Predicting drug efficacy based on the integrated breast cancer pathway model

This study is based on a simple hypothesis - "ideal" drugs for a patient can cure the patient's disease by modulating the gene expression profile of the patient to a similar level with those in healthy people, on the pathway level. To verify this hypothesis, we present a computational framework to evaluate drug effects on gene expression profiles in breast cancer. First, a breast cancer pathway model has been constructed by utilizing a computational connectivity maps (C-Maps) approach. This model includes important protein and drug information. In this pathway, specific drug-protein interactions (i.e. activation/inhibition) are annotated as edge attributes. Thus, we get a novel Pharmacology Effect Network, or PEN. We then develop a ranking algorithm called PET (i.e. Pharmacological Effect on Target) to combine gene expression information and our constructed PEN to evaluate specific drugs' efficacies. Finally, we applied PET and PEN to evaluate 23 breast cancer drugs. The ranking results clearly show the validity of our framework.