proteinglm/stability_prediction

Stability Stability数据集的主要任务是预测蛋白质在蛋白酶作用下保持折叠状态的浓度。数据集包含蛋白质序列字符串和稳定性评分浮点数两个特征，分为训练集、验证集和测试集三个部分。数据来源于Rocklin等人的研究，并通过TAPE项目收集。数据集采用Apache-2.0许可证发布，并提供了相关的引用信息。

The Stability Stability task is to predict the concentration of protease at which a protein can retain its folded state. The dataset contains protein sequences (seq) and stability scores (label), represented as strings and float numbers, respectively. The dataset is divided into train, valid, and test sets, containing 53,614, 2,512, and 12,851 instances, respectively. The average sequence length is 45, and the average stability score is 0.34. The dataset is sourced from the research of Rocklin et al. and further collected within the TAPE project. The dataset is released under the Apache-2.0 License.