Send Orders for Reprints to reprints@benthamscience.ae Current Pharmaceutical Design, 2016, 22, 3569-3575 3569 Systems Pharmacology: A Unified Framework for Prediction of Drug-Target Interactions Duc-Hau Le*a and Ly Leb a School of Computer Science and Engineering, Water Resources University, 175 Tay Son, Dong Da, Hanoi, Vietnam; bSchool of Biotechnology, International University, Vietnam National University, Vietnam Abstract: Background: Drug discovery is one important issue in medicine and pharmacology area Traditional methods using target-based approach are usually time-consuming and ineffective Recently, the problems are approached in a system-level view and therefore it is called systems pharmacology This research field deals with the problems in drug discovery by integrating various kinds of biomedical and pharmacological data and using advanced computational methods Ultimately, the problems are more effectively solved One of the most important problem in systems pharmacology is prediction of drug-target interactions Methods: In this review, we are going to summarize various computational methods for this problem Results: More importantly, we formed a unified Duc-Hau Le framework for the problem In addition, to study human health and disease in a more systematically and effectively, we also presented an integrated scheme for a wider problem of prediction of disease-gene-drug associations Conclusion: By presenting the unified framework and the integrated scheme, underlying computational methods for problems in systems pharmacology can be understood and complex relationships among diseases, genes and drugs can be identified effectively Keywords: Drug-target interaction, network-based approach, machine learning-based approach, drug-disease association, disease-gene association, drug-gene-disease association Received: March 14, 2016 Accepted: April 15, 2016 INTRODUCTION The development of a drug from an original idea to the market usually takes ten to seventeen years and costs about billion US dollar on average [1, 2] In addition, several of them have been withdrawn due to adverse/side effects The major reason is that the drugs not only interact to therapeutic targets but also to offtargets In addition, drugs can interact with other drugs or chemicals from food when they are used in the same patients Therefore, drug discovery are needed to be considered at system-level, in which many kinds of data for biomedical and pharmacological instances are investigated simultaneously [3] In addition, computational methods are also used to recognize interactions and mutual effects among the instances Therefore, system pharmacology, which is a combination between systems biology and pharmacology, becomes promising approach to deal with this problem In which, systems biology is the way to investigate interactions among cellular components at system-level Therefore, systems pharmacology is a biological and chemical approach for health and diseases [4-6] More specifically, systems pharmacology studies how drugs work (i.e., mode of action) at different levels such as whole human body, organs, tissues, or cellular components [7] Computational methods in systems pharmacology are usually used along with various kinds of data such as –omics (genomics, transcriptomics, proteomics, metabolomics, phenomics, interactomics) as well as pharmacological data, etc.… [8] The final goals is to elucidate therapeutic mechanisms of drugs (i.e., how drug work on different pathways, cell types and tissues) and enable the process of drug discovery and development to become more effective Based on these computational methods, many state-of-the-art technologies in modern computer-aided drug design have been developed [9] There is a number of problems in drug discovery such as identification of targets, drug repositioning, drug efficacy (e.g., safety, *Address correspondence to this author at School of Computer Science and Engineering, Water Resources University, 175 Tay Son, Dong Da, Hanoi, Vietnam; Tel/Fax: +84-912324564; E-mails: hauldhut@gmail.com 1381-6128/16 $58.00+.00 adverse effects and responses) [10] Many computational methods have been proposed to solve these problems [11-16] In this study, we focus on the most important problem in drug discovery, which is prediction of novel drug-target interactions [17-19] To identify targets, experimental studies usually approach the problem based on either target or phenotype [20] Meanwhile, computational methods for this problem are very diverse, in which novel interactions are predicted based on known drug-target interactions using various kinds of supporting biomedical and pharmacological data [21, 22] Irrespectively to this diversity, computational methods are classified into two main approaches: i) network-based, ii) machine learning-based There has been a number of studies which review the computational methods for predictions of drug-target interactions [15, 23, 24] However, these studies have simply listed proposed methods but not yet proposed a unified framework of the computational methods for the problem For instance, studies [23, 24] focused on only machine learning-based methods Recently, Chen et al [15] has reviewed both network-based as well as machine learning-based methods However, this study only enumerated proposed methods but not yet proposed any general framework for this problem In this study, we are going to review the computational methods as well as propose a unified framework for the prediction of drug-target interactions In addition, to study human health and disease in a more systematically and effectively, we present an integrated scheme for a wider problem of prediction of diseasegene-drug associations A UNIFIED FRAMEWORK FOR PREDICTION OF DRUG-TARGET INTERACTIONS Drug discovery by systems pharmacological approaches requires the integration of various kinds of data [8, 25] This is systems biology and chemistry-based approach [5] for human health and diseases [4] Therefore, all computational methods integrate different kinds of data such as –omics as well as pharmacological data [26, 27] Many computational methods have been proposed for the drug discovery [8], however, a majority of them are based on © 2016 Bentham Science Publishers 3570 Current Pharmaceutical Design, 2016, Vol 22, No 23 Le and Le Fig (1) A unified framework for prediction of drug-target interactions networks of biomedical and pharmacological instances (e.g., drugs, targets and diseases), and therefore called network pharmacology [10, 28-32] The network-based methods start with construction of heterogeneous networks of biomedical and pharmacological instances by combining homogeneous ones (i.e., drug similarity networks, target similarity networks and phenotypic disease similarity network) and bipartite ones (i.e., drug-target networks, drug-disease networks and disease-gene networks) In addition, network-based methods consider action modes of drugs in the context of interactions among cellular components [28] After heterogeneous networks are constructed, computational methods are then proposed to identify novel interactions between drugs and targets Besides the network-based approaches, other major approaches are based on machine learning techniques [23, 24] The machine learning-based approaches also require integrating various kinds of data to build feature vectors for biomedical and pharmacological instances as well as calculate similarity matrices among the instances After that, a suitable machine learning algorithm is proposed to build a predictive model for identification of novel drug-target interactions In general, these two main approaches are both based on similarity measures because they are all under an assumption that chemically or pharmacologically similar drugs target similar target proteins [33] Network-based approaches use similarity measures to construct the heterogeneous networks and calculate relative similarity between candidate interactions and known ones Meanwhile, machine learning-approaches use them for building kernel matrices as well as calculating similarity among feature vectors/interaction profiles In addition, three main types of data spaces are used in these approaches, they are genomic space (i.e., general proteins as well as target ones), chemical space (i.e., chemical structures of drugs) and pharmacological space (i.e., general phenotypes as well as drug-affected ones) However, in order to build kernel matrices, Systems Pharmacology: A Unified Framework for Prediction similarity networks and feature vectors, many types of data have been used such as chemical structure of drugs, sequence of proteins, network of instances and text (e.g., ontologies and literature) This requires different similarity measures to effectively estimate how similar two instances are In addition, the proposed methods are not only application of a network/machine learning-based algorithm on a raw data, but they are a processes containing sequential steps such as data preparation, model building, algorithm selection, model assessment and result analysis Therefore, the categorization of the studies into two main approaches (i.e., network-based and machine learning-based) are simply based on algorithm selection An example for the intersection between the two main approaches is that the kernel matrices in machine learning-based methods can be constructed using kernel functions on graph/network In addition, machine learning-based methods can be started by building a bipartite network such as known drug-target interaction networks [34] After that, learning algorithms were used to predict novel ones [35-37] Moreover, network-based methods are usually started with construction of heterogeneous network of instances, then a networkbased algorithm is proposed to rank candidates (e.g., candidate targets) based on their functional similarities to known ones (e.g., known drug-related targets) Therefore, this can be considered as a learning process where the model is built totally based on labeled data (e.g., known drug-related targets) (i.e., supervised learning) or based on both labeled and unlabeled data (e.g., other targets in the network) (i.e., semi-supervised learning) Another common issue in the two main approaches is data representation In general, the computational approaches represent data in two ways: i) drugs and targets are considered as separate instances (e.g., given a known drug and its known targets, predict novel targets of this drug), and ii) drugs and targets are consider as pairs (e.g., given known drugtarget interactions, predict novel ones) Fig (1) shows a unified framework for the two main approaches for the prediction of drugtarget interactions In the next section, we are going to summarize some typical studies in each approach to clarify the point 2.1 Network-Based Approaches Typical studies in this approach usually begin by constructing a heterogeneous network of biomedical and pharmacological instances, then a network-based algorithm is proposed to identify novel drug-target interactions For instance, Chen et al [33] built a heterogeneous network that combines a drug similarity network and a target protein similarity network by known drug-target interactions In which, the drug similarity network was constructed based on similarity in chemical structures of drugs and the target protein similarity network was constructed based on similarity among sequences of target proteins In addition, information of known drugtarget interactions was also embedded into these two similarity networks After that, a random walk with restart (RWR) algorithm, which was successfully used for prediction of disease-associated genes [38] and microRNA [39] on a heterogeneous network, was used to identify novel target proteins related to a given drug The underlying assumption of this method (both for the construction of the heterogeneous network as well as the selection of RWR algorithm) is that chemically similar drugs target to similar target proteins Therefore, this method was proven to outperform ones, which was solely based on target protein similarity network Using the same algorithm as in [33], however, Seal et al [40] additionally used an extensive drug-target network and a drug similarity network with links among drugs were built based on the molecular similarity with chemical fingerprints Another network-based method [41] was also proposed to predict novel drug-target interactions They did not specifically build the heterogeneous network of drugs and targets However, they still based chemical similarity between drugs and sequence similarity between target proteins to propose predictive drug-based similarity inference (DBSI) and target-based similarity inference (TBSI) models, respectively In addition, they proposed a network-based inference (NBI) model based Current Pharmaceutical Design, 2016, Vol 22, No 23 3571 on topological similarity in the drug-target bipartite network Taken together, network-based methods are usually based on the similarity between drugs or between targets (i.e., in the form of similarity networks or similarity matrices) and a known drug-target interaction network 2.2 Machine Learning-Based Approaches These approaches also mainly used the three data spaces as in network-based ones (i.e., genomic, chemical and pharmacological spaces) A main strategy of these methods is to build kernel matrices based on similarity among drugs and similarity among targets Basically, drugs are chemical compounds, which are in the form of graph structure, therefore graph-based kernel functions [42] are usually used to calculate kernel matrices Also, target proteins can be represented as sequences, therefore string-based kernel functions can be used to calculate kernel matrices [43] There are two main approaches for machine learning-based methods for the prediction of drug-target interactions, they are chemogenomics and pharmacogenomics In which, the former is based on an assumption that chemically similar drugs usually target similar target proteins [35, 44-46] Meanwhile, the latter is based on another that phenotypically similar drugs interact to similar target proteins [37, 47], in which, phenotypes of drugs can be represented by effects of drugs The built kernel matrices are then used for a learning model to predict novel drug-target interactions Most of the machine learning-based methods proposed for the problem are based on kernel-based supervised learning techniques For instance, Jacob et al [45] used a product kernel (i.e., a combination of chemical structure-based kernel matrix for ligands and sequence-based kernel matrix for target proteins) In addition to these kinds of data, Yamanishi et al [35] utilized topological properties of the drug-target network to build a kernel-based logistic regression prediction model By combining both chemical and pharmacological data of drugs, Yamanishi et al [37] proposed to use a supervised learning method on the bipartite drug-target network In addition, Takarabe et al [48] also used the product Kernel as in [45] and the kernel-based regression model as in [35], however, they built pharmacological kernel matrix instead of the chemical structure-based kernel matrix as in [35] Finally, Bleakley et al [36] used the same kernel matrices as [48] in a bipartite local models (BLM) on the bipartite drug-target network Based on the bipartite drug-target network, some studies built interaction profiles of drugs and targets [49-51] For instance, studies [50, 51] proposed a method, namely GIP, by building Gaussian kernels based on these interaction profiles After that, a regularized least square-based classifier was used to predict novel drug-target interactions [50] Besides, other machine learning models have been used for the problem such as conditional random field (CRF), a probabilistic graphical model [52] and Bayesian matrix factorization [53] Taken together, these machine learning-based methods mainly used similarity measures to build kernel matrices These kernel matrices were usually built separately for drugs and targets After that, they were combined for predictive models Other supervised approaches are feature vector-based, in which, drug-target pairs were represented as feature vectors Based on features of drugs and targets extracted from drug and target-related data as well as known drug-target interactions, classifiers were then constructed to predict novel drug-target interactions There have been several learning models used for the problem such as k-nearest neighbor (kNN) [54], logistic regression [55, 56], support vector machines (SVM) [57-59] and ensemble methods [60, 61] Interestingly, most of these methods are kernel-based, where kernel matrices are calculated from feature vectors of training instances [56-59, 61] A major limitation of the above supervised learning-based methods is that the predictive models were built based on only labeled data (i.e., known drug-target interactions) However, this kind 3572 Current Pharmaceutical Design, 2016, Vol 22, No 23 Le and Le Fig (2) An integrated scheme for prediction of drug-gene-disease associations of data is very limited due to the cost of labeling process In addition, some of them are binary-classification models This means that negative training instances (i.e., non-drug-target interactions) have to be specified in training process However, there is no such experimentally verified interactions in literature Therefore, semisupervised learning methods, which learn from both labeled and unlabeled data (i.e., unknown drug-target interactions), have been proposed to overcome this limitations [62-64] These methods also utilized kernel/similarity matrices constructed from chemical and genomic spaces as well as the bipartite drug-target network AN INTEGRATED SCHEME FOR PREDICTION OF DRUG-GENE-DISEASE ASSOCIATIONS Approaches in systems pharmacology are based on various kinds of data and computational methods In which, three main entities (i.e., drug, disease and gene/target protein) should be considered in the same context [65-67] There have been many individual databases for each entity, but few for all of them such as C2Maps [68], EU-ADR corpus [69] Therefore, there is a pressing need to build such the databases In the future, these databases can be formed and more comprehensive with results from experimental studies as well as text mining techniques [3, 70, 71] The relationships among these entities pose three challenges, one of them is the prediction of drug-target interactions The two remaining ones are prediction of disease-associated genes/proteins and prediction of novel drug-disease associations (also known as drug repositioning/repurposing) Although, the three problems have their own goals in biomedicine and pharmacology, but they have the same target in algorithmic view, which is prediction of novel binary interactions/associations This is also a popular problem in bioinformatics More importantly, they are all based on a “guilt-byassociation” assumption that functionally similar drugs/diseases target/relate to functionally similar target proteins/genes, respectively Among them, prediction of novel disease-gene associations is very popular problem in biomedicine research and well-studied by both network- and machine learning-based methods [72-80] Therefore, we could use or adapt the methods used for this problem for the prediction of drug-target interactions In addition, prediction of novel drug-disease associations have been also studied for years [16, 32, 71, 81, 82] As aforementioned, many computational methods have been proposed for prediction of drug-target interactions However, some of them were built and assessed based on simple settings and ultimately they may not be used for practical problems Therefore, computational methods are needed to take intensive consideration on the problem modeling as well as assessment methods [83] In addition, obviously, irrespective of the network- or machine learning-based methods, the calculation of similarity among instances in different data spaces to construct similarity/kernel matrices and similarity networks is the most important step Data integration/fusion from various spaces is also important since the problem is considered in multi-dimension as well as systematically Therefore, we should also take consideration on selection of integration methods, in which kernel-based data fusion methods can be suitable ones since the data is mostly represented as kernel matrices Indeed, Wang et al [84] is a pioneer applying kernel-based data fusion for the prediction of drug-target interactions However, it should be noted that this data fusion method has been used popularly for prediction of disease-gene association for years [85-87] In addition, many algorithms have been proposed for kernel-based data fusion [88] such as Lp-norm MKL [89], which could be used for kernelbased problems Therefore, they could be used to improve the prediction of drug-target interactions Besides, to meet practical problems, more suitable machine learning models should be considered for the prediction of drugtarget interactions such as positive and unlabeled (PU) learning technique, which has been successfully applied in prediction of disease-gene associations [90] In addition, the combination between the PU learning technique and kernel-based data fusion or ensemble methods could provide better way for the problem as did for prediction of disease-gene associations in [91] and [92], respectively For network-based approaches, the calculation of similarity between instances in constructed networks can be done with graphbased kernel functions [93, 94] Some of them have been used successfully for prediction of disease-gene associations Similarly, Systems Pharmacology: A Unified Framework for Prediction Current Pharmaceutical Design, 2016, Vol 22, No 23 these machine learning- and network-based methods could be used to predict drug-target interactions more effectively In parallel to proposal of computational methods and construction of databases, development of tools for prediction of novel drug-target interactions is also needed [95, 96] There has been a number of such the tools such as DINIES [97], iGPCR-Drug [98] Also, many tools have been developed for prediction of diseasegene associations [74, 99, 100] as well as several ones for prediction of drug-disease associations such as DRAR-CPI [101] and PROMISCUOUS [102] Obviously, the above mentioned tools are developed individually for each the problem of prediction Therefore, an integrated tool predicting drug-gene-disease associations is needed such as DT-Web [103], which can predict both drug-target interactions and drug-disease associations Taken together, to study human health and disease in a more systematically and effectively, the drug-gene-disease associations should be considered in the same context Therefore, all proposed computational prediction methods, constructed databases and developed tools should meet this requirement Figure (2) shows an integrated scheme for prediction of drug-gene-disease associations REFERENCES CONCLUSION After human genome project, it has been expected that several new effective drugs will be developed Unfortunately, most of drug targets are proteins which go through significant post translation modification or involve in complex network Systems pharmacology therefore becomes the modern approach for drug discovery and development in post genomic era In this study, we have reviewed various computational approaches proposed for this problem More importantly, by analyzing the proposed methods, we recognized that they are based on the same assumption that similar drugs target similar target proteins Based on this assumption, they use similarity measures to calculate similarity among instances in different data spaces to construct similarity/kernel matrices and similarity networks After that, these matrices and networks are used in network- or machine learning-based computational models to predict novel drug-target interactions To be more intuitively, we proposed a unified framework of computational methods for the prediction of drug-target interactions In addition, we extended the problem to effectively study human health and diseases by considering the relationships among drugs, gene/target proteins and diseases in the same context Interestingly, we found that the drug-gene-disease relationship pose three key challenges (i.e., prediction of drugtarget interactions, prediction of disease-associated genes/proteins and prediction of novel drug-disease associations) with similar underlying assumptions (i.e., guilt-by-association) and the same algorithmic view (i.e., prediction of binary interactions) Therefore, predictive models and data fusion methods could be used exchangeably among the three problems Similarly, we also presented an integrated scheme for prediction of drug-gene-disease associations Finally, with high-throughput technologies in next generation sequencing, GWAS has yielded large amount of genomic data Therefore, computational methods for the problems in drug discovery should deal with this wealth data [104, 105] which open a wide road for research in personalized medicine [106-108] In addition, novel predicted drug-target interactions by computational methods need to be experimentally verified in wet-laboratory [10] CONFLICT OF INTEREST The authors confirm that this article content has no conflict of interest ACKNOWLEDGEMENTS This research is funded by Vietnam National Foundation for Science and Technology Development (NAFOSTED) under grant number 102.01-2014.21 [1] [2] [3] [4] [5] [6] [7] [8] [9] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] [24] [25] [26] [27] 3573 Ashburn TT, Thor KB Drug repositioning: identifying and developing new uses for existing drugs Nat Rev Drug Discov 2004; 3: 673-83 Hughes JP, Rees S, Kalindjian SB, Philpott KL Principles of early drug discovery Br J Pharmacol 2011; 162: 1239-49 Yang Y, Adelstein SJ, Kassis AI Target discovery from data mining approaches Drug Discov Today 2009; 14: 147-154 Somvanshi P, Venkatesh KV A conceptual review on systems biology in health and diseases: from biological networks to modern therapeutics Syst Synth Biol 2014; 8: 99-116 Brown JB, Okuno Y Systems biology and systems chemistry: new directions for drug discovery Chem Biol 2012; 19: 23-8 Robinson R From Complex Behavior to New Drugs: Chemoinformatics Shows the Way PLoS Biol 2013; 11: e1001713 Zhao S, Iyengar R Systems pharmacology: Network analysis to identify multiscale mechanisms of drug action Ann Rev Pharmacol Toxicol 2012; 52: 505-21 Ou-Yang S-s, Lu J-y, Kong X-q, Liang Z-j, Luo C, Jiang H Computational drug discovery Acta Pharm Sin 2012; 33: 1131-40 Dalkas GA, Vlachakis D, Tsagkrasoulis D, Kastania A, Kossida S State-of-the-art technology in modern computer-aided drug design Brief Bioinform 2013; 14: 745-52 Chandra N, Padiadpu J Network approaches to drug discovery Expert Opin Drug Discov 2013; 8: 7-20 Iorio F, Saez-Rodriguez J, Bernardo Dd Network based elucidation of drug response: from modulators to targets BMC Syst Biol 2013; 7: 139 Ryall K, Tan A Systems biology approaches for advancing the discovery of effective drug combinations J Cheminform 2015; 7: Takarabe M, Shigemizu D, Kotera M, Goto S, Kanehisa M Network-based analysis and characterization of adverse drug-drug interactions J Chem Inf Model 2011; 51: 2977-85 Cheng F, Zhao Z Machine learning-based prediction of drug–drug interactions by integrating drug phenotypic, therapeutic, chemical, and genomic properties J Am Med Informatics Assoc 2014; 21: e278-86 Chen X, Yan CC, Zhang X, et al Drug–target interaction prediction: databases, web servers and computational models Brief Bioinform 2015; [Epub ahead of print] Li J, Zheng S, Chen B, Butte AJ, Swamidass SJ, Lu Z A survey of current trends in computational drug repositioning Brief Bioinform 2016; 17(1): 2-12 Overington JP, Al-Lazikani B, Hopkins AL How many drug targets are there? Nat Rev Drug Discov 2006; 5: 993-96 Imming P, Sinning C, Meyer A Drugs, their targets and the nature and number of drug targets Nat Rev Drug Discov 2006; 5: 821-34 Keiser MJ, Setola V, Irwin JJ, et al Predicting new molecular targets for known drugs Nature 2009; 462: 175-81 Schenone M, Dancik V, Wagner BK, Clemons PA Target identification and mechanism of action in chemical biology and drug discovery Nat Chem Biol 2013; 9: 232-40 Chandra N Computational systems approach for drug target discovery Exp Opin Drug Discov 2009; 4: 1221-36 Yamanishi Y Chemogenomic approaches to infer drug–target interaction networks In: Mamitsuka H, DeLisi C, Kanehisa M, Eds Data Mining for Systems Biology USA: Humana Press 2013; pp 97-113 Ding H, Takigawa I, Mamitsuka H, Zhu S Similarity-based machine learning methods for predicting drug–target interactions: a brief review Brief Bioinform 2014; 15: 734-47 Mousavian Z, Masoudi-Nejad A Drug–target interaction prediction via chemogenomic space: learning-based methods Exp Opin Drug Metabol Toxicol 2014; 10: 1273-87 Iskar M, Zeller G, Zhao XM, van Noort V, Bork P Drug discovery in the age of systems biology: the rise of computational approaches for data integration Curr Opin Biotechnol 2012; 23: 609-16 Masoudi-Nejad A, Mousavian Z, Bozorgmehr J Drug-target and disease networks: polypharmacology in the post-genomic era In Silico Pharmacol 2013; 1: 17 Leung EL, Cao Z-W, Jiang Z-H, Zhou H, Liu L Network-based drug discovery by integrating systems biology and computational technologies Brief Bioinform 2013; 14: 491-505 3574 Current Pharmaceutical Design, 2016, Vol 22, No 23 [28] [29] [30] [31] [32] [33] [34] [35] [36] [37] [38] [39] [40] [41] [42] [43] [44] [45] [46] [47] [48] [49] [50] [51] [52] [53] Berger SI, Iyengar R Network analyses in systems pharmacology Bioinformatics 2009; 25: 2466-72 Csermely P, Korcsmáros T, Kiss HJ, London G, Nussinov R Structure and dynamics of molecular networks: a novel paradigm of drug discovery: a comprehensive review Pharmacol Ther 2013; 138: 333-408 Ainsworth C Networking for new drugs Nat Med 2011; 17: 11668 Hopkins AL Network pharmacology: the next paradigm in drug discovery Nat Chem Biol 2008; 4: 682-90 Wu Z, Wang Y, Chen L Network-based drug repositioning Mol BioSyst 2013; 9: 1268-81 Chen X, Liu MX, Yan GY Drug-target interaction prediction by random walk on the heterogeneous network Mol BioSyst 2012; 8: 1970-8 Yildirim MA, Goh K-I, Cusick ME, Barabasi A-L, Vidal M Drugtarget network Nat Biotech 2007; 25: 1119-26 Yamanishi Y, Araki M, Gutteridge A, Honda W, Kanehisa M Prediction of drug–target interaction networks from the integration of chemical and genomic spaces Bioinformatics 2008; 24: i232-40 Bleakley K, Yamanishi Y Supervised prediction of drug–target interactions using bipartite local models Bioinformatics 2009; 25: 2397-403 Yamanishi Y, Kotera M, Kanehisa M, Goto S Drug-target interaction prediction from chemical, genomic and pharmacological data in an integrated framework Bioinformatics 2010; 26: i246-54 Li Y, Patra JC Genome-wide inferring gene-phenotype relationship by walking on the heterogeneous network Bioinformatics 2010; 26: 1219-24 Le D-H Disease phenotype similarity improves the prediction of novel disease-associated microRNAs 2015 2nd National Foundation for Science and Technology Development Conference on Information and Computer Science (NICS) 2015; pp 76-81 Seal A, Ahn Y-Y, Wild D Optimizing drug-target interaction prediction based on random walk on heterogeneous networks J Cheminform 2015; 7: 40 Cheng F, Liu C, Jiang J, et al Prediction of drug-target interactions and drug repositioning via network-based inference PLoS Comput Biol 2012; 8: e1002503 Mahé P, Ueda N, Akutsu T, Perret JL, Vert JP Graph kernels for molecular structureactivity relationship analysis with support vector machines J Chem Inf Model 2005; 45: 939-51 Saigo H, Vert J-P, Ueda N, Akutsu T Protein homology detection using string alignment kernels Bioinformatics 2004; 20: 1682-9 Faulon J-L, Misra M, Martin S, Sale K, Sapra R Genome scale enzyme–metabolite and drug–target interaction predictions using the signature molecular descriptor Bioinformatics 2008; 24: 22533 Jacob L, Vert J-P Protein-ligand interaction prediction: an improved chemogenomics approach Bioinformatics 2008; 24: 214956 Yabuuchi H, Niijima S, Takematsu H, et al Analysis of multiple compound–protein interactions reveals novel bioactive molecules Mol Syst Biol 2011; Campillos M, Kuhn M, Gavin A-C, Jensen LJ, Bork P Drug target identification using side-effect similarity Science 2008; 321: 2636 Takarabe M, Kotera M, Nishimura Y, Goto S, Yamanishi Y Drug target prediction using adverse event report systems: a pharmacogenomic approach Bioinformatics 2012; 28: i611-8 Mei JP, Kwoh CK, Yang P, Li XL, Zheng J Drug–target interaction prediction by learning from local information and neighbors Bioinformatics 2013; 29: 238-45 van Laarhoven T, Nabuurs SB, Marchiori E Gaussian interaction profile kernels for predicting drug–target interaction Bioinformatics 2011; 27: 3036-43 van Laarhoven T, Marchiori E Predicting drug-target interactions for new drug compounds using a weighted nearest neighbor profile PLoS One 2013; 8: e66952 Yang F, Xu J, Zeng J Drug-target interaction prediction by integrating chemical, genomic, functional and pharmacological data Pac Symp Biocomput 2014; 148-59 Gönen M Predicting drug–target interactions from chemical and genomic kernels using Bayesian matrix factorization Bioinformatics 2012; 28: 2304-10 Le and Le [54] [55] [56] [57] [58] [59] [60] [61] [62] [63] [64] [65] [66] [67] [68] [69] [70] [71] [72] [73] [74] [75] [76] [77] [78] [79] He Z, Zhang J, Shi XH, et al Predicting drug-target interaction networks based on functional groups and biological features PLoS One 2010; 5: e9603 Perlman L, Gottlieb A, Atias N, Ruppin E, Sharan R Combining drug and gene similarity measures for drug-target elucidation J Comput Biol 2011; 18: 133-45 Tabei Y, Pauwels E, Stoven V, Takemoto K, Yamanishi Y Identification of chemogenomic features from drug–target interaction networks using interpretable classifiers Bioinformatics 2012; 28: i487-94 Tabei Y, Yamanishi Y Scalable prediction of compound-protein interactions using minwise hashing BMC Syst Biol 2013; 7: S3 Cao D-S, Liu S, Xu QS, et al Large-scale prediction of drug–target interactions using protein sequences and drug topological structures Anal Chim Acta 2012; 752: 1-10 Wang Y-C, Yang Z-X, Wang Y, Deng N-Y Computationally probing drug-protein interactions via support vector machine Lett Drug Des Discov 2010; 7: 370-8 Yu H, Chen J, Xu X, et al A Systematic Prediction of multiple drug-target interactions from chemical, genomic, and pharmacological data PLoS One 2012; 7: e37608 Nanni L, Lumini A, Brahnam S A set of descriptors for identifying the protein–drug interaction in cellular networking Joretical Biol 2014; 359: 120-8 Xia Z, Wu L-Y, Zhou X, Wong S Semi-supervised drug-protein interaction prediction from heterogeneous biological spaces BMC Systems Biol 2010; 4: S6 Chen H, Zhang Z A Semi-supervised method for drug-target interaction prediction with consistency in networks PLoS One 2013; 8: e62975 Yu W, Yan Y, Liu Q, Wang J, Jiang Z Predicting drug–target interaction networks of human diseases based on multiple feature information Pharmacogenomics 2013; 14: 1701-7 Lamb J, Crawford ED, Peck D, et al The connectivity map: Using gene-expression signatures to connect small molecules, genes, and disease Science 2006; 313: 1929-35 Sun PG The human drug–disease–gene network Inf Sci 2015; 306: 70-80 Peng Gang S Co-controllability of drug-disease-gene network New J Phys 2015; 17: 085009 Huang H, Wu X, Pandey R, et al C2Maps: a network pharmacology database with comprehensive disease-gene-drug connectivity relationships BMC Genom 2012; 13: 1-14 Van Mulligen EM, Fourrier-Reglat A, Gurwitz D, et al The EUADR corpus: annotated drugs, diseases, targets, and their relationships J Biomed Inform 2012; 45: 879-84 Frijters R, van Vugt M, Smeets R, van Schaik R, de Vlieg J, Alkema W Literature mining for the discovery of hidden connections between drugs, genes and diseases PLoS Comput Biol 2010; 6: e1000943 Andronis C, Sharma A, Virvilis V, Deftereos S, Persidis A Literature mining, ontologies and information visualization for drug repurposing Brief Bioinform 2011; 12: 357-68 Piro RM, Di Cunto F Computational approaches to disease-gene prediction: rationale, classification and successes FEBS J 2012; 279: 678-96 Barabasi A-L, Gulbahce N, Loscalzo J Network medicine: a network-based approach to human disease Nat Rev Genet 2011; 12: 56-68 Oti M, Ballouz S, Wouters MA Web Tools for the Prioritization of Candidate Disease Genes Methods Mol Biol 2011; 760: 189-206 Wang X, Gulbahce N, Yu H Network-based methods for human disease gene prediction Brief Funct Genom 2011; 10: 280-93 Le D-H A novel method for identifying disease associated protein complexes based on functional similarity protein complex networks Algorithms Mol Biol 2015; 10: 14 Le D-H Network-based ranking methods for prediction of novel disease associated microRNAs Comput Biol Chem 2015; 58: 13948 Le DH, Nguyen MH Towards more realistic machine learning techniques for prediction of disease-associated genes USA: SOICT 2015 Proceeding of the Sixth International Symposium on Information and Comunication Technology 2015; pp 116-20 Le D-H, Xuan Hoai N, Kwon Y-K A Comparative Study of Classification-Based Machine Learning Methods for Novel Disease Gene Prediction In: Nguyen V-H, Le A-C, Huynh V-N, Eds., Systems Pharmacology: A Unified Framework for Prediction [80] [81] [82] [83] [84] [85] [86] [87] [88] [89] [90] [91] [92] [93] [94] Knowledge and Systems Engineering USA: Springer International Publishing 2015; pp 577-88 Le D-H, Kwon Y-K Neighbor-favoring weight reinforcement to improve random walk-based disease gene prioritization Comput Biol Chem 2013; 44: 1-8 Sanseau P, Koehler J Editorial: Computational methods for drug repurposing Brief Bioinform 2011; 12: 301-2 Dudley JT, Deshpande T, Butte AJ Exploiting drug–disease relationships for computational drug repositioning Brief Bioinform 2011; 12: 303-11 Pahikkala T, Airola A, Pietilä S, et al Toward more realistic drug– target interaction predictions Brief Bioinform 2015; 16: 325-37 Wang Y-C, Zhang C-H, Deng N-Y, Wang Y Kernel-based data fusion improves the drug–protein interaction prediction Comput Biol Chem 2011; 35: 353-62 De Bie T, Tranchevent L-C, Van Oeffelen LMM, Moreau Y Kernel-based data fusion for gene prioritization Bioinformatics 2007; 23: i125-32 Yu S, Tranchevent L-C, De Moor B, Moreau Y Gene prioritization and clustering by multi-view text mining BMC Bioinform 2010; 11: 28 Yu S, Falck T, Daemen A, et al L2-norm multiple kernel learning and its application to biomedical data fusion BMC Bioinform 2010; 11: 309 Gönen M, Alpaydin E Multiple kernel learning algorithms J Mach Learn Res 2011; 999999: 2211-68 Kloft M, Brefeld U, Sonnenburg S, Zien A Lp-norm multiple kernel learning J Mach Learn Res 2011; 999999: 953-97 Yang P, Li X-L, Mei J-P, Kwoh C-K, Ng S-K Positive-unlabeled learning for disease gene identification Bioinformatics 2012; 28: 2640-7 Mordelet F, Vert J-P ProDiGe: Prioritization Of Disease Genes with multitask machine learning from positive and unlabeled examples BMC Bioinformatics 2011; 12: 389 Yang P, Li X, Chua H-N, Kwoh C-K, Ng S-K Ensemble positive unlabeled learning for disease gene identification PLoS One 2014; 9: e97079 Ma X, Chen T, Sun F Integrative approaches for predicting protein function and prioritizing genes for complex phenotypes using protein interaction networks Brief Bioinform 2013: bbt041 Fouss F, Francoisse K, Yen L, Pirotte A, Saerens M An experimental investigation of kernels on graphs for collaborative recom- Current Pharmaceutical Design, 2016, Vol 22, No 23 [95] [96] [97] [98] [99] [100] [101] [102] [103] [104] [105] [106] [107] [108] 3575 mendation and semisupervised classification Neural Netw 2012; 31: 53-72 Farkas IJ, Korcsmáros T, Kovács IA, et al Network-based tools for the identification of novel drug targets Sci Signal 2011; 4: pt3-pt3 Huan T, Wu X, Chen JY Systems biology visualization tools for drug target discovery Exp Opin Drug Discov 2010; 5: 425-39 Yamanishi Y, Kotera M, Moriya Y, Sawada R, Kanehisa M, Goto S DINIES: drug–target interaction network inference engine based on supervised analysis Nucleic Acids Res 2014; 42: W39-45 Xiao X, Min J-L, Wang P, Chou K-C iGPCR-drug: a web server for predicting interaction between GPCRs and drugs in cellular networking PLoS One 2013; 8: e72234 Le D-H, Kwon Y-K GPEC: A Cytoscape plug-in for random walkbased gene prioritization and biomedical evidence collection Comput Biol Chem 2012; 37: 17-23 Tranchevent L-C, Capdevila FB, Nitsch D, et al A guide to web tools to prioritize candidate genes Brief Bioinform 2010; 12: 2232 Luo H, Chen J, Shi L, et al DRAR-CPI: a server for identifying drug repositioning potential and adverse drug reactions via the chemical–protein interactome Nucleic Acids Res 2011; 39: W4928 von Eichborn J, Murgueitio MS, Dunkel M, Koerner S, Bourne PE, Preissner R PROMISCUOUS: a database for network-based drugrepositioning Nucleic Acids Res 2011; 39: D1060-6 Alaimo S, Bonnici V, Cancemi D, Ferro A, Giugno R, Pulvirenti A DT-Web: a web-based application for drug-target interaction and drug combination prediction through domain-tuned networkbased inference BMC Syst Biol 2015; 9: S4 Sanseau P, Agarwal P, Barnes MR, et al Use of genome-wide association studies for drug repositioning Nat Biotech 2012; 30: 317-20 Cao C, Moult J GWAS and drug targets BMC Genom 2014; 15: S5 Li Y, Jones S Drug repositioning for personalized medicine Genom Med 2012; 4: 1-14 Reynolds KS Achieving the promise of personalized medicine Clin Pharmacol Ther 2012; 92: 401-5 Fernald GH, Capriotti E, Daneshjou R, Karczewski KJ, Altman RB Bioinformatics challenges for personalized medicine Bioinformatics 2011; 27: 1741-8 The author has requested enhancement of the downloaded file All in-text references underlined in blue are linked to publications on ResearchGate ... supervised approaches are feature vector-based, in which, drug- target pairs were represented as feature vectors Based on features of drugs and targets extracted from drug and target- related data as well... Pharmaceutical Design, 2016, Vol 22, No 23 Le and Le Fig (1) A unified framework for prediction of drug- target interactions networks of biomedical and pharmacological instances (e.g., drugs, targets... main approaches is data representation In general, the computational approaches represent data in two ways: i) drugs and targets are considered as separate instances (e.g., given a known drug and