PyGVAMP Documentation¶
PyGVAMP is a refactored implementation of GraphVAMPNets built on the PyTorch Geometric architecture, achieving up to 50x speedup compared to the original implementation.
What is PyGVAMP?¶
PyGVAMP trains variational autoencoder models to identify and classify conformational states in protein molecular dynamics trajectories. It processes MD data by:
- Constructing graphs from atomic coordinates
- Learning state representations through neural message-passing
- Predicting state classifications and transition kinetics
Quick Links¶
- Installation & Getting Started
- Pipeline Execution
- Codebase Summary
- Testing Guide
- GitHub Repository
- Interactive Analysis Results
Generated Reports¶
PyGVAMP produces interactive analysis reports that include:
- 3D Embeddings — Visualize learned state representations in 3D space
- Protein Structure Viewer — Inspect conformational ensembles with 3Dmol.js
- Transition Matrices — Explore state-to-state transition kinetics
- Attention Patterns — Understand residue-level interactions driving state classification
Browse published results at pygvamp.com/analyses.