LOBSTER (Ligand Overlays from Binding SiTe Ensemble Representatives)

LOBSTER ("Ligand Overlays from Binding SiTe Ensemble Representatives") is a dataset of ligand overlays designed to evaluate small molecule superposition tools. Based on all structures from the RCSB PDB, the dataset generation and filtering protocols are fully automated to avoid subjectivity in the selection of protein-ligand complexes and to gain the largest possible set of refined compounds. Affinity and activity data have been processed to select ligands with a high ligand efficiency.Ligands were superimposed in their crystal pose by aligning the corresponding binding pockets to so-called ensembles. For poses generated in benchmark experiments, this offers an objective comparison to the superimposed ligand crystal poses. A clustering of ensembles created with the same protein-ligand complexes ensures the diversity of the LOBSTER set.The 671 ligand ensembles comprise a total of 3212 unique ligands from 3521 different protein-ligand complexes. A total of 72 734 ligand pairs have been derived from the ensembles. Ten subsets were generated from the pairs according to the shape overlap of the pairs, quantified by the Shape Tversky Index.LOBSTER（全称Ligand Overlays from Binding SiTe Ensemble Representatives，即结合位点集合代表物配体叠合数据集）是一款专为评估小分子叠合工具构建的配体叠合数据集。 该数据集的生成与过滤流程完全自动化，其数据覆盖RCSB蛋白质数据库（RCSB PDB）的全部结构，旨在规避蛋白质-配体复合物筛选环节的主观性，并获取尽可能多的精制化合物。研究人员通过处理亲和力与活性数据，筛选出配体效率较高的配体。配体通过将对应结合口袋对齐至指定结合位点集合，以其晶体构象完成叠合；对于基准实验中生成的构象而言，该方式可实现与配体晶体叠合构象的客观对比。针对使用同一蛋白质-配体复合物构建的结合位点集合进行聚类操作，确保了LOBSTER数据集的多样性。 该数据集包含671个配体集合，共涵盖来自3521个不同蛋白质-配体复合物的3212种独特配体，从这些集合中共衍生出72734对配体。研究人员依据配体对的形状重叠程度（通过形状特沃斯基指数（Shape Tversky Index）量化），将全部配体对划分为10个子集。