PathCNN: Interpretable convolutional neural networks for survival prediction and pathway analysis applied to glioblastoma

Jung Hun Oh ^1,†,∗, Wookjin Choi ^2,†, Euiseong Ko ³, Mingon Kang ^3,∗, Allen Tannenbaum ⁴ and Joseph O. Deasy ¹

¹Department of Medical Physics, Memorial Sloan Kettering Cancer Center, New York, USA, ²Department of Computer Science, Virginia State University, Petersburg, USA, ³Department of Computer Science, University of Nevada, Las Vegas, USA and ⁴Departments of Computer Science and Applied Mathematics & Statistics, Stony Brook University, New York, USA

^*To whom correspondence should be addressed.
^†The authors wish it to be known that, in their opinion, the first two authors should be regarded as Joint First Authors.

Contact: [email protected] or [email protected]

1. Model Building

   • PathCNN.py

2. GradCAM

   • PathCNN_GradCAM_modeling.py: to generate a model for GradCAM (PathCNN_model.h5)
   • PathCNN_GradCAM.py: to generate GradCAM images and a resultant file (pathcnn_gradcam.csv)

3. Multi-omics data

   • GBM multi-omics data including mRNA expression, CNV, and DNA methylation were downloaded from the CBioPortal database.
   • Pathway information was downloaded from the KEGG database.
   • PCA was performed for each pathway in individual omics types.

   Five PCs in each omics type are in the following files:

   • PCA_EXP.xlsx, PCA_CNV.xlsx, PCA_MT.xlsx

   Clinival variables are in the following file:

   • Clinical.xlsx