Dataset for article "AFM-Fold: Rapid Reconstruction of Protein Conformations from AFM Images"

This repository contains data used in a paper "AFM-Fold: Rapid Reconstruction of Protein Conformations from AFM Images" by Kawai et al. 

• [flhac_conformations/]: Structural ensembles of the C-terminal domain of FlhA (FlhA_C) generated by different models. 
  • [alphaflow/]
    • [aflow_md_3a5i_1.dcd, aflow_md_3a5i_1.pdb]: Structural ensemble of FlhA_C generated by AlphaFlow MD. 
    • [aflow_pdb_3a5i_1.dcd, aflow_pdb_3a5i_1.pdb]: Structural ensemble of FlhA_C generated by AlphaFlow PDB. 
  • [bioemu/]
    • [bioemu_3a5i_1.dcd, bioemu_3a5i_1.pdb]: Structural ensemble of FlhA_C generated by BioEmu. 
  • [msa_subsampling/]
    • [msa_subsampling_3a5i.dcd, msa_subsampling_3a5i.pdb]: Structural ensemble of FlhA_C generated by MSA subsampling ( using ColabFold ). 
• [guidance_scheduling/]: Results of structure generation under different guidance strength parameters (t_start and y_max). 
  • [ak/*-*-*/results_*_*.npz]: Structure generation results of Adenylate Kinase (AK) with varied guidance strength hyperparameters. The three numbers in each directory name correspond to the inter-domain distances (ATPbd–Core, Core–AMPbd, AMPbd–ATPbd, in Å). 
  • [flhac/*-*/results_*_*.npz]: Structure generation results of FlhA_C with varied guidance strength hyperparameters. The two numbers in each directory name correspond to the inter-domain distances (A_CD_2–A_CD_3, A_CD_4–A_CD_1, in Å). 
• [noise_robustness/]
  • [*-*.npz]: Results obtained by varying the noise standard deviation of the training AFM images and the “experimental” images used for structure estimation. 
• [real_afm/]
  • [frame_*.npz]: Estimated 3D structures of FlhA_C from time-series real AFM images. 
• [real_afm_histogram.npy]: Histogram of the real AFM image intensity distribution. 
• [1ake.pdb]: Chain A of PDB ID: 1AKE. 
• [4ake.pdb]: Chain A of PDB ID: 4AKE. 
• [3a5i.pdb]: Chain A of PDB ID: 3A5I. 
• [ak_train.dcd, ak_train.pdb]: Candidate conformations of Adenylate Kinase used for model training. 
• [ak_train.npz]: Evaluation results for the candidate conformations of Adenylate Kinase. 
• [flhac_train.dcd, flhac_train.pdb]: Candidate conformations of FlhA_C used for model training. 
• [flhac_train.npz]: Evaluation results for the candidate conformations of FlhA_C. 
• [ak.pt]: CNN parameters for Adenylate Kinase structural estimation. 
• [flhac.pt]: CNN parameters for FlhA_C structural estimation. 
• [ak_summary.npz]: Summary of structural estimation results for Adenylate Kinase, trained on noise-free pseudo-AFM images. 
• [ak_summary_noisy.npz]: Summary of structural estimation results for Adenylate Kinase, trained on pseudo-AFM images with noise std. dev. = 0.3 nm. 
• [ak.dcd, ak.pdb]: Conformations randomly sampled from a 450 ns MD trajectory for validation.