Application of Network Science in Drug‐Protein‐Disease Interaction and Drug Repositioning

Network science is exploited in various disciplines for better understanding and analysis of components of a complex system. In recent years, several studies in the fields of computational prediction of drug-target, drug-disease, and drug-drug interaction as well as drug repositioning based on network modeling have been done in the laboratory of Complex Biological Systems and Bioinformatics (CBB), Institute of Biochemistry and Biophysics (IBB), University of Tehran, with collaboration of some other research labs. In this article, a brief review of these research works is presented. Prediction of Drug-Target Interaction (DTI), plays important role in drug design. In our recent research, drug interactions with four groups of target proteins has been investigated.These groups include Enzymes, Ion Channels, G-protein Coupled Receptors (GPCR) and Nuclear Receptors. In one of studies, a recommendation system was designed, named DTI- BECM (Drug Target Interaction Based on Eigenvector Centrality Measure). This recommender system uses well-known SIMCOMP algorithm for similarity measurement between chemical structures of drugs. Also, it uses the generalized Eigen vector centrality measure for bi-partite graphs. In a complementary study, molecular descriptors of drugs were used for improving the performance of system. These methods, successfully repositioned some known drugs for new targets.One of the latest projects, was performed to establish a framework for discovering unknown Disease-Drug Interactions. A random walk algorithm with restart was developed on a three-layer heterogeneous network to prioritize candidate diseases associated with drugs. Nine different networks (drug similarity, target similarity, disease similarity, gene similarity, drug-target interaction, target-pathway interaction, pathway-gene interaction, gene-disease interaction and drug-disease interaction network) were composed to shape a single heterogeneous network. The proposed method predicts therapeutic indices for existing or new potential drugs.Proposing a solution for Drug-Drug Interaction (DDI), is another discipline in the drug research field that we have focused on it. Drug-Drug Interaction (DDI) is an unsafe drug side effect occurs unexpectedly when more than one drug is utilized. Identification of drugs with undesirable interactions with experimental methods is almost impossible. In the proposed method, different features are extracted from drugs for DDI prediction. These features include the fingerprint of chemical structure, label side effect, reported side effects, drug targets, and Anatomical Therapeutic Chemical (ATC) classification codes. Besides, we used the molecular descriptors of drugs as a new source of information for DDI prediction. Applying appropriate feature selection methods and using Support Vector Regression (SVR), alongside integration different information sources, showed significant results in comparison with other methods.