alessandronascimento/zinc20_chembl36

---
tags:
- SMILES
- chemistry
- SELFIES
- ZINC
- CHEMBL
size_categories:
- 10M<n<100M
---
# Dataset Card for ZINC20_Chembl36

This dataset contains about 13.6M molecules from ZINC20 and from Chembl26. The molecules are provided as SMILES and SELFIES, as well as tokens using MolGen tokenizer.

### ZINC20

The 2D tranches from ZINC20 were used and the *In Stock* molecules were selected and downloaded. The tranches were concatenated and converted to a dataset. Afterwards, the selfies library was used to convert SMILES to SELFIES.

### Chembl36

Chembl36 molecules were downloaded and converted to selfies using the selfies library. The SDF file was used to generate molecules in smiles with rdkit.

## Dataset Details

- **Curated by:** Alessandro S. Nascimento
- **License:** MIT

### Dataset Sources

- **Web pages:** [ZINC20](https://zinc.docking.org/) and [Chembl](https://www.ebi.ac.uk/chembl/)
- **Tokenizer used:** zjunlp/MolGen-large

## Dataset Split Strategy

The train/test/validation splits were generated randomly using a 80/10/10% ratio.

---
标签：
- 简化分子线性输入规范（SMILES）
- 化学
- SELFIES
- ZINC
- CHEMBL
规模类别：
- 1000万 < 样本量 < 1亿
---
# ZINC20_Chembl36 数据集卡片
本数据集包含来自ZINC20与Chembl26的约1360万条分子数据。这些分子以简化分子线性输入规范（SMILES）、SELFIES格式，以及通过MolGen分词器（MolGen tokenizer）生成的Token形式提供。

## ZINC20
本部分使用了ZINC20的二维数据集分块，并筛选并下载了其中的**现货**分子。将这些分块拼接后转换为数据集，随后借助SELFIES库将SMILES格式转换为SELFIES格式。

## Chembl36
下载Chembl36的分子数据，并通过SELFIES库将其转换为SELFIES格式。借助RDKit工具从SDF文件中生成SMILES格式的分子数据。

## 数据集详情
- **数据整理者：** Alessandro S. Nascimento
- **授权协议：** MIT

### 数据集来源
- **来源网页：** [ZINC20](https://zinc.docking.org/) 与 [Chembl](https://www.ebi.ac.uk/chembl/)
- **所用分词器：** zjunlp/MolGen-large

## 数据集划分策略
训练集/测试集/验证集以80/10/10%的比例随机划分生成。