scbirlab/steshin-2023-lohi

Lo-Hi Benchmark数据集是一个用于化学和生物学领域的机器学习基准测试数据集，主要用于药物发现中的命中识别和先导化合物优化任务。数据集包含多个子数据集，每个子数据集有三个分割的训练和测试文件。数据集的列包括SMILES字符串、实验值、结构标识符、分子量、Murcko骨架、Crippen LogP和拓扑极性表面积等信息。数据集由Simon Steshin整理，使用MIT许可证发布。

The Lo-Hi Benchmark dataset is a machine learning benchmark dataset for the fields of chemistry and biology, primarily used for hit identification and lead optimization tasks in drug discovery. The dataset includes multiple sub-datasets, each with three splits of training and test files. The columns of the dataset include SMILES strings, assay results, structure identifiers, molecular weight, Murcko scaffolds, Crippen LogP, and topological polar surface area. The dataset was curated by Simon Steshin and is released under the MIT license.