Refined AI-predicted Models

This record is of the same structure as the relaxed AI-predicted models, however, it contains the refined version of the energy minimized (relaxed) structural models. Refinement includes, as described in the Methods section, the removal of low confidence regions with a pLDDT score below 70, and filtering for models with 10 or more amino acid residues remaining. These folders include, as individual PDB files, 39,484 refined AlphaFold 2 models, 39,748 refined OpenFold models, and 31,870 refined ESMFold models.

本数据集的结构与弛豫（relaxed）AI预测模型保持一致，但其收录的是经能量最小化（弛豫）处理后的结构模型的精修版本。正如方法章节所述，本次精修流程包括移除pLDDT（Predicted Local Distance Difference Test）评分低于70的低置信度区域，并筛选出剩余氨基酸残基数不少于10的模型。该数据集的文件夹以独立PDB（Protein Data Bank）文件的形式，收录了39,484个精修后的AlphaFold 2模型、39,748个精修后的OpenFold模型以及31,870个精修后的ESMFold模型。