Drug-target and disease networks: polypharmacology in the post-genomic era
© Masoudi-Nejad et al.; licensee Springer. 2013
Received: 26 November 2013
Accepted: 27 November 2013
Published: 5 December 2013
With the growing understanding of complex diseases, the focus of drug discovery has shifted away from the well-accepted “one target, one drug” model, to a new “multi-target, multi-drug” model, aimed at systemically modulating multiple targets. Identification of the interaction between drugs and target proteins plays an important role in genomic drug discovery, in order to discover new drugs or novel targets for existing drugs. Due to the laborious and costly experimental process of drug-target interaction prediction, in silico prediction could be an efficient way of providing useful information in supporting experimental interaction data. An important notion that has emerged in post-genomic drug discovery is that the large-scale integration of genomic, proteomic, signaling and metabolomic data can allow us to construct complex networks of the cell that would provide us with a new framework for understanding the molecular basis of physiological or pathophysiological states. An emerging paradigm of polypharmacology in the post-genomic era is that drug, target and disease spaces can be correlated to study the effect of drugs on different spaces and their interrelationships can be exploited for designing drugs or cocktails which can effectively target one or more disease states. The future goal, therefore, is to create a computational platform that integrates genome-scale metabolic pathway, protein–protein interaction networks, gene transcriptional analysis in order to build a comprehensive network for multi-target multi-drug discovery.
Identification of the interaction between drugs and target proteins plays an important role in genomic drug discovery, in order to discover new drugs or novel targets for existing drugs. Due to the laborious and costly experimental process of drug-target interaction prediction, in silico prediction could be an efficient way of providing useful information in supporting experimental interaction data. Since, a small number of experimental drug-target interaction data has been reported in current and publicly available databases, this has motivated many researchers to develop high performance computational approaches capable of detecting new pairs of drug-target interaction efficiently.
Drug-target network (DTN) is a bipartite graph in which every link connects a drug to a protein if the protein is a known target of the drug (Yildirim et al.2007). To generate a DT network, all FDA-approved drugs and their known targets are used and the information about drug-target interactions could be extracted from known databases, including KEGG DRUG (Goto et al.2002), DrugBank (Wishart et al.2008) and others.
Yildirim et al. released some significant features related to the network topology of the DT network by applying network analysis to drugs and target proteins. Based on their findings, the membrane proteins are mostly targets of FDA-approved drugs which belong to the same Anatomical Therapeutic Chemical (ATC) class that naturally target the same proteins. To investigate the relationships between approved drugs, they integrated all available DT interaction data along with genetic-disease associations, gene expression and protein-protein interaction data.
A variety of computational methods have been proposed to analyze and detect new protein-ligand interactions. To the best of our knowledge, traditional computational approaches can be categorized in three classes, namely ligand-based, target-based, and text-mining methods. The Ligand-based approach like QSAR (Quantitative Structure Activity Relationship) uses machine learning methods to predict protein-ligand interaction by comparing a new ligand to the known ligands of a target protein (Butina et al.2002; Byvatov et al.2003). When the number of known ligands for a target protein of interest is insufficient, this approach couldn’t be effective in the prediction of interaction. Target-based approach or docking simulation rely on the 3D structure of proteins to predict protein-ligand interaction and it can’t be applied to proteins with unknown 3D structure (Cheng et al.2007; Donald2011; Morris et al.2009). This limitation is very serious for membrane proteins such as Ion channels and G-Protein Coupled Receptors (GPCRs) due to the complexity of determining 3D structures of most of these proteins. Another approach involves text mining methods which are based on keyword searching in literatures but the redundancy in the name of the gene/compound in the literatures is a major concern in this approach (Zhu et al.2005).
To predict the drug-target interaction, another interesting approach was proposed by Campillos et al. based on the side-effect similarities between known drugs (Campillos et al.2008). Nevertheless, this approach has been verified by in vitro experiments in some cases, it can only apply to the marketed drugs with known side effects and to the interaction between new drugs cannot be predicted by this approach. Recently, the importance of chemogenomic approaches in the domain of protein-ligand interaction prediction has grown fast (Dobson2004; Kanehisa et al.2006; Stockwell2000). These methods integrate both genomic spaces of target proteins, and chemical space of compounds, to predict new drug-target pairs. A unified space, namely “pharmacological space”, could be created by the integration of drug chemical structures, protein sequences and drug-target network topology to infer unknown drug-target interactions. The underlying idea is that drugs with similar chemical structures are likely to interact with similar proteins, and the prediction could be performed by extracting different features for drugs and proteins to define the similarity between two compounds or two proteins.
Based on this concept, a variety of statistical and learning methods have been developed to predict drug-target interaction (Bleakley and Yamanishi2009; Chen et al.2012; Cheng et al.2012; Mei et al.2012; van Laarhoven et al.2011; Xia et al.2010; Xie et al.2012; Yamanishi et al.2008). Although promising results have been gained by using these methods, most of them can only predict whether a drug interacts with a target protein, but reveal no further information about how this interaction occurred. Unlike these methods, Wang et al. have proposed a method based on the restricted Boltzmann machine to predict different types of interaction between drugs and targets (Wang and Zeng2013). This state-of-the-art technique involves research that takes into account the drug-target network in their analyses. “Network pharmacology” or “Systems pharmacology” is therefore a possible next paradigm in drug discovery which is generated by the advances in these areas (Hopkins20072008).
The improvements in drug discovery for complex diseases could be achieved by studying drug action through the network biology. With the growing understanding of complex diseases, the focus of drug discovery has shifted away from the well-accepted “one target, one drug” model, to a new “multi-target, multi-drug” model, aimed at systemically modulating multiple targets. In this context, polypharmacology has emerged as a new paradigm to overcome the recent decline in productivity of pharmaceutical research (Wermuth2004). Drug designers traditionally consider the polypharamacology as an unwanted property that must be removed or reduced to produce drugs that interact with a single target. But in the modern pharmaceutical industry, the hitherto strategy of, “one drug for one target for one disease”, has been considered responsible for the more-funding-less-drug problem. Recent research has shown that effective treatments of complex diseases are not possible by interventions at single nodes. To modify phenotypes, it is required that multiple proteins be modulated simultaneously. Thus, drugs which act on two or more targets of interest should be more impressive than single-target ones. Assessing the role of polypharmacology in drug action through drug-target network analysis may provide insights into which drugs are more efficacious in the treatment of complex diseases.
The integration of drug-target network with the human disease network revealed that drug targets are often involved in multiple diseases. Another analysis of the OMIM database of genetic associations manifest that most disease shares the genetic origins with others. According to these findings, drug repositioning has been growing in importance in the last few years for the application of known drugs in new directions. In addition, the mapping of drug targets on to the human protein interaction network reveals that the drug targets tend to be highly connected. Thus, the integration of drug-target network with other biological networks can help with drug discovery in producing drugs with more efficacies and less toxicity. Recent research also indicates that besides primary network parameters, more complex network metrics such as motifs and clusters may also be appropriate parameters for controlling the metabolic systems. Moreover, for metabolic networks, enzyme-centric networks could be more reliable in the context of controllability, although little attention has been paid to such networks in systems controllability. The outcomes of metabolic network controllability could create insights into the discovery of novel drug targets (Asgari et al.2013).
An important notion that has emerged in post-genomic drug discovery is that the large-scale integration of genomic, proteomic, signaling and metabolomic data can allow us to construct complex networks of the cell that would provide us with a new framework for understanding the molecular basis of physiological or pathophysiological states. Such an integrated view has important implications in improving our understanding of the disease phenotypes by viewing them as perturbations in a complex system rather than as effects on a selective set of proteins. Using such a framework, network based drug discovery aims to harness this knowledge to investigate and understand the impact of interventions, such as candidate drugs, on the molecular networks that define different states and therefore can significantly complement the existing drug discovery pipelines.
An emerging paradigm of polypharmacology in the post-genomic era is that drug, target and disease spaces can be correlated to study the effect of drugs on different spaces and their interrelationships can be exploited for designing drugs or cocktails which can effectively target one or more disease states (Janga and Tzakos2009). This understanding could lead to the introduction of new multidrug treatments, side-effect prediction and the identification of new drug targets. The future goal, therefore, is to create a computational platform that integrates genome-scale metabolic pathway, protein–protein interaction networks, gene transcriptional analysis in order to build a comprehensive network for multi-target multi-drug discovery (Figure 1).
Authors appreciate great help from LBB’s members during their course of study.
- Asgari Y, et al.: Controllability in metabolic cancer networks according to drug targets as driver nodes. PLOS ONE 2013, 8: 11. e79397 e79397View ArticleGoogle Scholar
- Bleakley K, Yamanishi Y: Supervised prediction of drug-target interactions using bipartite local models. Bioinformatics 2009, 25: 2397–2403. 10.1093/bioinformatics/btp433View ArticleGoogle Scholar
- Butina D, et al.: Predicting ADME properties in silico: methods and models. Drug Discov Today 2002, 7: 83–88. 10.1016/S1359-6446(01)02099-2View ArticleGoogle Scholar
- Byvatov E, et al.: Comparison of support vector machine and artificial neural network systems for drug/nondrug classification. J Chem Inf Comput Sci 2003, 43: 1882–1889. 10.1021/ci0341161View ArticleGoogle Scholar
- Campillos M, et al.: Drug target identification using side-effect similarity. Science 2008, 321: 263–266. 10.1126/science.1158140View ArticleGoogle Scholar
- Chen X, et al.: Drug-target interaction prediction by random walk on the heterogeneous network. Mol Biosyst 2012, 8: 1970–1978. 10.1039/c2mb00002dView ArticleGoogle Scholar
- Cheng AC, et al.: Structure-based maximal affinity model predicts small molecule druggability. Nat Biotechnol 2007, 25: 71–75. 10.1038/nbt1273View ArticleGoogle Scholar
- Cheng F, et al.: Prediction of drug-target interactions and drug repositioning via network-based inference. PLoS Comput Biol 2012, 8: e1002503. 10.1371/journal.pcbi.1002503View ArticleGoogle Scholar
- Dobson C: Chemical space and biology. Nature 2004, 432: 824–828. 10.1038/nature03192View ArticleGoogle Scholar
- Donald BR: Algorithms in Structural Molecular Biology. Cambridge, MA, USA: MIT Press; 2011.Google Scholar
- Goto S, et al.: LIGAND: database of chemical compounds and reactions in biological pathways. Nucleic Acids Res 2002, 30: 402–404. 10.1093/nar/30.1.402View ArticleGoogle Scholar
- Hopkins AL: Network pharmacology: the next paradigm in drug discovery. Nat Chem Biol 2008, 4: 682–690. 10.1038/nchembio.118View ArticleGoogle Scholar
- Hopkins AL: Network pharmacology. Nat Biotechnol 2007, 25: 1110–1111. 10.1038/nbt1007-1110View ArticleGoogle Scholar
- Janga SC, Tzakos A: Structure and organization of drug-target networks: insights from genomic approaches for drug discovery. Mol Biosyst 2009, 5: 1536–1548. 10.1039/b908147jView ArticleGoogle Scholar
- Kanehisa M, et al.: From genomics to chemical genomics: new developments in KEGG. Nucleic Acids Res 2006, 34: 354–357. 10.1093/nar/gkj102View ArticleGoogle Scholar
- Mei JP, et al.: Drug-target interaction prediction by learning from local information and neighbors. Bioinformatics 2012, 29: 238–245.View ArticleGoogle Scholar
- Morris GM, et al.: AutoDock4 and AutoDockTools4: automated docking with selective receptor flexibility. J Comput Chem 2009, 30: 2785–2791. 10.1002/jcc.21256View ArticleGoogle Scholar
- Stockwell B: Chemical genetics: ligand-based discovery of gene function. Nat Rev Genet 2000, 1: 116–125.View ArticleGoogle Scholar
- Van Laarhoven T, et al.: Gaussian interaction profile kernels for predicting drug-target interaction. Bioinformatics 2011, 27: 3036–3043. 10.1093/bioinformatics/btr500View ArticleGoogle Scholar
- Wang Y, Zeng J: Predicting drug-target interactions using restricted Boltzmann machines. Bioinformatics 2013, 29: 126–134. 10.1093/bioinformatics/bts637View ArticleGoogle Scholar
- Wermuth CG: Multitarget drugs: the end of the ’one-target-one-disease’ philosophy? Drug Discov Today 2004, 9: 826–827. 10.1016/S1359-6446(04)03213-1View ArticleGoogle Scholar
- Wishart DS, et al.: Drugbank: a knowledgebase for drugs, drug actions and drug targets. Nucleic Acids Res 2008, 36: 901–906.View ArticleGoogle Scholar
- Xia Z, et al.: Semi-supervised drug-protein interaction prediction from heterogeneous biological spaces. BMC Syst Biol 2010,4(suppl. 2):S6.View ArticleGoogle Scholar
- Xie L, et al.: Novel computational approaches to polypharmacology as a means to define responses to individual drugs. Annu Rev Pharmacol Toxicol 2012, 52: 361–379. 10.1146/annurev-pharmtox-010611-134630View ArticleGoogle Scholar
- Yamanishi Y, et al.: Prediction of drug-target interaction networks from the integration of chemical and genomic spaces. Bioinformatics 2008, 24: 232–240. 10.1093/bioinformatics/btn162View ArticleGoogle Scholar
- Yildirim MA, et al.: Drug-target network. Nat Biotechnol 2007, 25: 1119–1126. 10.1038/nbt1338View ArticleGoogle Scholar
- Zhu S, et al.: A probabilistic model for mining implicit ‘chemical compound-gene’ relations from literature. Bioinformatics 2005, 21: 245–251.Google Scholar
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.