Graph Neural Network-State Predictive Information Bottleneck (GNN-SPIB) approach for learning molecular thermodynamics and kinetics
This work addresses the problem of timescale limitations in molecular dynamics simulations for computational chemists and biophysicists by providing an automated approach that eliminates the need for pre-defined reaction coordinates, representing an incremental improvement over existing enhanced sampling methods.
The researchers tackled the challenge of molecular dynamics simulations being limited by timescale constraints and reliance on expert-selected features by developing the GNN-SPIB framework, which automatically learns low-dimensional representations from atomic coordinates using graph neural networks and the State Predictive Information Bottleneck, successfully predicting structural, thermodynamic, and kinetic information for slow processes across three benchmark systems.
Molecular dynamics simulations offer detailed insights into atomic motions but face timescale limitations. Enhanced sampling methods have addressed these challenges but even with machine learning, they often rely on pre-selected expert-based features. In this work, we present the Graph Neural Network-State Predictive Information Bottleneck (GNN-SPIB) framework, which combines graph neural networks and the State Predictive Information Bottleneck to automatically learn low-dimensional representations directly from atomic coordinates. Tested on three benchmark systems, our approach predicts essential structural, thermodynamic and kinetic information for slow processes, demonstrating robustness across diverse systems. The method shows promise for complex systems, enabling effective enhanced sampling without requiring pre-defined reaction coordinates or input features.