GraphCDR

Source code and data for "GraphCDR: A graph neural network method with contrastive learning for cancer drug response prediction"

Requirements

• Python >= 3.6
• PyTorch >= 1.4
• PyTorch Geometry >= 1.6
• hickle >= 3.4
• DeepChem >= 2.4
• RDkit >= 2020.09

Usage

• please unzip the file: data/Drug/drug_graph_feat.zip first.
• python graphCDR.py <parameters>
• python graphCDR-ccle.py <parameters>

Case study

As GDSC database only measured IC50 of part cell line and drug pairs. We applied GraphCDR to predicted the missing types of responses. The predicted results can be find at data/Case study (missing pairs).xlsx