Graph Network Code for Analyzing Molecular Dynamics Trajectories

Code for calculating inter-species cluster network diameters for various MD simulations.

Contents

• src/ core code
  • network_analysis.py functions for creating graphs and calculating graph metrics
    • make_graph: create NetworkX Graph
    • compute_metric: calculate NetworkX Graph diameter
    • get_network: apply cell-linked list algorithm to generate edge list
  • visualization.py plotting diameters of graph network
  • util.py functions for computing distances, angles, and creating cell maps
• examples/
  • water/ See Int. J. Mol. Sci. 2020, 21(2), 403
  • N1888/ See J. Phys. Chem. B 2024.

Example: Water cluster calculations

Reproducing the water cluster analysis results requires 3 steps. Following the `examples/water' directory:

1. Installation

git clone https://github.com/shehan807/graph-network-analysis.git

2. Set up confing.yaml

The config.yaml contains all tunable parameters (number of cells, cutoff distances, edge criteria definitions, etc) to obtain a cluster of like-species from an MD trajectory. The example below specifies a water molecule residue, HOH along with a distance criteria of 0.36 nanometers to generate two oxygen atoms, O-O, as nodes connected by an edge.

num_cores: 1 # Number of cores/processes for joblib.Parallel 
num_cells: 10 # Number of cells to divide along each dimension of MD box 
cutoff: 0.4 # cutoff distance, nm
residue_name: "HOH" # species for inter-species cluster analysis
criteria: # list of criteria for network edge
  - name: "O-O" # atom symbols 
    distance: 0.36 # nm 
    min_true: 1
    angle: None # deg
      #  - name: "H1-O"
      #    distance: 0.245
      #    angle: 30.0
      #  - name: "H2-O"
      #    distance: 0.245
      #    angle: 30.0
check_pkl: True # check if edges.pkl exists in code

For automating the process of creating numerous config.yaml files, see examples/N1888/create_config.sh.

3. Run Shell Script

Create an inputs directory to place a <topology>.pdb and <trajectory>.dcd file and run

./run_analysis

The /outputs directory will contain a pickle file for the edges (to avoid having to rerun NetworkX edge generation in the future) along with the generated probability ditribution of diameters, $P(d)$, that should look something like:

Copyright

Copyright (c) 2024, McDaniel Group at Georgia Institute of Technology and and Authors: Authors:

• Shehan Parmar ([email protected])
• John P. Stoppleman ([email protected])