Computational Identification of Drugs Synergy in Cancer Treatment

Abstract

Abstract Background: Combinational pharmacotherapy with the synergistic/additive effect is a powerful treatment strategy for complex diseases such as malignancies. Identifying synergistic combinations with various compounds and structures requires testing a large number of compound combinations. Detecting synergistic chemotherapeutic agents from a considerable number of candidates is highly challenging. Aim: This study proposes a model to predict the synergistic drug combinations by using the deep neural network on cancer cell lines and genomic and physicochemical features. Methods: The base of work is on the information of drugs and genomic features information which compiles into the network of features. The network of features comprises the interaction among the features.Then the vector of features is extracted based on the interaction among the drugs, cell lines, and features. Finally, the feature vectors are fed into the deep neural network to achieve the synergistic prediction results. Results: The proposed model was compared to other machine learning methods such as Gradient Boosting Machines, Random Forests, and Support Vector Machines on the publicly available synergy dataset. Applying the model to classify the drug combinations resulted in the high predictive performance of an AUC of 0.97, and the best accuracy performance was 92.17%. We envision that the proposed model could be a valuable tool for selecting novel synergistic drug combinations. Conclusion: In this study, we analyzed the effect of drug chemical structure features and pharmacogenomics features on identifying synergistic drug combinations. Our prediction model may help reduce search space and accelerate the identification of clinically effective synergistic drug combinations.