PathExpSurv: Pathway Expansion for Explainable Survival Analysis and Cancer Driver Gene Discovery

Requirements

• torch
• numpy
• pandas

Data Preparation

We obtain 3 different survival datasets from UCSC Xena: (1) Thyroid Cancer (THCA) Dataset, (2) Lower Grade Glioma (LGG) Dataset, (3) Breast Cancer (BRCA) Dataset. We took the prior pathways as the functional modules. The source of the prior signaling pathways is：KEGG DISEASE Database. We put them into the Dataset/ folder.

Model Training

# Two-Phase Training
python main.py --task='LGG'

Downstream Analysis

We performed several downstream analysis including pathway expansion, Kolmogorov-Smirnov test, two-phase result analysis, single gene survival analysis.

python downstream_analysis.py

Results

Performance of Survival Analysis

Disease Drivers Discovery

Contact

Please open an issue or contact zchou0807@gmail.com with any questions.