PathFold: Predicting the Entire Protein Folding Pathway from Protein Sequence Alone

We provide PathFold-predicted trajectories for 175 test entries in PathFold.zip. The dataset is organized into three subsets corresponding to different model variants: multi_step1_batch10 – Trajectories generated by PathFold-1 for 175 test entries. multi_step3_batch10 – Trajectories generated by PathFold-3 for 175 test entries. multi_step6_batch10 – Trajectories generated by PathFold-6 for 175 test entries. Each subset contains 175 subdirectories, one for each test entry. Each subdirectory includes the predicted structures in PDB format, named: [PDBID]_[time_step]_epoch_6.pdb The time_step starts from 0, which represents the extended (unfolded) conformation, and increases sequentially along the predicted folding trajectory. In addition, we provide PathFold_SMD_test.zip. This archive contains five independent SMD trajectories for each of the 175 test entries, provided as PDB files. The frame number is indicated in each filename, with a time interval of 1 ps between consecutive frames. The number of frames varies depending on the protein.