State Predictive Information Bottleneck (SPIB) is a deep learning-based framework that learns the reaction coordinates from high dimensional molecular simulation trajectories. Please read and cite this manuscript when using SPIB: https://aip.scitation.org/doi/abs/10.1063/5.0038198. Application of SPIB to two practical, and biophysically relevant systems was demonstrated in this recently published work: https://pubs.acs.org/doi/abs/10.1021/acs.jctc.2c00058.
Objective: This two part tutorial series is designed to guide a user in employing SPIB to analyze molecular dynamics simulations trajectories (4 state potential, and alanine dipeptide in vacuum). Users will also learn to perform metadynamics by biasing SPIB learned reaction coordinates. We recommend using Google Colab (GPU accelerator) while following these tutorials.
Prerequisites: Users should have a working knowledge of Python. Since SPIB is implemented in PyTorch, some familiarity with it can be useful but not essential. The notebooks contain installation commands for all the necessary softwares, and libraries including Gromacs, and PLUMED.