scbirlab/hudson-2023-dosedo

Dosedo数据集是一个包含了约3.7百万个多样性导向合成（DOS）化合物的SMILES字符串的集合，这些化合物来自一个DNA编码库。数据集已经经过规范化和分割，包含训练集、验证集和测试集，并提供了化合物的分子量和拓扑极性表面积等额外特征。该数据集旨在为开发化学模型提供支持，并为学术研究社区提供访问新颖结构特征的机会，以加速早期阶段的药物发现。

The Dosedo dataset is a collection of SMILES strings for approximately 3.7 million diversity-oriented synthesis (DOS) compounds from a DNA-encoded library. The dataset has been canonicalized and split into training, validation, and test sets, and includes additional features such as molecular weight and topological polar surface area. It is designed to support the development of chemical models and provide the academic research community with access to novel structural features to accelerate early-phase drug discovery.