D2R-MD

D2R-MD 数据集包含人类多巴胺 D2 受体在膜环境中的 2 微秒分子动力学轨迹（MD），共 12000 帧。该数据集用于训练和评估 LD-FPG 框架，该框架旨在从现有的 MD 模拟数据中学习和生成特定蛋白质的多样化、全原子构象集，包括侧链细节。LD-FPG 框架利用 Chebyshev 图神经网络 (ChebNet) 将蛋白质构象编码为低维潜在嵌入，然后使用扩散模型生成新的样本，并由解码器将潜在表示解码回全原子笛卡尔坐标。D2R-MD 数据集为探索可切换蛋白质的复杂动力学提供了一个计算高效的工具，支持基础机制研究和药物发现应用。

The D2R-MD dataset contains a 2-microsecond molecular dynamics (MD) trajectory of the human dopamine D2 receptor in a membrane environment, with a total of 12,000 frames. This dataset is used to train and evaluate the LD-FPG framework, which aims to learn and generate diverse, all-atom conformational ensembles for specific proteins from existing MD simulation data, including side-chain details. The LD-FPG framework utilizes the Chebyshev Graph Neural Network (ChebNet) to encode protein conformations into low-dimensional latent embeddings, then uses a diffusion model to generate new samples, and employs a decoder to decode the latent representations back to all-atom Cartesian coordinates. The D2R-MD dataset provides a computationally efficient tool for exploring the complex dynamics of switchable proteins, supporting fundamental mechanistic studies and drug discovery applications.