IQuestLab/UBio-MolLR25

该数据集用于论文《通过长程感知消息传递实现大分子系统的可扩展机器学习力场》中。数据集包含通过密度泛函理论（DFT）计算生成的数据，确保物理准确性用于机器学习训练。数据集包含三个主要部分：1) deshaw_protein.tar.gz，包含从DE Shaw Research分子动力学轨迹中提取的蛋白质结构和构象数据；2) di_molecule_interaction.tar.gz，专注于非键相互作用，通过系统增加两个不同分子（二聚体）之间的距离构建；3) md_traj.tar.gz，包含内部生成的各种分子动力学（MD）轨迹。数据以LMDB格式存储，包含原子力、原子位置、总分子能量等关键信息。

This dataset is used in the paper "Scalable Machine Learning Force Fields for Macromolecular Systems Through Long-Range Aware Message Passing". It contains data generated using Density Functional Theory (DFT) calculations to ensure physical accuracy for machine learning training. The dataset consists of three primary components: 1) deshaw_protein.tar.gz, containing protein structures and conformational data extracted from DE Shaw Research molecular dynamics trajectories; 2) di_molecule_interaction.tar.gz, focusing on non-bonded interactions, constructed by systematically increasing the distance between two distinct molecules (dimers); 3) md_traj.tar.gz, containing various molecular dynamics (MD) trajectories generated in-house. The data is stored in LMDB format and includes key information such as atomic forces, atomic positions, and total molecular energy.