Link to bioRxiv paper: http://biorxiv.org/cgi/content/short/2020.09.13.295063v1?rss=1 Authors: Wang, Q., Yan, G. Abstract: (1) Background: Compared with monotherapy, efficacious drug combinations can increase the therapeutic effect, decrease drug resistance of experimental subjects and the side effects of drugs. Therefore, efficacious drug combinations are widely used in the treatment of complex diseases, such as various cancers. However, compared with the mathematical model and computational method, experimental screening efficacious drug combinations is time-consuming, costly, laborious, and inefficient; (2) Methods: we predicted efficacious drug combinations in cancer based on random walk with restart (ERWR). An efficacious score can be obtained between any two individual drugs by ERWR; (3) Results: As a result, we analyzed the rationality of the efficacious score first. Besides, compared with the other methods by leave-one-out cross-validation, all the Area Under Receiver Operating Characteristic Curves (AUROCs) of ERWR were higher for data sets of breast cancer, colorectal cancer, and lung cancer. Moreover, the case study of breast cancer showed that ERWR could discover potential efficacious drug combinations; (4) Conclusions: These results suggest that ERWR is a novel way to discover efficacious drug combinations in cancer, which provides new prospects for cancer treatment. Furthermore, ERWR is a semi-supervised learning framework that can be used to predict combinations of drugs for other complex diseases. Copy rights belong to original authors. Visit the link for more info