Predicting effective drug combinations via network propagation

Drug combinations are frequently used in treating complex diseases including cancer, diabetes, arthritis and hypertension. Most drug combinations were found in empirical ways so there is a need of efficient computational methods. Here we present a novel method based on network analysis which estimates the efficacy of drug combinations from a perturbation analysis performed on a protein-protein association network. The results suggest that those drugs are likely to form effective combinations that perturb a large number of proteins in common, even if the original targets are found in seemingly unrelated pathways.