• hetgnn_main.py
  • Main script for HetGNN. Layer configurations can be modified in gat_dependency/GAT_model.py.

• DLP_main.py
  • Implements DLP with the Whole split method.
• DLP_cellsplit.py
  • Implements DLP with the Cell Only split method.

• construct_heterogeneous_graph_jihwan.py
  • Script for constructing a heterogeneous graph.
• construct_heterogeneous_graph_pyG_jihwan.py
  • Script for constructing a heterogeneous graph with biological features using Pytorch Geometric.

• Conda list
  • Conda environment settings required for running the project.

• Final_Result
  • Contains code for generating figures, tables, and lists of reprioritized genes. (See Jupyter Notebook in folder)
  • Contains generated figures by pptxm can be checked in Final_Result/Fugures made.pptx.

You can run the scripts with the following format, adjusting arguments using --arg:

python hetgnn_main.py --epochs 30 --lr 0.0001 --batch_size 256 --emb_dim 512 --seed 42 --exp_name example_test

All resources and data sources mentioned in this thesis are publicly accessible.

• Protein-protein network: Download FIsInGene with annotations

• DepMap Public 23Q2 Files: View all data
  • DepMap Score: CRISPRGeneEffect.csv
  • Cell Line Name: Model.csv
  • Expression: OmicsExpressionProteinCodingGenesTPMLogp1.csv
  • Copy Number Variation: OmicsCNGene.csv
  • Control: AchillesCommonEssentialControls.csv

• CGP: Download CGP Symbols (GMT)
• CP: Download CP Symbols (GMT)
• GO: Download GO Symbols (GMT)
• BP: Download BP Symbols (GMT)

All results can be found in WandB and check comments in overview first!

• DLP: View DLP Project
• HetGNN: View HetGNN Project
• HetGNN-Lin: View HetGNN-Lin Project