vector-institute/atom3d-msp

Mutation Stability Prediction (MSP)任务涉及使用提供的蛋白质结构对SKEMPI 2.0数据库中的突变进行分类，判断其是否稳定。数据集包含4148个突变结构和316个野生型结构，每个突变包括一个PDB文件和一个原生PDB文件。数据集通过Kd值的变化来标记突变是否稳定，Kd值小于野生型蛋白的突变标记为1，否则为0。数据集分为训练集、验证集和测试集，每个数据项包含input_ids、coords、labels和token_type_ids等特征。

The Mutation Stability Prediction (MSP) dataset is designed for classifying whether mutations in the SKEMPI 2.0 database are stabilizing or not, using provided protein structures. It includes 4148 mutant structures and 316 wild-type (WT) structures. The dataset excludes non-point mutations, mutations causing non-binding, those involving disulfide bonds, and specific PDBs due to processing difficulties. Each item in the dataset contains input_ids (atomic numbers), coords (3D coordinates), labels (binding classification), and token_type_ids (mask for atom types). The dataset is split into train, validation, and test sets with specified sizes and configurations.