Robust-Equilibration-Detection-Paper

To reproduce the paper:

git clone https://github.com/michellab/Robust-Equilibration-Detection-Paper.git
cd Robust-Equilibration-Detection-Paper
make all

Inputs and code to reproduce the results and analysis from "Robust Equilibration Detection for Molecular Simulations". Analysis is performed using the RED Python package and all data is available on Zenodo.

├── analysis
│   ├── analysis_free.ipynb
│   ├── analysis.ipynb
│   └── analysis_single.ipynb
├── compute_equil_times
│   ├── compute_equil_times.py
│   └── compute_equil_times.sh
└── synthetic_data_creation
    ├── dataset_creation.ipynb
    ├── dataset_creation_free.ipynb
    └── dataset_creation_single.ipynb

This repository is split into three sections:

• Synthetic data creation (see synthetic_data_creation): Notebooks required to create the synthetic datasets from the real ABFE data
• Computation of equilibration times (see compute_equil_times): Scripts to calculate the equilibration times for the synthetic dataset using RED
• Analysis (see analysis): Notebooks to analyse the performance of the equilibration detection methods

For all notebooks, _free denotes the free vanish multi-window data and _single denotes the bound vanish single-window data - otherwise the notebooks deal with the standard bound vanish multi-window data.

To reproduce the paper, you will require make installed on your system. If you would like to play around with the data and analysis using the same synthetic data as in the paper, run:

make env
make download_data # Downloads the large files from Zenodo (https://zenodo.org/records/13902735)
make figures_only

This will pull all data from Zenodo (https://zenodo.org/records/13902735) and recreate all figures, without running computationally expensive steps. This will take a few minutes.

If you would like to reproduce the entire paper, including generation of the synthetic data and computation of equilibration times, run:

make all

Note that this is computationally expensive (due to the repeated calculation of equilibration times on thousands of synthetic datasets) and will take a few hours to complete.

You can clean any generated files with:

make clean