MOSCITO - Molecular Dynamics Subspace Clustering with Temporal Observance

Code for master thesis

Requirements

For running the code the following libraries are required:

• Pyemma 2.5.12
• pypardiso 0.4.2

Usage

1. Get the desired features

Depending on the format of the trajectory file a topology file has to be provided! Available features:

• Coordinates of all atoms
• Coordinates of C-alpha atoms
• Backbone torsions
• Minimal distance between residues
• Solvent accessible surface area (SASA)
• Dihedral angles between the chi1 - chi5 dihedral

from feature_selection import FeatureSelector

trajectory = 'path/to/trajectory_file'
topology = 'path/to/topology_file'
feature_selector = FeatureSelector(trajectory, topology)
feature = feature_selector.get_backbone_torsions()

2. Run MOSCITO

from moscito import MOSCITO

moscito = MOSCITO()
affinity = moscito.fit_predict(feature)

3. Run spectral clustering

from sklearn.cluster import SpectralClustering

num_clusters = 10
sc = SpectralClustering(num_clusters, affinity='precomputed')
labels = sc.fit_predict(affinity)

4. Visualize clustering

from visualization.visualize_clustering import show_clustering

show_clustering(labels)