Deep Signature: Characterization of Large-Scale Molecular Dynamics

---

This repository provides the official implementations and experiments for our paper "Deep Signature: Characterization of Large-Scale Molecular Dynamics". The core code for implementing our method can be found in models/deep_signature.py. Please follow the steps to prepare running environment, and run our code.

Prepare Environment

conda create -n torch python=3.9

conda activate torch

conda install scipy Pillow Pyyaml tqdm pandas scikit-image scikit-learn matplotlib seaborn networkx numba h5py

conda install -c conda-forge mdtraj mdanalysisdata mdanalysis

pip install torch==1.9.0+cu111 torchvision==0.10.0+cu111 torchaudio==0.9.0 -f https://download.pytorch.org/whl/torch_stable.html

pip install torch_scatter==2.0.9 torch_sparse==0.6.12 torch_cluster==1.5.9 torch_spline_conv==1.2.1 -f https://pytorch-geometric.com/whl/torch-1.9.0+cu111.html

pip install signatory==1.2.6.1.9.0 --no-cache-dir --force-reinstall
pip install moviepy
pip install torchdiffeq
pip install rdkit
pip install wandb

🚀 Quick Start

1. Download Datasets

You can download our processed data for GPCR by the link, after download and unzip the file, just specify the file path in the script.

2. Running

# 1. Model training for the GPCR dataset
chmod +x ./scripts/*
./scripts/gpcr_graph_sig.sh

You can find the running commands for other baselines in the scripts file.

Citations

If you find this repo useful for your research, please cite the following papers:

@inproceedings{Qin2025ICLR,
    title={Deep Signature: Characterization of Large-Scale Molecular Dynamics},
    author={Tiexin Qin, Mengxu Zhu, Chunyang Li, Terry Lyons, Hong Yan and Haoliang Li},
    booktitle={The Thirteenth International Conference on Learning Representations},
    year={2025},
    url={https://openreview.net/forum?id=xayT1nn8Mg}
}