Supplementary data for "Correlation-Free Transition Path Sampling through Shooting Point Generation Guided by Committor Learning"
This repository contains the simulation and analysis code and notebooks for
Negedly, M., Falkner, S., Coretti, A., & Dellago, C. (2026). Correlation-Free Transition Path Sampling through Shooting Point Generation Guided by Committor Learning. arXiv:2609.20461. DOI: 10.48550/arXiv.2609.20461
In the paper, we introduce the GenAIMMD method and test it on two systems: the two-dimensional Wolfe-Quapp (WQ) potential and a polymer model with 11 degrees of freedom.
The code and notebooks for the two test systems are located in this repo's accordingly named branches --- wolfe_quapp and polymer. Since the corresponding simulation data are over 300 GB in size, they are available separately on Zenodo (DOI: 10.5281/zenodo.22875161). In each branch, there are multiple directories which only contain a README.md each. This is where the extracted data from Zenodo must be placed in for the scripts to find them. The respective README.md files provide more details. This README will update for each branch, describing the files contained therein.
From a working Python installation (tested on 3.14), install the required packages by changing to the repo's root and running
$ pip install -r requirements.txt