DynamicGT: a dynamic-aware geometric transformer model to predict protein binding interfaces in flexible and disordered regions

This dataset contains conformational Representatives of protein structures derived from two sources: Molecular Dynamics (MD) Simulations (three replicates of 500 microseconds for each entry) – Representative structures were selected based on RMSD clustering, with dynamically chosen cutoffs to retain between 20 to 50 representatives per trajectory. AlphaFlow-Generated Conformations – A distilled version of AlphaFlow was used to generate 25 conformations per protein.本数据集包含两类来源的蛋白质结构构象代表性样本： 1. 分子动力学（Molecular Dynamics, MD）模拟数据：每个条目包含3组重复模拟，每组模拟时长为500微秒；通过均方根偏差（Root Mean Square Deviation, RMSD）聚类选取代表性结构，采用动态设定的截断阈值，每条模拟轨迹保留20至50个代表性结构。 2. AlphaFlow生成构象数据：采用经蒸馏的AlphaFlow版本为每个蛋白质生成25个构象。