Robust and fast containment charaterization for (periodic) point clouds and voxel masks.
Molecular Dynamics Voxel Containment yields a robust characterization of the inside and outside hierarchy for point clouds in periodic spaces of dimensionality three or lower (e.g. R^3/Z^3).
Using an MDAnalysis atomgroup, bead selection and resolution, a density grid is created. This density grid is segmented using connected components, and graph logic is utilized to solve the topological identification of containment (insides and outsides). The final output is a set of Directed Acyclic Graphs (DAGs) running from the largest container to the smallest (from outside inwards in graph space). This containment logic can then be used to analyse or manipulate the systems.
Any complex configuration of (non)periodic segments is supported by this algorithm in a fast, robust, unambiguous, deterministic and rot+trans invarient (up to voxel discretization) manner.
Note
MDVContainment has had a rework offering much improved performance and quality. However, here is the legacy version if you insist on using it.
Figure 1 | Containment hierarchy in self-assembled acyl chain bicelles. The main solvent (segment -2) is the most outer segment in this system. It containes three non-periodic segments (1, 2, 3), where segment 1 is split over the periodic boundary. Segment 3 contains a bubble of inner solvent (segment -1).
Figure 2 | A periodic hollow cylinder in solution. The cylinder (segment 1) splits the solution into two segments (segment -2, -1), the solid cylinder inside the hollow cylinder (segment -1), and all of the space outside of the cylinder (segment -2). Both cylindrical segments (1 and -1) are said to be contained by the solvent segment (-2), although only the hollow cylinder (1) is a child of the solvent segment (-2).
This work is currently in the process of publication, the DOI of the mansucript will be placed here once it is available.
Please cite this work if you use it for scientific publications. It helps me to continue to work on this kind of software, thanks! On that note, if you would offer me a job to work on this, I would take your offer very seriously.
MDVContainment is available under the Apache-2.0 license.
MDVContainment has been tested to work with python >= 3.8 and Ubuntu 20.04.6 LTS.
Direct install into the current python environment libraries:
pip install git+https://github.com/BartBruininks/mdvcontainment
Create a folder in a custom location to have access to the examples folder:
git clone git@github.com:BartBruininks/mdvcontainment.git
cd mdvcontainment
pip install .
Important
If you need any help with MDVContainment or have ideas for future functionalities, please raise an issue!
Input
# `minimal_example.py` for a CG Martini structure file
# Import the required libraries
import MDAnalysis as mda
from mdvcontainment import Containment
# Import the structure file
path = 'your_structure.pdb' # Or any MDA supported structures file
universe = mda.Universe(path)
selection_string = 'not resname W WF ION' # Useful for CG Martini
selection = universe.select_atoms(selection_string)
# Run the containment analysis
containment = Containment(selection, resolution=0.5, closure=True)
# Show the containment graph with voxel counts
print(containment)
Note
For atomistic structures use closure=False
. Take a look at
closing (link to wikipedia) to learn more about what it does.
Output
Containment Graph with 3 components (component: nvoxels):
└── [-2: 54461]
└── [1: 15403]
└── [-1: 5136]
For worked examples in jupyter notebooks, take a look at the examples/notebooks folder. Some example structure files are added under examples/structures.