HemanthKari/rcsb-ligand-retrieval-v0.1

RCSB蛋白质-配体检索索引是一个公开许可的检索数据集，包含来自RCSB蛋白质数据库的72,388个蛋白质-配体复合物，专为基于结构的药物发现而设计。数据集包括蛋白质序列、配体SMILES、结合位点嵌入、配体嵌入（Morgan指纹）、FAISS HNSW索引、结合亲和力标签（来自BindingDB）以及规范化的训练/验证分割。所有组件均可商业再分发，适用于多种药物发现任务，如姿势预测、支架搜索和虚拟筛选基准构建。

The RCSB Protein-Ligand Retrieval Index is an openly licensed retrieval dataset of 72,388 protein-ligand complexes from the RCSB Protein Data Bank, built for retrieval-augmented structure-based drug discovery. It includes protein sequences, ligand SMILES, pocket embeddings, ligand embeddings (Morgan fingerprints), a FAISS HNSW index, binding affinity labels (from BindingDB), and canonical train/val splits. Every component is commercially redistributable and suitable for various drug discovery tasks such as pose prediction, scaffold search, and virtual screening benchmark construction.