msk-moi/dynosarc

A toolbox for performing Wasserstein-based subtyping and computing dynamic Ollivier-Ricci curvature demonstrated with respect to sarcoma

Article

The preprint of the manuscript "Dynamic Network Curvature Analysis of Gene Expression Reveals Novel Potential Therapeutic Targets in Sarcoma" can be found at https://doi.org/10.1101/2022.03.09.483487

Cite

Please cite our paper if you use this code in your work

@article{elkin2023geometry,
  title={Geometry of gene expression network reveals potential novel indicator in Ewing sarcoma},
  author={Elkin, Rena and Oh, Jung Hun and Dela Cruz, Filemon and Norton, Larry and Deasy, Joseph O and Kung, Andrew L and Tannenbaum, Allen R},
  journal={Cancer Research},
  volume={83},
  number={7\_Supplement},
  pages={6541--6541},
  year={2023},
  publisher={AACR}
}

Introduction

DYNOsaRC is a general purpose package for performing Wasserstein-based hierarchical clustering and computing dynamic Ollivier-Ricci curvature on weighted graphs.

Examples running the code to reproduce the analysis performed in the manuscript can be found in the notebooks folder. Briefly, DYNOsaRC analysis was applied to analyze weighted transcriptomic networks associated with pediatric sarcomas. The two-fold analysis entailed:

1. Performing Wasserstein-based subtyping

2. Analyzing persistent functional gene associations via dynamic Ollivier-Ricci curvature

Details of the approach can be found in the manuscript.

Dependencies

• NetworkX
• NumPy
• pandas
• Pot
• SciPy

Acknowledgements

The code for computing dynamic Ollivier-Ricci curvature was largely based off of the code written by Gosztolai and Arnaudon

@article{GosztolaiArnaudon2021,
author = {Gosztolai, Adam and Arnaudon, Alexis},
doi = {10.1038/s41467-021-24884-1},
issn = {2041-1723},
journal = {Nat. Commun.},
number = {1},
pages = {4561},
title = {{Unfolding the multiscale structure of networks with dynamical Ollivier-Ricci curvature}},
volume = {12},
year = {2021}
}