This release introduces the first pseudocode scaffold for DrugResponse-GNN, a pathway-bottleneck GNN for drug sensitivity prediction and interpretability.
Highlights
Harmonization framework for CCLE, GDSC, NCI-60, and CTRP panels.
Pathway-bottleneck GNN architecture with modular encoder and graph integration stubs.
Baseline adapters for drGAT, DRPreter, and a simple regression baseline.
Cross-panel evaluation scaffold (e.g., CCLE→GDSC).
Interpretability engine: pathway attributions, stability across seeds, and optional subgraph highlighting.
Case study stubs for PI3K, PARP, and EGFR inhibitor families.
Next steps
Implement preprocessing of panels and omics alignment.
Add evaluation metrics and visualization.
Expand case studies and cross-panel generalization.