/GCNCC

Graph Convolutional Neural Networks for clustering and disease classification

Primary LanguageShellMIT LicenseMIT

Graph Convolutional Network for Clustering and Disease Classification

Official implementation of Graph Convolutional Network for Clustering and Classification (GCNCC).

Omar Maddouri, Xiaoning Qian, and Byung-Jun Yoon, [Deep graph representations embed network information for robust disease marker identification]

NOTE: The bash/ folder is intended to reproduce the validation and evaluation experiments from the paper.

Installation

python setup.py install

Dependencies

pip install -r requirements.txt

GCNCC workflow

alt text

Usage

Note: The validation and evaluation pipelines under the bash/ folder contain files ordered by prefix numbers in the file names for sequential execution.

  1. Download the GitHub repository locally.
  2. Create a new folder data/ with all required sub-hierarchies as indicated in the next steps.
  3. Download the PPI network "9606.protein.links.v11.0.txt" or a newer version for homo sapiens from STRING (https://string-db.org/cgi/download.pl) and place it under folder data/reference/ppi_network/
  4. Download the dataset of interest under data/raw_input/
  5. For hyperparameter tuning and validation experiments, consider the pipeline under bash/validation/
  6. For evaluation experiments, consider the pipeline under bash/process/

Note: The output results are saved under data/output/

Cite

To be added!