Welcome to EnGens documentation!

EnGens is a Python library for anyone interested in studying protein structures and their flexibility. EnGens provides a computational framework for generation and analysis of representative protein conformational ensembles.

With EnGens you will be able to process your datasets of protein structure both:

• static (gathered from the PDB databank or modelled)

• dynamic (generated by MD simulation)

EnGens will analyze your dataset and extract a representative ensemble with the following steps:

1. Data loading and preprocessing

2. Data featurization

3. Dimensionality reduction

4. Clustering and ensemble extraction

5. Visualization and further analysis

You can use the resulting ensemble for downstram tasks such as docking, structural drug design or other analysis!

Contents

Check out the Installation instructions section for further information regarding the Installation of the project. Check out the Workflow Description section for detailed description of the workflows. Finally, check out EnGens Usage for the instructions on how to use EnGens.

• Installation instructions
  • Installation
    • Docker pull
    • Docker build
    • Conda build
• Workflow Description
  • Flow Chart
  • Workflows detailed description
  • Workflow 1S: Extracting featurized representations from raw data (static use-case)
    • Workflow 1S-1: input description
    • Workflow 1S-2: preprocessing input
    • Workflow 1S-3: featurization
    • Workflow 1S-4: output description
  • Workflow 1D: Extracting featurized representations from raw data (dynamic use-case)
    • Workflow 1D-1: input description
    • Workflow 1D-2: preprocessing input
    • Workflow 1D-3: featurization
    • Workflow 1D-4: VAMP featurization tuning
    • Workflow 1D-5: output description
  • Workflow 2(S&D): Projecting the featurized representation into an embedding in low-dimensional space (static and dynamic use-case)
    • Workflow 2-1: input description
    • Workflow 2-2a: PCA
    • Workflow 2-2b: UMAP
    • Workflow 2D-2c: TICA
    • Workflow 2D-2d: SRV
    • Workflow 2-3: output description
  • Workflow 3(S&D): Clustering embeddings and extracting the ensemble (static and dynamic use-case)
    • Workflow 3-1: input description
    • Workflow 3-2a: K-means clustering
    • Workflow 3-2b: GMM clustering
    • Workflow 3-2c: hierarchical clustering
    • Workflow 3-3: representative selection
    • Workflow 3-3: output description
  • Workflow 4(S&D): Visualizing the data and analyzing the ensemble
    • Embedding plot
    • Cluster membership plot
    • Multiple structures visualization
    • Ensemble summary box plots
    • Ensemble summary scatter plots
• EnGens Usage
  • Usage
    • Static use-case
      • Run static workflow of EnGens
      • Extract features from the PDB files
        • Step 1 - load the structure and trajectory
        • Step 2 - select different featurizations
        • Step 3 - evaluate the featurizations
        • Not an option for crystal structure input!!
        • Step 4 - pick the featurization
        • Step 5 - save the results as input for Workflow2 - dimensionality reduction
    • Dynamic use-case
      • Workflow 1 - extract features from the trajectory
        • Step 1 - load the structure and trajectory
        • Step 2 - select different featurizations
        • Step 3 - evaluate the featurizations
        • Step 4 - pick the featurization
        • Step 5 - save the results as input for Workflow2 - dimensionality reduction
      • Workflow 2 - reduce the dimensionality of the featurized trajectory
        • Step 0 - Load the results from Workflow1
        • Step 1 - Choose the method for dimensionality reduction and apply it
        • Step 2 - inspect the results visually
        • Step 3 - save results for Workflow3 and clustering
      • Workflow 3 - cluster the trajectory
        • Step 0 - Load data from Workflow2
        • Step 1 - choose the clustering method
        • Step 2 - run the clustering with different parameters
        • Step 3 - optionally pick a subset of clusters with the heighest weight
        • Step 4 - extract conformations for the ensemble
        • Step 5 - save results for analysis
      • Workflow 4 - analyze the results
        • Cluster plots
        • View multiple structures per cluster
        • Trajectory movie
        • Additional info per cluster - distance between two residues
        • Additional info per cluster - RMSD to a frame
        • Additional info per cluster - RMSD to another structure (given a PDB file)
        • Additional info per cluster - component from dimensionality reduction

Acknowledgments

This work was built up on a large amount of community effort including the following tools:

• Structural bioinformatics (and MD) software

  • MDTraj

  • MDAnalysis

  • PyTraj

  • MSMTools

  • PyEmma

  • OpenMM

  • PDBFixer

  • SRV

  • mTM-align

  • BioPython

  • PDBRenum

• Visualization

  • NGLViewer

  • py3Dmol

  • ProtVista

• General ML tools

  • scikit-learn

  • scipy

  • umap-learn

  • deeptime

• Others

  • plotly

  • pandas

  • numpy

We would like to thank the authors and the maintainers of the tools!