GGNN

A Gemetric informed Graph neural network based model for Cancer survival prediction

Requirements

• pytorch
• networkx
• numpy
• pandas
• POT
• lifelines
• matplotlib

Use GGNN

1. Download multi-omics data and biological network from

• https://themmrf.org/finding-a-cure/our-work/the-mmrf-commpass-study/
• https://www.cbioportal.org/
• https://www.hprd.org/
• https://www.genome.jp/kegg/

2. Run preprocessing: Preprocessing.R

• Input multi-omics data is filtered by available genes intersect with HPRD gene set and KEGG gene set
• Output RNA.csv, CNA.csv, Methy.csv and clinn.csv in a folder "out/" in the original data folder

3. Compute Ollivier-Ricci curvature: Compute_ORC.py

• Use csv files generated from last step
• Curvature results are saved in "out/" as RNA_curv.csv, CNA_curv.csv and Methyl_curv.csv

4. Run GGNN model: Test_multi_curv_net_surv.py

• A processed data set of TCGA study of LGG with HPRD network and KEGG pathway can be download via the link bellow:
• https://stonybrookmedicine.box.com/s/7t8qbrmni7n16lcfu3kmlnwje9l6qgdy