GTsR

GTsR is a graph neural network based tool for solvent identification, solvent removal, and activation-stability prediction in metal-organic frameworks (MOFs).

Pretrained Models

Checkpoint	File	Purpose
free	ckpt/free_best.pth	Remove free solvent
all	ckpt/all_best.pth	Remove all solvent
stability	ckpt/stability_best.pkl	Predict activation stability

Installation

git clone https://github.com/coollkr/GTsR.git
cd GTsR
conda env create -f environment.yml
conda activate gtsr
pip install -e .

or

conda install -c conda-forge zeopp-lsmo
pip install gtsr

Usage

• Remove solvent

from gtsr import GTsRunner

runner = GTsRunner(checkpoint="free")
result = runner.clean(
    cif="input.cif",
    output="prediction",
    threshold=0.5,
)

You can also use:

runner = GTsRunner(checkpoint="all")
runner = GTsRunner(checkpoint="path/to/ckpt.pth", device="cpu")

runner.clean() returns a dictionary with the following fields:

Field	Description
input	Absolute path to the input CIF
output	Output directory
framework	Path to the cleaned framework CIF
solvent	Path to the solvent CIF, or None if not generated
checkpoint	Path to the checkpoint used for prediction
task	Task name stored in the checkpoint
threshold	Atom classification threshold
num_atoms	Total number of atoms
num_framework_atoms	Number of framework atoms
num_solvent_atoms	Number of solvent atoms
probabilities	Solvent probability for each atom
labels	Predicted class label for each atom
solvent_smiles	SMILES strings of identified solvents

• Predict activation stability

from gtsr import GTsRunner

runner = GTsRunner(checkpoint="stability")
score = runner.stability(cif="cleaned_framework.cif")

if score == 1:
    print("The cleaned structure is stable.")
else:
    print("The cleaned structure is not stable.")

Web Interface

Streamlit demo

Citation

Update the following entry when the associated publication becomes available:

@article{
doi:10.26434/chemrxiv.15006593/v1,
author = {Guobin Zhao and Kairui Liang and Xiao-Yan Li},
title = {Stability-Aware Structural Cleaning of Metal–Organic Frameworks for Reliable Molecular Simulation and High-Throughput Screening},
journal = {ChemRxiv},
volume = {2026},
number = {0727},
year = {2026},
doi = {10.26434/chemrxiv.15006593/v1},
URL = {https://chemrxiv.org/doi/abs/10.26434/chemrxiv.15006593/v1},
eprint = {https://chemrxiv.org/doi/pdf/10.26434/chemrxiv.15006593/v1},
}

License

This project is released under the MIT License. See LICENSE for details.