Prediction of drug-target interactions and drug repositioning via network-based inference.

Cheng F, Liu C, Jiang J, Lu W, Li W, Liu G, Zhou W, Huang J, Tang Y. · 2012

Read Prediction of drug-target interactions and drug repositioning via network-based inference. by Cheng F, Liu C, Jiang J, Lu W, Li W, Liu G, Zhou W, Huang J, Tang Y. (2012) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/europmc-pmc3349722.txt.

Introduction

Cheng and co-authors developed a computational method that predicts new drug targets by reasoning over networks. Their network-based inference approach exploits the web of known drug, target, and disease relationships, ranking candidate interactions that have not yet been tested in the laboratory.

The method performed well at recovering known drug–target pairs and, more strikingly, pointed to new uses for existing medicines. Five older drugs flagged by the algorithm were checked experimentally and confirmed in vitro to act on predicted targets. The authors propose the approach as a route to drug repositioning: finding fresh indications for approved compounds whose safety profiles are already understood.

Catalog record

CREATORCheng F, Liu C, Jiang J, Lu W, Li W, Liu G, Zhou W, Huang J, Tang Y.
YEAR2012
COLLECTIONPAPERS
FORMATTXT
FILE SIZE54 KB
LICENSECC BY 4.0
RIGHTSCC BY
SOURCEEurope PMC

Source: https://europepmc.org/articles/PMC3349722

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Prediction of Drug-Target Interactions and Drug Repositioning via Network-Based Inference By Feixiong Cheng, Chuang Liu, Jing Jiang, Weiqiang Lu, Weihua Li, Guixia Liu, Weixing Zhou, Jin Huang, Yun Tang, Russ B Altman Published in: PLoS Computational Biology DOI: 10.1371/journal.pcbi.1002503 ABSTRACT Abstract Drug-target interaction (DTI) is the basis of drug discovery and design. It is time consuming and costly to determine DTI experimentally. Hence, it is necessary to develop computational methods for the prediction of potential DTI. Based on complex network theory, three supervised inference methods were developed here to predict DTI and used for drug repositioning, namely drug-based similarity inference (DBSI), target-based similarity inference (TBSI) and network-based inference (NBI). Among them, NBI performed best on four benchmark data sets. Then a drug-target network was created with NBI based on 12,483 FDA-approved and experimental drug-target binary links, and some new DTIs were further predicted. In vitro assays confirmed that five old drugs, namely montelukast, diclofenac, simvastatin, ketoconazole, and itraconazole, showed polypharmacological features on estrogen receptors or dipeptidyl peptidase-IV with half maximal inhibitory or effective concentration ranged from 0.2 to 10 µM. Moreover, simvastatin and ketoconazole showed potent antiproliferative activities on human MDA-MB-231 breast cancer cell line in MTT assays. The results indicated that these methods could be powerful tools in prediction of DTIs and drug repositioning. INTRODUCTION Over the past decade, the rate of new chemical entities transferred to therapeutic agents has been significantly decreased [1]. Interestingly, this phenomenon is concurrent with the dominant assumption that the goal of drug discovery is to design exquisitely selective ligands against a single target. However, this ‘one gene, one drug, one disease’ paradigm was challenged in many cases, and the concept of polypharmacology was hence proposed for those drugs acting on multiple targets rather than one target [1]. For example, serotonin and serotonergic drugs not only bind to G protein-coupled receptors (GPCRs) such as 5-hydroxytryptamine receptors 1, 2 and 4–7 (5-HT1,2,4–7), but also might bind to an ion channel, i.e. 5-HT3 [2], [3]. Such polypharmacological features of drugs enable us to understand drug side effects or find their new uses, namely drug repositioning [4]. Some good examples are thalidomide, sildenafil, bupropion and fluoxetine [4],[5]. To date, several in silico methods have been developed to address the issues of drug-target interaction (DTI) prediction and drug repositioning [6]–[11]. The conventional methods can be either ligand-based or receptor-based. Ligand-based methods like quantitative structure-activity relationships (QSAR) and similarity search are very useful in this context. For example, Keiser et al. predicted new molecular targets for known drugs using chemical two-dimensional (2D) structural similarity, namely similarity ensemble approach [6], [7]. Twenty-three new DTIs were confirmed and five of which were potent with Ki values<100 nM. Recently, Humberto et al. developed a multi-target QSAR (mt-QSAR) classifier and built a web server for DTI prediction [8]. Receptor-based methods like reverse docking have also been applied in drug-target (DT) binding affinity prediction, DTI prediction and drug repositioning [9]–[11]. However, those methods could not be used for targets whose three-dimensional (3D) structures are unknown. More recently, several network-based and phenotype-based methods were developed for such purposes. Yildirim et al. constructed a bipartite graph composed of US Food and Drug Administration (FDA)-approved drugs and proteins linked by DT binary associations [12]. This method quantitatively showed an overabundance of ‘follow-on’ drugs. Campillos et al. identified new DTIs using side-effect similarity [13]. They tested 20 of unexpected DTIs and validated 13 ones by in vitro binding assays. Iorio et al. predicted and validated new drug modes of action and drug repositioning from transcriptional responses [14]. Recently Butte group also reported two successful examples of drug repositioning based on public gene expression data [15], [16]. Furthermore, Yamanishi et al. developed a bipartite graph learning method to predict DTI by integrating chemical and genomic spaces [17]. Though high overall predictive accuracy was obtained in Yamanishi's work, the sensitivity was anomaly low and the method was not validated experimentally. In this study, three inference methods were developed to predict new DTI: drug-based similarity inference (DBSI), target-based similarity inference (TBSI) and network-based inference (NBI), all derived from complex network theory [18]–[21]. Four benchmark data sets with known drugs targeting enzymes, ion channels, GPCRs, and nuclear receptors respectively, were used to assess the performance of the methods in comparison with literature reports. The best-performed method was then selected to create a drug-target network of FDA-approved and experimental drugs and to predict new DTIs subsequently. Some of the predictions were further validated by in vitro assays. This work would provide new powerful tools for DTI prediction and drug repositioning. RESULTS The methods developed here were derived from the recommendation algorithms of complex network theory, and proposed for DTI prediction and hence drug repositioning. In principle, the DBSI method (Figure 1A) is very similar to the item-based collaborative filtering method in recommendation algorithms [20], while TBSI (Figure 1B) is similar to the user-based collaborative filtering method [21]. Different from DBSI and TBSI, the NBI method (Figure 1C) only uses known DT bipartite network topology similarity to predict unknown DTI, which employs a process analogous to mass diffusion in physics across the DT network [18], [19]. In NBI method, predictive scores are calculated for each given drug (pink circle) and each unlinked target, and a recommendation list of drugs was created for a given target (pink square) in a descending order after the diffusion process. [Figure: Schematic diagram of (A) drug-based similarity inference (DBSI), (B) target-based similarity inference (TBSI) and (C) network-based inference (NBI) methods.The entire workflow includes five steps: (i) collection of known drug-target interaction data and construction of bipartite drug-target graphs; (ii) calculation of drug-drug two dimensional structural similarity (SC), target-target genomic sequence similarity (Sg) and drug-target topology network similarity; (iii) application of new methods in prediction of new drugs for a given target (pink square) or new targets for a given drug (pink circle); (iv) validation of new drug-target interactions by experimental assays (D); (v) visualization of experimental results using drug-target-disease associations network analysis (E). In A–C, given drug node (pink circle) denotes the drug which we want to predict new target for, given target node (pink square) denotes the target which we want to predict new drug for, drug with resource (green circle) denotes that this drug have resource, target with resource (green square) denotes that this target have resource, the more resource a node possesses, the darker the color is, blue edges denote the drug-target interactions with known experimental evidence, black arrows denote the resource diffusion direction. In E, green circle: drug node, red square: on-target node, blue square: off-target node, yellow square: new off-target node, violet square: disease node.] Performance of the methods on benchmark data sets Four benchmark data sets were used to assess the performance of the methods. The data sets were named after four major drug targets, i.e. enzymes, ion channels, GPCRs, and nuclear receptors. At first, all known DTIs (Table S1) involved in the data sets were used to generate a DT bipartite network (Figure S1), in which a drug (circle) and a target (square) were connected if the target was known to the drug according to experimental evidence. Figure 2 illustrated the receiver operating characteristic (ROC) curves calculated by the methods on the

Read the complete text