILES Dataset
dataset_info:
features:
- name: smiles
dtype: string
- name: source_id
dtype: int64
splits:
- name: train
num_bytes: 747528582
num_examples: 2752563
- name: validation
num_bytes: 93440971
num_examples: 344070
- name: test
num_bytes: 93441242
num_examples: 344071
download_size: 406610470
dataset_size: 934410795
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
- split: validation
path: data/validation-*
- split: test
path: data/test-*
---

# Reaction SMILES Dataset

A collated dataset of 3.4M unique chemical reaction SMILES strings compiled from multiple public sources for use in pre-training and fine-tuning chemical language models.

Reaction SMILES (Simplified Molecular Input Line Entry System) extend the standard SMILES notation to represent complete chemical reactions. They encode reactants, reagents/catalysts, and products in a single text string using the `>` delimiter:

```text
reactants>reagents>products
```

For example: `CC(=O)O.CCO>[H+]>CC(=O)OCC.O` represents the esterification of acetic acid with ethanol to form ethyl acetate.

## Dataset Features

- **smiles**: The reaction SMILES string
- **source_id**: Original data source identifier

## Data Sources

| Source ID | Source URL |
| --------- | -------------------------------------------------------------------------------------------------------------------------------------- |
| 1 | [US Patents 1976-Sep2016 Grants](https://figshare.com/articles/dataset/Chemical_reactions_from_US_patents_1976-Sep2016_/5104873) |
| 2 | [US Patents 2001-Sep2016 Applications](https://figshare.com/articles/dataset/Chemical_reactions_from_US_patents_1976-Sep2016_/5104873) |
| 3 | [CRD 1.37M Dataset (2024)](https://figshare.com/articles/dataset/Reaction_SMILES_b_CRD_b_1_37M_dataset/28230053/1) |
| 4 | [USPTO Year 2023](https://figshare.com/articles/dataset/Reaction_SMILES_USPTO_year_2023/24921555) |
| 5 | [Reaction SMILES Dataset (2023)](https://figshare.com/articles/dataset/Reaction_SMILES_dataset/22491730) |

## License

This dataset aggregates publicly available reaction data. Please refer to the individual source links for specific licensing terms.