---
license: mit
dataset:
- chembl
- zinc15
---
## Overview
This dataset contains Functional Group (FG)-enhanced SMILES designed for molecule representation learning. It consists of approximately **20 million** FG-enhanced SMILES collected from a wide range of chemical suppliers and databases, as detailed below:

| **Supplier**                  | **Number of Compounds** | **Source**                                               |
|-------------------------------|-------------------------|----------------------------------------------------------|
| Targetmol                     | 22,555                  | [Targetmol](https://www.targetmol.com/)                  |
| Chemdiv                       | 1,741,620               | [Chemdiv](https://www.chemdiv.com/)                      |
| Enamine                       | 862,698                 | [Enamine](https://enamine.net/)                          |
| Life Chemical                 | 347,657                 | [Life Chemicals](https://lifechemicals.com/)             |
| Chembridge                    | 1,405,499               | [Chembridge](https://chembridge.com/)                    |
| Vitas-M                       | 1,430,135               | [Vitas-M](https://vitasmlab.biz/)                        |
| InterBioScreen                | 560,564                 | [InterBioScreen](https://www.ibscreen.com/)              |
| Maybridge                     | 97,367                  | [Maybridge](https://chembridge.com/)                     |
| Asinex                        | 601,936                 | [Asinex](https://www.asinex.com/)                        |
| Eximed                        | 61,281                  | [Eximed](https://eximedlab.com/)                         |
| Princeton BioMolecular        | 1,647,078               | [Princeton BioMolecular](https://princetonbio.com/)      |
| Otava                         | 9,203,151               | [Otava](https://www.otava.com/)                          |
| Alinda Chemical               | 733,152                 | [Alinda Chemical](https://www.alinda.ru/synthes_en.html) |
| ChEMBL 25                     | 1,785,415               | [ChEMBL](https://www.ebi.ac.uk/chembl/)                  |
| ZINC15                        | 4,000,000               | [ZINC15](https://zinc15.docking.org/)                    |
| **Total**                     | **20,000,000**          |                                                          |

## Dataset Details
This dataset contains **36,748** unique tokens (vocabulary size = 36,748), which is an extension from 93 tokens in the corresponding standard SMILES dataset. This expansion helps bridge the gap between SMILES and natural language, which has a large set of vocabulary to sufficiently express the nuances of language.

For more details about the method of generating FG-enhanced SMILES, please refer to our paper, [FARM: Functional Group-Aware Representations for Small Molecules](https://arxiv.org/pdf/2410.02082), or visit our [GitHub repository](https://github.com/thaonguyen217/farm_molecular_representation) for implementation details.

Below is a breakdown of the number of token types that represent different chemical elements:
![image](output.png)
*Number of functional groups associated with different chemical elements in the FG-enhanced SMILES dataset. The y-axis represents the natural logarithm (log, base e) of the count.*

## Examples of FG-enhanced SMILES
Below are some examples of FG-enhanced SMILES from the dataset. Each SMILES string is augmented with functional group annotations (e.g., `C_alkyl`, `O_ether`, `c_6`, `n_tertiary_amine`, etc.) to enhance the representation of molecular structures for machine learning models.

1. ```C_alkyl O_ether c_6 1 c_6 c_6 c_6 ( - c_5-6 2 n_5-6 c_5-6 3 n_tertiary_amine_5-6 ( C_alkyl c_6 4 c_6 c_6 c_6 c_6 c_6 4 F_fluoro ) c_5-6 ( C_alkyl ) c_5-6 ( - c_5-6 4 c_5-6 c_5-6 c_5-6 5 c_5-6 ( c_5-6 4 ) O_ether_5-6 C_5-6 O_ether_5-6 5 ) c_amide_5-6 ( = O_amide ) n_amide_5-6 3 c_5-6 2 C_alkyl N_tertiary_amine ( C_alkyl ) C_alkyl C_alkyl c_6 2 c_6 c_6 c_6 c_6 c_6 2 ) c_6 c_6 1```

2. ```C_alkyl C_tertiary_carbon ( C_alkyl ) C_tertiary_carbon ( C_ester ( = O_ester ) O_ester C_ester c_5-6 1 c_5-6 n_tertiary_amine_5-6 2 c_5-6 c_5-6 c_5-6 c_5-6 c_5-6 2 n_5-6 1 ) C_tertiary_carbon ( C_alkyl ) N_secondary_amine C_ester ( = O_ester ) O_ester C_ester ( C_alkyl ) ( C_alkyl ) C_alkyl```

3. ```C_alkyl C_tertiary_carbon_6 1 C_6 N_tertiary_amine_6 ( c_5 2 n_5 n_5 c_5 ( C_tertiary_carbon_3 3 C_3 C_3 3 ) n_tertiary_amine_5 2 C_alkyl C_alkyl N_secondary_amine C_ketone ( = O_ketone ) c_5 2 c_5 c_5 c_5 s_5 2 ) C_6 C_amide_6 ( = O_amide ) N_amide_6 1 C_alkyl```

4. ```C_alkyl O_ether C_alkyl C_alkyl n_tertiary_amine_5 1 c_5 ( C_alkyl c_5 2 c_5 c_5 ( C_alkyl ) n_secondary_amine_5 n_5 2 ) n_5 n_5 c_5 1 N_tertiary_amine_7 1 C_7 C_7 C_7 N_amide_7 ( C_amide ( = O_amide ) C_tertiary_carbon ( C_alkyl ) C_alkyl ) C_7 C_7 1```

5. ```C_alkyl C_alkyl n_amide_6-6 1 c_amide_6-6 ( = O_amide ) n_6-6 c_6-6 ( O_hydroxyl ) c_6-6 2 c_6-6 ( C_ketone ( = O_ketone ) N_secondary_amine N_tertiary_amine_5 3 C_5 C_5 C_5 C_tertiary_carbon_5 3 C_alkyl O_ether C_alkyl ) c_6-6 c_6-6 ( C_tertiary_carbon ( C_alkyl ) C_alkyl ) n_6-6 c_6-6 1 2```

6. ```C_alkyl C_tertiary_carbon_3 1 C_3 C_tertiary_carbon_3 1 C_alkyl n_tertiary_amine_5 1 c_5 ( C_alkyl c_6 2 c_6 c_6 n_6 c_6 c_6 2 ) n_5 n_5 c_5 1 N_tertiary_amine_6 1 C_6 C_6 N_tertiary_amine_6 ( C_alkyl C_amide ( = O_amide ) N_amide ( C_alkyl ) C_alkyl ) C_6 C_6 1```