PathNetDRP: A Novel Biomarker Discovery Framework Using Pathway and Protein-Protein Interaction Networks for Immune Checkpoint Inhibitor Response Prediction
This source code is the implementation code of the PathNetDRP algorithm to accurately predict ICI response.
python 3.8.18
pandas 2.0.3
numpy 1.23.5
scikit-learn 1.3.0
statsmodels 0.14.0
networkx 3.1
scipy 1.10.1
Create PathNetGene score matrix files for each cohort.
#first,
python 1_get_score_by_pathway.py
# If the 1_get_score_by_pathway.py process has been completed,
python 2_calculate_score.py
If the 2_calculate_score.py process is completed, the PathNetGene score matrix will be created in the result folder.
Select Biomarker related to immunotherapy responsiveness for each cohort.
python 3_select_feature.pyCheck LOOCV prediction performance for selected biomarkers.
python 4_LOOCV.py- This source code calculates the PathNetGene Score based on pre-treatment RNA-seq data from patients treated with ICI.
- You may use raw count data, but please ensure it is normalized (e.g., FPKM, TPM, TMM, etc.).
- The gene expression data should be a matrix with genes as rows and sample IDs as columns.
- Our experiments were conducted on cohorts with melanoma, metastatic gastric cancer, and bladder cancer. Other cancer types can be used but may not exhibit superior predictive performance.
- Please refer to the paper for the data required for the experiments. Due to storage limitations, only one cohort is provided as a sample.
- The PPI network data was downloaded from https://string-db.org/ (human version 11.0). For detailed preprocessing steps, please refer to the paper.
- For pathway data, we downloaded it from msigDB (https://www.gsea-msigdb.org/gsea/msigdb) and extracted only the Reactome data.
If you have any issues or questions about execution, please leave a comment through the repository and we will check. Or, please contact me via email: dohee_lee_@inu.ac.kr