MDGraphEmb

MDGraphEmb converts protein MD simulation trajectories into graph-based representations and corresponding graph embeddings, facilitating machine learning analysis of protein dynamics.

Key Features:

• Graph-based representation: Transforms protein MD simulation trajectories into graph-based representations.
• Graph embedding generation: Produces graph embeddings that compress high-dimensional, noisy trajectories into tabular formats suitable for machine learning.
• Framework for graph embedding techniques: Supports a range of graph embedding techniques and machine learning models for workflow creation.
• Performance comparison: Provides comparison of different graph embedding methods combined with machine learning models.

Scientific Applications:

• Protein conformation analysis: Analyses protein dynamics and identifies important protein conformations from molecular dynamics simulations.
• Functional state classification: Classifies functional states by encoding and detecting important protein conformations, demonstrated with adenylate kinase (ADK) as the main case study.
• Generalization validation: Validated on additional systems including Plantaricin E (PlnE) and HIV-1 protease.
• Downstream machine learning classification: Applies graph embedding to diverse downstream machine-learning classification tasks using structural information from protein MD simulation data.

Methodology:

MDGraphEmb is built on MDAnalysis. Protein MD simulation trajectories are converted into graph-based representations, from which graph embeddings are generated to create low-dimensional representations that preserve graph topology and node properties.