All published workflows

Published workflow · Temple Compute

BioExcel Autoencoder - MD Trajectory Analysis with Machine Learning

W-25 · AutoEncoders for MD Analysis

v116 stages3 clones0

The pipeline

Every stage picks its own compute target. Temple Compute OS resolves the dependencies between them and moves the data across each boundary.

  1. 01Download apo training trajectory (6W9C_apo) from MDDB
    Consumes
    fetch_train_trajectory_config, src_conda_env_yaml
    Produces
    train_structure, train_trajectory
  2. 02Fit and align apo training trajectory (rot+trans)
    Consumes
    fit_train_trajectory_config, train_trajectory, train_structure, src_conda_env_yaml
    Produces
    train_trajectory_fit
  3. 03Featurize apo training trajectory (CA Cartesian coordinates)
    Consumes
    featurize_train_trajectory_config, train_trajectory_fit, train_structure, src_conda_env_yaml
    Produces
    train_dataset, train_stats
  4. 04Build autoencoder model architecture
    Consumes
    build_model_config, train_stats, src_conda_env_yaml
    Produces
    initial_model
  5. 05Train autoencoder model on apo trajectory dataset
    Consumes
    train_model_config, initial_model, train_dataset, src_conda_env_yaml
    Produces
    trained_model, training_metrics
  6. 06Download holo test trajectory (6W9C_holo) from MDDB
    Consumes
    fetch_test_trajectory_config, src_conda_env_yaml
    Produces
    test_structure, test_trajectory
  7. 07Fit and align holo test trajectory to apo reference (rot+trans)
    Consumes
    fit_test_trajectory_config, test_trajectory, train_structure, src_conda_env_yaml
    Produces
    test_trajectory_fit
  8. 08Featurize holo test trajectory (CA Cartesian coordinates)
    Consumes
    featurize_test_trajectory_config, test_trajectory_fit, test_structure, src_conda_env_yaml
    Produces
    test_dataset, test_stats
  9. 09Evaluate trained autoencoder on apo dataset
    Consumes
    evaluate_model_config, trained_model, train_dataset, src_conda_env_yaml
    Produces
    eval_results
  10. 10Create GROMACS index file for apo CA atoms
    Consumes
    make_ndx_apo_config, train_structure, src_conda_env_yaml
    Produces
    train_index
  11. 11Create GROMACS index file for holo CA atoms
    Consumes
    make_ndx_holo_config, test_structure, src_conda_env_yaml
    Produces
    test_index
  12. 12Compute RMSF of apo trajectory
    Consumes
    gmx_rmsf_apo_config, train_structure, train_trajectory, train_index, src_conda_env_yaml
    Produces
    rmsf_apo
  13. 13Compute RMSF of holo trajectory
    Consumes
    gmx_rmsf_holo_config, test_structure, test_trajectory, test_index, src_conda_env_yaml
    Produces
    rmsf_holo
  14. 14Reconstruct holo trajectory from autoencoder latent space
    Consumes
    eval_results, test_stats, src_conda_env_yaml
    Produces
    recon_trajectory, recon_structure
  15. 15Compute RMSF of reconstructed holo trajectory
    Consumes
    gmx_rmsf_holo_recon_config, recon_structure, recon_trajectory, test_index, src_conda_env_yaml
    Produces
    rmsf_holo_recon
  16. 16Generate PLUMED input files for enhanced sampling with autoencoder CVs
    Consumes
    make_plumed_config, trained_model, train_index, train_structure, train_stats, src_conda_env_yaml
    Produces
    plumed_model, plumed_dat, plumed_features_dat

Run this workflow

This release is frozen and self-contained: the pipeline definition, its input files, and its plugin environment. Open it in Temple Compute OS to clone it into your own account and run it on HPC or any cloud. Temple Compute OS is in private beta, so you will need an invitation first.

Run this in Temple Compute OSRequest an invitation