Large-Scale Predictive Drug Safety: From Structural Alerts to

Perspective

Large Scale Predictive Drug Safety: From Structural Alerts to Biological Mechanisms Ricard Garcia-Serna, David Vidal, Nikita Remez, and Jordi Mestres Chem. Res. Toxicol., Just Accepted Manuscript • DOI: 10.1021/acs.chemrestox.5b00260 • Publication Date (Web): 11 Sep 2015 Downloaded from http://pubs.acs.org on September 14, 2015

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a free service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are accessible to all readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

Chemical Research in Toxicology is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 33

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 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 54 55 56 57 58 59 60

Chemical Research in Toxicology

Large Scale Predictive Drug Safety: From Structural Alerts to Biological Mechanisms †

†

Ricard Garcia-Serna, David Vidal, Nikita Remez,

†,‡

∗,†,‡

and Jordi Mestres

†

Chemotargets SL, Parc Científic de Barcelona, Baldiri Reixac 4 (TI-05A7), 08028 Barcelona, Catalonia, Spain ‡ Systems Pharmacology, Research Program on Biomedical Informatics (GRIB), IMIM Hospital del Mar Medical Research Institute and University Pompeu Fabra, Parc de Recerca Biomèdica, Doctor Aiguader 88, 08003 Barcelona, Catalonia, Spain

ABSTRACT: The recent explosion of data linking drugs, proteins, and pathways with safety events has promoted the development of integrative systems approaches to large-scale predictive drug safety. The added value of such approaches is that, beyond the traditional identification of potentially labile chemical fragments for selected toxicity endpoints, they have the potential to provide mechanistic insights for a much larger and diverse number of safety events in a statistically-sound non-supervised manner, based on the similarity to drug classes, the interaction with secondary targets and the interference with biological pathways. The combined identification of chemical and biological hazards enhances our ability to assess the safety risk of bioactive small molecules with higher confidence than using structural alerts only. We are still a very long way from reliably predicting drug safety but advances towards gaining a better understanding of the mechanisms leading to adverse outcomes represent a step forward in this direction.

∗

To whom correspondence should be addressed: [email protected]

ACS Paragon Plus Environment

1


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 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 54 55 56 57 58 59 60

Page 2 of 33

CONTENTS Introduction Sources of Data Links Safety-Drug Links Drug-Target Links Target-Pathway Links Target-Safety Links From Structural Alerts to Biological Mechanisms Hazard Detection Risk Assessment Conclusions and Outlook INTRODUCTION Drug safety has been recognized as one of the main causes of attrition during clinical trials, ranking in third place for drugs in phase II1,2 and second for phase III and submission stages.1,3 In addition, the complex nature of most drug-mediated adverse events complicates their detection at clinical phases and, in many instances, they can only be perceived clearly once the drug is in the market and thus widely available to the population. Current estimates suggest that in the United States alone the annual number of patients affected by drug adverse events is around 2 million, of which approximately 5% are ultimately fatal.4 These numbers make drug safety one of the top five causes of death in the United States, with similar figures observed in other developed countries. In economic terms, it has been estimated that a pharmaceutical company affected by a drug withdrawal is expected to loose around $1 billion.4 Also, the costs for the healthcare system associated with drug safety treatments are estimated to be around $40 to $50 billion per year worldwide. If one takes into consideration that 19 drugs were withdrawn from the US market between 1998 and 2007, the overall societal costs of being exposed to unsafe drugs are enormous.5 In view of the situation, both pre-clinical anticipation and post-marketing surveillance and detection (pharmacovigilance) of drug-induced safety issues have become priority areas of research in recent years. Consequently, national and supra-national authorities worldwide have promoted and funded multi-disciplinary consortia to advance in our understanding and predictive ability of drug safety issues.6 In the European Union, the EU-ADR project7 was awarded by the Information and Communication Technologies (ICT) Area of the European Commission aiming at the early detection of adverse drug reactions through the exploitation of data contained in


2

Page 3 of 33

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 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 54 55 56 57 58 59 60


medical records of over 30 million patients from several European countries. The project started in 2008 and it was completed in 2011. A web platform is still available where users can have access to epidemiological, textual, in vitro, and in silico data on a selected number of adverse drug reactions.8 Subsequently, the eTOX project9,10 was funded by the Innovative Medicines Initiative Joint Undertaking (IMI-JU) with the objective of constructing a preclinical database and developing software solutions for the analysis of toxicological data contained in the internal legacy reports of pharmaceutical industry and their exploitation to predict major toxicity endpoints.11 Initiated in 2010, the eTOX project will run until the end of 2016. In the United States, the ToxCast12 project started in 2007 funded by the US Environmental Protection Agency (EPA) with the aim of reducing the need for animal-based toxicity tests in the drug discovery process by exposing living cells or isolated proteins to chemicals and detecting the alterations these may cause. Later on, in 2008, the Toxicity Testing in the 21st Century (Tox21) inter-agency collaboration was established with the focus on setting up a robotic system capable of screening a collection of 10,000 environmental chemicals in triplicate in a week.13 Until now, these initiatives have evaluated thousands of pharmaceutical and environmental chemicals across more than 700 high-throughput assays that cover a wide range of high-level cell responses and approximately 300 signaling pathways.14,15 In Japan, the Toxicogenomics Project (TGP)16 was initiated in 2002 by the National Institute of Biomedical Innovation, from the National Institute of Health Sciences, with 15 pharmaceutical companies involved. This project spent its first four years collecting data on observed gene expression and toxicity after exposing living rats and primary cultured hepatocytes of rats and humans to 150 compounds. All these data was used to build a large-scale high-quality toxicogenomics database called Toxicogenomics Project-Genomics Assisted Toxicity Evaluation system (TG-GATEs), which integrates also several analysis and prediction tools.17 In spite of all the worldwide efforts put on those highly ambitious projects, a global generic solution to large scale predictive drug safety remains elusive. Computational toxicology has been traditionally focused on identifying chemical fragments significantly present in drug structures linked to safety events.18,19 However, good progress has been made recently in generating all types of safety relevant data, storing them properly, and making them accessible in public resources, which has motivated and highly facilitated their integration and, consequently, the


3


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 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 54 55 56 57 58 59 60

Page 4 of 33

development of novel systems approaches to drug safety. In particular, advances in predicting the off-target pharmacology of bioactive small molecules, and assessing the degree of perturbation of biological pathways, offer now the possibility to complement traditional structural alerts with insights on the plausible biological mechanism leading to adverse outcomes. In the next sections, a summary of the main data sources will be provided, followed by a brief overview of some of the most recent attempts to data integration, hazard detection, and risk assessment.

SOURCES OF DATA LINKS As it happened almost a decade ago with the explosion of public access to pharmacological data,20 the amount of data available on drug safety has increased dramatically in recent years.21 In parallel, coordinated international efforts, such as the Open PHACTS project,22 are ongoing with the aim to assemble and integrate different types of data from a variety of sources and make them available in an appropriate format to the scientific community. The potential to link chemical structures (drugs) with safety outcomes, protein targets, and biological pathways23,24 opens an avenue to build a next generation of predictive drug safety systems that can offer the possibility to perform signal substantiations and derive mechanistic hypotheses beyond traditional structure alerts.25-29 The following sections summarise the main individual sources from which pairwise links between those four pillar entities are available (Figure 1). Safety-Drug Links. The association between the use of a given drug and the observation of an adverse reaction in a patient is the essential piece of information to connect chemical structures and safety events. Traditionally, these data have been manually and individually collected through post-marketing pharmacovigilance adverse event reporting systems30 but more recently they can be identified by means of large scale mining of bibliographic sources,31 web search log data,32 and even social networks content.33 Unfortunately, drug-safety links are still scattered in different national databases that contain information stored in a diversity of languages and do not make use of a standard unified ontology of safety terms.34,35 As a consequence, most of the efforts in assembling current drug-safety repositories have to be devoted to term curation, translation, and integration (synonymia). The need to closely monitor adverse drug events occurring in the population as a result of exposure to drugs has lead to the creation of information systems where clinicians,


4

Page 5 of 33

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 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 54 55 56 57 58 59 60


pharmaceutical companies, and even patients themselves, can deposit safety-relevant data. The largest of these resources, gathering data since 1969 and currently receiving almost one million reports per year, is the Adverse Events Reporting System (AERS) supported by the WHO, FDA and Health Canada organizations.36 However, detecting significant signals directly from this type of data sources may be problematic due to the existence of many confounding factors that interfere with statistical analyses. It has been shown that some of these factors can be corrected, or at least partly alleviated, by applying adaptive data-driven approaches such as the one called statistical correction of uncharacterized bias (SCRUB).37 As a result, a new database of drug effects was created (Offsides) containing 438,801 significant annotations between 1,332 drugs and 10,097 side effects. Another important source of drug safety information is the use of clinical patient records, which in 1989 were the starting point of a project carried out by Erasmus MC in Rotterdam called Interdisciplinary Processing of Clinical Information (IPCI).38 Designed originally to be a longitudinal observational database with data from computer-based patient records of a selected group of general practitioners who voluntarily agreed to supply data to the database, it currently contains information on more than 1.5 million patients. This information includes demographics, medical notes, prescriptions or indications for therapy, referrals, hospitalizations or laboratory results and can be accessed through a strict protocol in order to carry out both academic and commercial studies. Safety information coming from drug labels, package inserts, and other public documents is also worth considering. The Side Effect Resource (SIDER), hosted by the European Molecular Biology Laboratory, is a web-based platform that contains information on marketed drugs and their recorded adverse drug reactions,39 standardized using the MedDRA dictionary.40 Currently containing around 100,000 annotations between almost 1,000 drugs and more than 4,000 side effects, this resource includes also the structures of drugs as well as data related to the observed frequency of the different side effects. The World Drug Index41 and Pharmapendium42 databases are alternative commercial solutions offering highly curated safety data extracted also from scientific literature, as well as from clinical and preclinical sources. Finally, preclinical data from in vivo animal toxicity studies represent also an important surrogate source of human safety information. ToxRefDB,43 hosted by the US Environmental Protection Agency, is a web-based public repository that collects this kind of information in a


5


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 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 54 55 56 57 58 59 60

Page 6 of 33

searchable format. In addition, the DrugMatrix project,44 funded by the US Department of Health, is a vast resource populated with the results of toxicological experiments in which rats or primary rat hepatocytes were systematically treated with chemicals to study their effects in multiple target organs at many levels. Collecting, integrating and exploiting this type of information across pharmaceutical companies is an ongoing effort within the eTOX IMI-JU European project mentioned above.9-11 Drug-Target Links. It is now widely recognized that most drugs interact with multiple proteins,45-47 a property often referred to as polypharmacology.48-50 Some of those drug-protein interactions are essential for exerting the therapeutic effect but other may well be responsible for some of the observed adverse drug reactions.51-56 In fact, it has been estimated that around 80% of drug safety events are mediated by some degree of off-target pharmacology. The rest would be deeply rooted into the inherent genetic polymorphisms of human populations,57 the diversity of immune system responses,58 and drug metabolism,59 all aspects being difficult to anticipate due to its idiosyncratic nature.60,61 Therefore, if one could anticipate only a portion of the cases due to off-target pharmacology it would be already a significant achievement. Accordingly, collecting binding affinity data has become of utmost interest for predictive safety. In recent years, the amount of pharmacology data available in public repositories has increased dramatically. The largest of those repositories is PubChem BioAssays62 hosted by the National Center for Biotechnology Information (NCBI) that contains 200,000,000 bioactivity outcome summaries representing biological properties (including binding affinities) for 1,900,000 chemical structures. Another reference source for this type of data is ChEMBL,63 a database of bioactivity data for drug-like molecules mainly funded by the European Bioinformatics Institute (EBI) that contains almost 12.5 million data points for more than 1.3 million unique compounds.64 Other databases worth mentioning are BindingDB,65,66 currently containing over 1,100,000 binding data for almost 0.5 million small molecules, DrugBank,67 containing almost 8,000 drug entries, IUPHARdb,68 the official database of the International Union of Basic and Clinical Pharmacology, and the Comparative Toxicogenomics Database (CTD),69 gathering data about environmental chemicals and their relation with proteins, genes, and diseases. Last but not least, the NIH’s Psychoactive Drug Screening Program70 is a unique resource that makes available its own generated experimental data on the binding affinity of small molecules to cloned human or rodent CNS receptors, channels, and transporters.


6

Page 7 of 33

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 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 54 55 56 57 58 59 60


Beyond these resources, the recent capacity to generate genome-wide association study (GWAS) data for drugs has emerged as an invaluable approach to identifying disease genes linked to genetic markers of the disease or phenotype under investigation, extending current knowledge on their mechanism of action, and thus ultimately contributing to therapeutic repositioning and target safety assessment.71,72 Target-Pathway Links. Proteins are organized and interconnected in biological pathways that, when perturbed, may lead to adverse health outcomes.73 In order to explore possible causal associations between safety events and biological pathways through the previously described safety-drug and drug-target links, one requires a final extension linking targets with pathways. Several highly-curated well-maintained resources have been created over the years with information on target-pathway links. Among them, Reactome,74 funded by the Ontario Institute for Cancer Research (OICR) and the European Bioinformatics Institute (EBI), is a peer-reviewed open-access pathway database that aims at bringing online detailed information and schemas present in textbooks and articles in relation with human pathways. The latest version describes over 7,000 human proteins participating in almost 7,000 reactions extracted from over 15,000 research publications. Another public-access web resource is Pathway Commons75 that collects data on over 1,400 biological pathways from multiple organisms. Other public sources of targetpathway links include WikiPathways,76 a community effort that has developed a public wiki for pathway curation and ontology annotations, PharmGKB,77 centered around drugs and their interaction with proteins with known genotypic variations, the Therapeutic Targets Database,78 containing data on therapeutic proteins and their links to drugs, pathways, and diseases, and the Adverse Outcome Pathway Knowledge Base, from the Organisation for Economic Co-operation and Development (OECD).79 Finally, KEGG80 is a widely used multifocus resource with information on many key elements for systems biology, from metabolites to organisms, that changed from public to subscribed access in 2011. Target-Safety Links. Ideally, one would like to be able to infer potential safety concerns directly from the in vitro affinity between a drug and its associated target.81 Indeed, some adverse drug reactions are directly linked to the affinity of the drug for its primary therapeutic target (on-target effects). For example, adrenergic β-blockers, widely used in the treatment of cardiovascular diseases, are well recognized for the potential to cause bradycardia and hypotension82 (excessive on-target effects) and for inducing bronchoconstriction in asthmatic


7


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 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 54 55 56 57 58 59 60

Page 8 of 33

subjects83 (on-target effects in secondary organ systems). Unfortunately, most drugs have lowaffinity interactions for multiple proteins beyond their primary target (off-target effects),45-50 and this polypharmacology makes hypotheses on target-safety links extremely difficult and potentially misleading. There are, however, some well-known exceptions. One of the best established cases is the hERG potassium channel, the affinity to which has been strongly linked to QT prolongation that can result in a potentially fatal type of arrhythmia, called Torsades de Pointes (TdP), one of the main causes of drug withdrawals in recent years.84 Another off-target example is the serotonin 5HT2B receptor, the activation of which has been associated to druginduced valvular heart disease.85 Beyond these specific cases, a rather long list of proteins potentially implicated in major adverse events has been assembled and constitutes the basis for current in vitro safety pharmacology campaigns.51-53 However, again, one ought to take into consideration that many of those associations may be confounded by the drug’s pharmacokinetic parameters,86 tissue distribution, and affinity to other proteins, and that all those one-to-one target-safety links should be regarded with caution, particularly since many of those proteins are aminergic G protein-coupled receptors, a protein family well recognized for its promiscuous pharmacology compared to other families.46,87

FROM STRUCTURAL ALERTS TO BIOLOGICAL MECHANISMS The exact mechanism of action by which the majority of safety events occur upon drug administration is still, in many instances, a big unknown. In this respect, if there is one mechanistic aspect in drug safety that has been well explored in recent years is the association of drug toxicity with metabolic transformations leading to the formation of reactive species, systemic accumulation of metabolites, or induction of metabolic pathways.88,89 A wide range of both free and commercial software for predicting drug metabolism is available.90 Unfortunately, experts often argue that many of those methods are not very good at prediction and remain unproven in their usefulness. In this respect, reducing the false positive rate remains a challenge for future developments. In contrast, what is perfectly known is the chemical structure of the molecule which in fact is, directly or indirectly, ultimately responsible for the observed adverse events. Therefore, it is not surprising that most of the efforts to date in computational toxicology have focused mainly on providing structural alerts of particular fragments or chemical patterns, also referred to as


8

Page 9 of 33

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 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 54 55 56 57 58 59 60


toxicophores,91-96 commonly present in drugs linked to particular safety outcomes.19 This has resulted in a variety of freely available solutions such as the QSAR Toolbox, from the OECD,97,98 ToxTree, a decision tree approach commissioned by the EU Joint Research Centre,98,99 OpenTox, an open source predictive toxicology framework developed within the EU FP7 program,100 and ToxAlerts, available from the Online Chemical Modeling Environment (OCHEM) platform.101,102 There are also various commercial solutions available such as TOPKAT,103-106 LeadScope Model Applier,107 and HazardExpertPro.104,108 Among those, two software packages are widely used within pharmaceutical industry for regulatory purposes to assess the risk of compounds to adverse drug reactions, environmental toxicity or genotoxicity. One of them is Derek Nexus,104-106,109 an expert system that uses information stored in a certified knowledge base, composed of both public and private sources and reviewed and verified by experts, to detect rules and relationships between chemical features and different key toxicity endpoints such as carcinogenicity, mutagenicity, genotoxicity and skin sensitisation. The other is MultiCASE,105,110 a knowledge-based system using a fragment-based QSAR methodology to automatically identify the molecular substructures that have a high probability of being responsible for the toxicity effects observed in a set of compounds. However, in spite of the widespread use and demonstrated usefulness of those packages in many cases, all approaches mentioned above still rely solely on the structure of molecules. In the last few years, the need to go beyond mere structural alerts to advance in the field of predictive drug safety has become evident. As exposed above, given the current availability of such large and increasingly diverse amounts of systemic data, a next generation of tools addressing drug safety from a systems perspective is required. In this respect, several efforts have been recently made to perform statistical analyses of chemical and biological data for large scale hazard detection of chemical fragments, protein targets, and biological pathways linked to safety events, but also to develop novel computational approaches that exploit those detected hazards for large scale risk assessment of safety events. The following sections highlight some of the most recent advances in both directions. Hazard Detection. Access to big data on links connecting pairs of systemic entities allows for detecting statistically significant chemical and biological hazards to drug safety. In a pioneering work, Campillos et al.111 showed that two drugs having similar side-effect profiles have also biologically-relevant affinities for the same protein. The concept was later developed


9


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 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 54 55 56 57 58 59 60

Page 10 of 33

further by Garcia-Serna & Mestres54 to show that drugs with similar safety profiles had also similar polypharmacology. Later, Lounkine et al.56 applied a similarity-based method to predict novel unknown targets for old known drugs and detect correlations between those new targets and the side effects linked to some of the old drugs. As an example, they established a link between epigastralgia and COX-1 inhibition, supported by evidence connecting COX-1 and upper gastrointestinal bleeding.112 Along the same lines, Kuhn et al.81 performed a statistical analysis of toxicity data obtained during clinical trials linked to known targets of drugs to identify overrepresented targets among drugs associated with a specific side effect. In their work, they were able to show that the majority of the analyzed side effects were significantly related to one or more target proteins and, often, these links could be supported by bibliographic references or knock-out mice models. As a proof of concept, the authors showed that it was possible to reduce the severity of hyperesthesia in mice by the administration of a second drug that was counteracting the effect of the main one over the serotonin 5-HT7 receptor, an off-target believed to be responsible for the appearance of that adverse reaction. One important aspect in some of the works mentioned above is the impact of predicted offtarget pharmacology in detecting potential off-target hazards.113,114 Given the relatively low levels of completeness,45,46 it is envisaged that any attempt to complete our current knowledge of drug-target interactions with computational predictions will have an effect on our ability to identify those proteins that appear to be linked to safety events. To illustrate the effect here, we took a set of 1,271 drugs annotated to a list of 333 terms associated with cardiotoxicity (including long QT syndrome, TdP, tachycardia, bradycardia, cardiac arrhythmia, palpitations, etc.) and compared their experimentally known in vitro target profiles (Figure 2a, left) with the corresponding completed profiles that result after adding in silico predicted off-target affinities (Figure 2a, right).115,116 The differences between the list of highly enriched targets linked to cardiotoxicity as extracted from the experimentally known (Figure 2b, left) and the predictionenriched (Figure 2b, right) target profiles become evident. For the sake of simplicity, enrichment is assessed here as the difference between the percentage of drugs being active on a certain target and associated with a given safety event, and those that, while being active on that same target, are not linked (yet) to that safety event. As an example, 79% of the drugs active on the Potassium voltage-gated channel subfamily H member 2 (hERG) were found to be linked also to some cardiotoxicity term, whereas the average percentage of drugs active on hERG but associated with


10

Page 11 of 33

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 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 54 55 56 57 58 59 60


other toxicity categories is 52%. Therefore, taking solely experimentally known drug-target interactions, this is translated into an enrichment factor for hERG of 0.27 (Figure 2b, left; second target from top), reflecting the relevance of this particular off-target in cardiotoxicity, relative to other safety categories. A more systematic comparative analysis of the respective lists of targets found to be enriched in drugs associated with cardiotoxicity reveals that only five of the top twelve targets (or target sets) shown are actually present in both lists (Figure 2b). This means that inclusion of predicted affinities does not affect the detection and relevance of those five protein hazards for cardiotoxicity. These are the voltage-dependent L-type calcium channels, hERG, β-adrenergic receptors, sodium channels and the muscarinic acetylcholine receptors. In all five cases, the link between those targets and different cardiotoxicity events could be established: blocking of L-type calcium channels and β-adrenergic receptors is associated with bradycardia and hypotension,82 hERG blockade is involved in a potentially fatal type of arrhythmia leading to long QT syndrome and TdP,84 blocking of the cardiac sodium channels is thought to be the cause of the cardiovascular toxicity of tricyclic antidepressants,117 and inhibition of the cardiac muscarinic M2 receptor has been linked to tachycardia.118 In contrast, some targets that were found highly relevant when using only known interactions (Figure 2b, left) are replaced by other targets that gain relevance when predicted interactions are considered (Figure 2b, right). For example, among the most relevant targets associated with cardiotoxicity when only known interactions are taken into account we found calmodulin, peroxisome proliferator-activated receptor gamma (PPARγ), and cholinesterase. Indeed, it has been reported that mutations in calmodulin can lead to syncope and sudden cardiac death,119 PPARγ ligands, such as rosiglitazone and pioglitazone, have been related to increased risk of heart failure,120 and the use of cholinesterase inhibitors has been reported to be related with a small but significant increase in the risk of adverse cardiac events.121 Even though they still retain positive enrichment factors, their relevance in cardiotoxicity decreases when in silico predictions are included and other targets emerge as more hazardous for cardiac safety, such as DNA topoisomerase 2α, high-affinity choline transporter 1, nischarin (imidazoline receptors), inducible nitric oxide synthase (iNOS), and HIV-1 protease. In this respect, it was confirmed that DNA topoisomerase 2α is indeed associated with acute cardiotoxic effects and show a doserelated cardiomyopathy,122 diminished expression of presynaptic choline transporter is related to


11


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 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 54 55 56 57 58 59 60

Page 12 of 33

tachycardia and ventricular dysfunction,123 imidazoline I1 receptor agonists can cause hypotension and bradycardia,124 iNOS plays a pivotal role in ischemic tolerance and thus iNOS inhibitors could potentially lead to cardiotoxic effects,125 and prolonged use of HIV-1 protease inhibitors is linked to an increased risk of myocardial infarction.126 It is thus of key importance for hazard detection to have the most complete knowledge of drug-target interactions and, when completeness is limited, complement that knowledge with high-confidence predictions. If taking off-target pharmacology into consideration offers a new dimension to predictive drug safety, adding information on the biological pathways in which those off-targets are involved would add yet another layer of complexity to it. In this respect, Scheiber et al.127 used Bayesian classification models based on structural fingerprints to predict the binding profiles of compounds producing a common safety event. Then, the most frequent targets in those profiles were cross-linked with information on the pathways those targets were involved, highlighting potential links between pathways and safety events. As an example, the study detected that the most frequent pathways annotated to rhabdomyolysis where cAMP-signaling and steroid metabolism, both of which have a clear relation to possible causes of that particular adverse reaction. Using a similar approach, Aloy and coworkers128-130 explored the interface between chemistry and biology to detect explanations to various drug side effects. With known drugs at the center of their schema, they gathered data coming from several databases with information on therapeutic targets, pathways, protein-protein interactions, molecular functions and biological processes, on one side, and chemical structure terms, molecular scaffolds and small fragments, on the other side. Then, they matched these different entities to safety events reported for those same drugs and calculated a score to reflect their ability to classify drugs according to specific adverse reactions. With this procedure, they were able to detect overrepresented chemical or biological hazardous features, with acceptable discriminant precision, for 164 side effects. All hazards detected at this stage can then be incorporated into novel mechanistic approaches to risk assessment. Risk Assessment. As stated above, safety models based solely on the presence of particular chemical fragments in the structure of molecules combined with statistical analyses form still the backbone of current computational toxicology methods.109,110 For example, Pauwels et al.131 used advanced statistics, including sparse canonical correlation analysis (SCCA), to detect co-occurring sets of side effects and chemical structures in common drugs.


12

Page 13 of 33

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 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 54 55 56 57 58 59 60


Once these significant sets were found, uncharacterized molecules containing fragments present in those sets could be related with the co-occurrent adverse drug reactions. However, beyond traditional structure alerts, current trends are towards offering a more mechanistic view to predictive safety on the basis of the target and pathway hazards detected from known and predicted data links.132 Along these lines, Bender et al.52 used substructure-based Bayesian categorization algorithms to model, on one hand, the affinity of drugs over a panel of targets and, on the other hand, the observation of side effects, both models showing an acceptable performance in the validation against a few drugs. One of the features of that methodology is that the findings can be back-mapped to specific sets of compound fragments and, comparing the sets relevant for targets and adverse events, the authors could detect mechanistic evidence for links between common adverse events and proteins long known for being related with off-target toxicity, such as muscarinic or opioid receptors and COX-1. Mizutani et al.133 used also advanced statistical methodologies on drug-related information gathered from protein affinity databases and side effect profiles repositories to detect pathways that could be behind specific pools of side effects. As stated before, these sets of side effects enriched in drugs altering specific pathways could explain why compounds with different chemical features and thus, with affinity for different proteins with non-matching biological functions, end up producing similar adverse events. This finding allowed the authors to predict, to a certain extent, the adverse reactions that several probe compounds altering those relevant pathways could cause. Both works emphasized the added value of adding target and pathway information to structural information to develop a new generation of more sensitive and precise safety models. To further exemplify this aspect, we collected some of the fragments, drugs, targets, and pathways that link amitriptyline to some safety events, namely, rash, palpitations, dyskinesia, and miosis (Figure 3a). Essentially, the analysis is based on assessing whether the drug being processed (amitriptyline, in this case) contains some of the chemical and biological hazards (fragments, targets, and pathways) detected in drugs linked to specific safety events. In addition, results are complemented with a list of drugs reported to be linked to those safety events and found to be similar to amitriptyline,115 under the assumption that similar drugs are likely to interact with similar targets and thus to have similar side effects.111 On this basis, it is found for example that 60.3% of the drugs linked to dyskinesia contain the fragment shown in Figure 3a, whereas only 18.6% of the drugs not associated with dyskinesia contain that fragment. The


13


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 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 54 55 56 57 58 59 60

Page 14 of 33

presence of that fragment in amitriptyline is thus considered a chemical hazard and its enrichment in drugs linked to dyskinesia (41.7%) can be regarded as the associated risk from a pure chemical perspective. In addition, it is also found that 40.5% of the drugs linked to dyskinesia have biologically-relevant binding affinity (pKi>5) for the dopamine D1A receptor,134 whereas only 8.5% of the drugs not associated with dyskinesia bind to that receptor. In this respect, the dopamine D1A receptor is a protein hazard for dyskinesia and the enrichment (32.0%) and affinity of amitriptyline for it will reflect the associated risk from a pharmacological perspective. Also, it appears that 56.2% of the drugs linked to dyskinesia have affinity for one or more targets involved in dopamine receptor signaling, whereas only 13.4% of the drugs not associated with dyskinesia interfere also with that pathway. Therefore, interfering with the dopamine receptor signaling emerges as a pathway hazard and its enrichment in connection with dyskinesia (42.8%) as the risk associated to it. Finally, amitriptyline is also found to be similar to doxepin, a drug linked to dyskinesia.39 Both the individual and combined significances of the fragment, target, pathway, and drug links to dyskinesia are illustrated in Figure 3b. As can be observed, significant percentages of drugs linked to dyskinesia (Figure 3b, left) contain the fragment, have affinity for D1A, interfere with the dopamine receptor signaling pathway, and/or are similar to doxepin. In contrast, a relatively much lower percentage of drugs not associated with dyskinesia (Figure 3b, right) fulfill those criteria. Therefore, it seems obvious that taking all these evidences together, that is, the fact that amitriptyline contains that fragment, it has affinity for the dopamine D1A receptor, it interferes with the dopamine receptor signaling pathway, and it is similar to doxepin, provides a higher confidence for the risk of amitriptyline being linked to dyskinesia than just using a structure alert only.109,110 Obviously, this is just an illustrative example of one fragment, one target, one pathway, and one drug linked to dyskinesia. But one ought to take into consideration that there is an ensemble of fragments, targets, pathways, and drugs identified as being significantly linked to dyskinesia. Taking it all together, the risk linking amitriptyline to dyskinesia is further strengthened compared to just considering structural alerts. Similar lines of argumentation are applicable to the other safety events, namely, rash, palpitations, and miosis, as illustrated in Figure 3a.


14

Page 15 of 33

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 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 54 55 56 57 58 59 60


CONCLUSIONS AND OUTLOOK Adverse drug reactions have traditionally been considered to be caused mainly by the potential for reactive metabolite formation (idiosyncratic adverse drug reactions).135 But more attention is now being paid to understand the risks associated with excessive on-target effects, on-target effects in secondary organ systems, and/or off-target pharmacology (mechanistic adverse drug reactions).51-53 Modern drug discovery has come a long way to understand, and adapt to the fact, that the human body is a complex system sustained by a balance of multiple imbricated connections between chemical and biological entities with specific functions at different levels.132 Their superposition can be described by a set of visible characteristics that we refer to as phenotype. A wide variety of different internal and external factors can affect this delicate but resilient system, producing alterations in the phenotype that are often detrimental for the overall fitness. These processes lead to what we call diseases. To achieve the desired rebalance, drug discovery focuses on the design of small molecules with key features that emulate the simplest elements of the system.136 These molecules should be able to interact with concrete entities of the cellular machinery, proteins, and modulate their behavior thus, restoring the system to its initial healthy state. Unfortunately, many diseases are not caused by the malfunction of a single protein and, even when that is the case, modulation of a protein usually triggers pleiotropic effects due to interferences with several other network nodes. This means that restoring the properties of a malfunctioning node through the addition of a new element does not guarantee that the original phenotype is going to be restored. In addition, selective interaction of a small molecule with the desired protein target is more difficult than initially expected and quite often affinities for multiple off-target proteins are encountered. In many cases, this is the cause of unwanted adverse drug reactions, a new set of phenotypic variations related to the drug treatment that can be severe and even fatal. Both human and economic costs of adverse drug reactions have been displacing the focus of drug discovery field towards predictive drug safety. Nowadays the first databases and profiling methodologies are already available for classical toxic endpoints, such as mutagenicity or skin sensitization, but researchers are far from reaching the same level with target-mediated and more complex safety events. The main obstacles to achieve this point are the lack of available information motivated by several causes including the high cost of in vivo and in vitro toxicity assays, the absence of published negative data from compounds tested for a specific adverse


15


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 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 54 55 56 57 58 59 60

Page 16 of 33

event and resulted safe, the bureaucratic limitations to access databases of medical data, or the country specific legal singularities that dictate the information to be presented in drug labels, among others. It ought to be stressed that the lack of a drug-safety link should not be interpreted as “drug devoid or clean of that safety event” but more as the fact that “the particular safety event has not yet been observed in an individual patient taking that drug, reported in a pharmacovigilance system, and properly stored in the corresponding database”. Along these lines, a recent report from the EMA Pharmacovigilance Risk Assessment Committee concluded that “the amount of time a drug has been on the market is correlated with the number of signals detected”,137 a clear message alerting on the incompleteness of current drug-safety links and encouraging to develop more confident predictive systems to continuously look for new signals and anticipate their risk. In this respect, the construction of a well-curated database of negative drug-safety links would be extremely helpful in building more reliable safety prediction models.138,139 We have already started to see an increase in the number of data handling and research efforts, as well as in the quantity and quality of their contents, paving the way for statistical applications to large scale safety-related hazard detection and risk assessment. With these methodologies, the analysis of networks of multiple entities involved in adverse drug reactions may reveal unsuspected relationships between phenotypic alterations and the different systemic entities, such as small molecules, proteins, and pathways. Within this new scenario, are we to abandon structural alerts? Certainly not. Some long-studied well-understood safety events may be directly linked to the presence of labile functional groups or chemical fragments and, for these cases, traditional structural alerts will be sufficient to get a fair assessment of any potential safety risk. In contrast, other safety events may be mediated by the drug’s primary and/or secondary pharmacology. For these cases, currently used structural alerts may actually reflect the main chemical features involved in drug-target binding and still provide useful information that adds confidence in risk assessment. Precisely in an attempt to reduce the risk of non-obvious offtarget affinities that could lead to some of those events, drug candidates are now routinely processed against selected in vitro assays of safety-relevant targets.140 But even then, the rather limited target coverage of these diversity panels should be always supported by the use of in silico pharmacology approaches to safely rule out undesirable late surprises due to distant polypharmacology.141 However, for the more complex adverse drug reactions one could be


16

Page 17 of 33

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 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 54 55 56 57 58 59 60


facing multiple initiating events involving the presence of labile chemical fragments, the interaction with various off-targets, and the interference with certain biological pathways, all acting in synergistic synchrony. It is in these particularly complex cases where integrative structural and mechanistic methods, such as the one exemplified here, will be most useful. We are still a very long way from rationalizing and reliably anticipating drug safety but efforts towards understanding toxicity cascades from a systems perspective are worth pursuing and should be continued to be explored and further developed.

AUTHOR INFORMATION Corresponding Author E-mail: [email protected] Funding This research was supported by the Spanish Ministerio de Economía y Competitividad (project BIO2014-54404-R). Funding was received from the European Community’s Innovative Medicines Initiative, under Grant agreement No. 115002 (eTOX project). Notes The authors are affiliated with Chemotargets SL (Barcelona), the company that develops CTlink, a software for predicting the off-target pharmacology and safety profiles of small molecules.

ACKNOWLEDGMENTS We are grateful for the ideas and comments received from all partners involved in the eTOX project.

ABBREVIATIONS AERS, adverse events reporting system; CNS, central nervous system; COX-1, cyclooxygenase type 1; CTD, comparative toxicogenomics database; EBI, European Bioinformatics Institute; EMA, European Medicines Agency; EPA, US environmental protection agency; FDA, US food and drug administration; hERG, human

ether-a-go-go-related

gene; HIV-1, human

immunodeficiency virus type 1; IMI-JU, innovative medicines initiative joint undertaking; iNOS,


17


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 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 54 55 56 57 58 59 60

Page 18 of 33

inducible nitric oxide synthase; IPCI, interdisciplinary processing of clinical information; KEGG, Kyoto encyclopedia of genes and genomes; NCBI, national center for biotechnology information; NIH, national institutes of health; OECD, organization for economic co-operation and development, OICR, Ontario Institute for Cancer Research; PPARγ, peroxisome proliferatoractivated receptor gamma; QSAR, quantitative structure-activity relationships; SIDER, side effect resource; TdP, Torsades de Pointes; TG-GATEs, toxicogenomics project genomics assisted toxicity evaluation system; WHO, world health organization.


18

Page 19 of 33

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 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 54 55 56 57 58 59 60


REFERENCES (1) Arrowsmith, J., Miller, P. (2013) Trial watch: phase II and phase III attrition rates 2011-2012. Nat. Rev. Drug Discov. 12, 569. (2) Arrowsmith, J. (2011) Trial watch: phase II failure 2008-2010. Nat. Rev. Drug Discov. 10, 328-329. (3) Furberg, C. D., and Pitt, B. (2001) Withdrawal of cerivastatin from the world market. Curr. Control Trials Cardiovasc. Med. 2, 205-207. (4) EU-ADR project. http://www.euadr-project.org (accessed 20.3.15). (5) Oliveira, J. L., Lopes, P., Nunes, T., Campos, D., Boyer, S., Ahlberg, E., van Mulligen, E. M., Kors, J. A., Singh, B., Furlong, L. I., Sanz, F., Bauer-Mehren, A., Carrascosa, M. C., Mestres, J., Avillach, P., Diallo, G., Díaz Acedo, C., and van der Lei, J. (2013) The EU-ADR web platform: delivering advanced pharmacovigilance tools. Pharmacoepidemiol. Drug Saf. 22, 459-467. (6) Giacomini, K. M., Krauss, R. M., Roden, D. M., Eichelbaum, M., Hayden, M. R., and Nakamura, Y. (2007) When good drugs go bad. Nature 446, 975-977. (7) eTOX project. http://www.etoxproject.eu (accessed 20.3.15). (8) Arrowsmith, J. (2011) Trial watch: phase III and submission failures 2007-2010. Nat. Rev. Drug Discov. 10, 87. (9) Cases, M., Briggs, K., Steger-Hartmann, T., Pognan, F., Marc, P., Kleinöder, T., Schwab, C. H., Pastor, M., Wichard, J., and Sanz, F. (2014) The eTOX data-sharing project to advance in silico drug-induced toxicity prediction. Int. J. Mol. Sci. 15, 21136-21154. (10) Briggs, K., Cases, M., Heard, D. J., Pastor, M., Pognan, F., Sanz, F., Schwab, C. H., Steger-Hartmann, T., Sutter, A., Watson, D. K., and Wichard, J. D. (2012) Inroads to predict in vivo toxicology – an introduction to the eTOX project. Int. J. Mol. Sci. 13, 3820-3846. (11) Mestres, J., Bryant, S. D., Zamora, I., and Gasteiger, J. (2011) Shaping the future of safer innovative drugs in Europe. Nat. Biotechnol. 29, 789-790. (12) ToxCast project. http://www.epa.gov/ncct/toxcast (accessed 20.3.15). (13) Tox21 platform. http://epa.gov/ncct/Tox21 (accessed 20.3.15). (14) Kleinstreuer, N. C., Yang, J., Berg, E. L., Knudsen, T. B., Richard, A. M., Martin, M. T., Reif, D. M., Judson, R. S., Polokoff, M., Dix, D. J., Kavlock, R. J., and Houck, K. A. (2014) Phenotypic screening of the ToxCast chemical library to classify toxic and therapeutic mechanisms. Nat. Biotechnol. 32, 583-591. (15) Attene-Ramos, M. S., Miller, N., Huang, R., Michael, S., Itkin, M., Kavlock, R. J., Austin, C. P., Shinn, P., Simeonov, A., Tice, R. R., and Xia, M. (2013) The Tox21 robotic platform for the assessment of environmental chemicals – from vision to reality. Drug Discov. Today 18, 716-723. (16) Open TG-GATEs. http://toxico.nibio.go.jp/english/index.html (accessed 20.3.15). (17) Igarashi, Y., Nakatsu, N., Yamashita, T., Ono, A., Ohno, Y., Urushidani, T., and Yamada, H. (2015) Open TGGATEs: a large-scale toxicogenomics database. Nucleic Acids Res. 43, D921-D927. (18) Williams, D. P. (2006) Toxicophores: investigations in drug safety. Toxicology 226, 1-11.


19


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 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 54 55 56 57 58 59 60

Page 20 of 33

(19) Ahlberg, E., Carlsson, L., and Boyer, S. (2014) Computational derivation of structural alerts from large toxicology data sets. J. Chem. Inf. Model. 54, 2945-2952. (20) Gaulton, A., Bellis, L. J., Bento, A. P., Chambers, J., Davies, M., Hersey, A., Light, Y., McGlinchey, S., Michalovich, D., Al-Lazikani, B., and Overington, J. P. (2012) ChEMBL: a large-scale bioactivity database for drug discovery. Nucleic Acids Res. 40, D1100-D1107. (21) Williams, A. J., Ekins, S., Spjuth, O., and Willighagen, E. L. (2012) Accessing, using, and creating chemical property databases for computational toxicology modeling. Methods Mol. Biol. 929, 221-241. (22) Williams, A. J., Harland, L., Groth, P., Pettifer, S., Chichester, C., Willighagen, E. L., Evelo, C. T., Blomberg, N., Ecker, G., Goble, C., and Mons, B. (2012) Open PHACTS: semantic interoperability for drug discovery. Drug Discov. Today 17, 1188-1198. (23) Bouhifd, M., Andersen, M. E., Baghdikian, C., Boekelheide, K., Crofton, K. M., Fornace, A. J., Jr., Kleensang, A., Li, H., Livi, C., Maertens, A., McMullen, P. D., Rosenberg, M., Thomas, R., Vantangoli, M., Yager, J. D., Zhao, L., and Hartung, T. (2015) The human toxome project. ALTEX 32, 112-124. (24) Tollefsen, K. E., Scholz, S., Cronin, M. T., Edwards, S. W., de Knecht, J., Crofton, K., Garcia-Reyero, N., Hartung, T., Worth, A., and Patlewicz, G. (2014) Applying adverse outcome pathways (AOPs) to support integrated approaches to testing and assessment (IATA). Regul. Toxicol. Pharmacol. 70, 629-640. (25) Chiang, A. P., and Butte, A. J. (2009) Data-driven methods to discover molecular determinants of serious adverse drug events. Clin. Pharmacol. Ther. 85, 259-268. (26) Pouliot, Y., Chiang, A. P., and Butte, A. J. (2011) Predicting adverse drug reactions using publicly available PubChem BioAssay data. Clin. Pharmacol. Ther. 90, 90-99. (27) Nigsch, F., Lounkine, E., McCarren, P., Cornett, B., Glick, M., Azzaoui, K., Urban, L., Marc, P., Müller, A., Hahne, F., Heard, D. J., and Jenkins, J. L. (2011) Computational methods for early predictive safety assessment from biological and chemical data. Expert Opin. Drug Metab. Toxicol. 7, 1497-1511. (28) Lee, S., Lee, K. H., Song, M., and Lee, D. (2011) Building the process-drug-side effect network to discover the relationship between biological processes and side effects. BMC Bioinformatics 12, S2. (29) Bauer-Mehren, A., van Mullingen, E. M., Avillach, P., Carrascosa, M. C., Garcia-Serna, R., Piñero, J., Singh, B., Lopes, P., Oliveira, J. L., Diallo, G., Ahlberg Helgee, E., Boyer, S., Mestres, J., Sanz, F., Kors, J. A., and Furlong, L. I. (2012) Automatic filtering and substantiation of drug safety signals. PLoS Comput. Biol. 8, e1002457. (30) Declerck, G., Hussain, S., Daniel, C., Yuksel, M., Laleci, G. B., Twagirumukiza, M., and Jaulent, M. C. (2015) Bridging data models and terminologies to support adverse drug event reporting using EHR data. Methods Inf. Med. 54, 24-31. (31) Liu, M., Hu, Y., and Tang, B. (2014) Role of text mining in early identification of potential drug safety issues. Methods Mol. Biol. 1159, 227-251. (32) White, R. W., Tatonetti, N. P., Shah, N. H., Altman, R. B., and Horvitz, E. (2013) Web-scale pharmacovigilance: listening to signals from the crowd. J. Am. Med. Inform. Assoc. 20, 404-408. (33) Freifeld, C. C., Brownstein, J. S., Menone, C. M., Bao, W., Filice, R., Kass-Hout, T., and Dasgupta, N. (2014) Digital drug safety surveillance: monitoring pharmaceutical products in twitter. Drug Saf. 37, 343-350.


20

Page 21 of 33

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 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 54 55 56 57 58 59 60


(34) Bousquet, C., Sadou, É., Souvignet, J., Jaulent, M. C., and Declerck, G. (2014) Formalizing MedDRA to support semantic reasoning on adverse drug reaction terms. J. Biomed. Inform. 49, 282-291. (35) Cai, M. C., Xu, Q., Pan, Y. J., Pan, W., Ji, N., Li, Y. B., Jin, H. J., Liu, K., and Ji, Z. L. (2015) ADReCS: an ontology database for aiding standardization and hierarchical classification of adverse drug reaction terms. Nucleic Acids Res. 43, D907-D913. (36) Strauss, A., Lawrence Gould, A., Nelsen, L. M., Wiholm, B. E., and Jones, J. K. (2005) Public-released version of the Adverse Event Reporting System (AERS) database. Clin. Ther. 27, 355-357. (37) Tatonetti, N. P., Ye, P. P., Daneshjou, R., and Altman, R. B. (2012) Data-driven prediction of drug effects and interactions. Sci. Transl. Med. 4, 125ra31. (38) Vlug, A. E., van der Lei, J., Mosseveld, B. M., van Wijk, M. A., van der Linden, P. D., Sturkenboom, M. C., and van Bemmel, J. H. (1999) Postmarketing surveillance based on electronic patient records: the IPCI project. Methods Inf. Med. 38, 339-344. (39) Kuhn, M., Campillos, M., Letunic, I., Jensen, L. J., and Bork, P. (2010) A side effect resource to capture phenotypic effects of drugs. Mol. Syst. Biol. 6, 343. (40) Brown, E. G., Wood, L., and Wood, S. (1999) The medical dictionary for regulatory activities (MedDRA). Drug Saf. 20, 109-117. (41) World Drug Index. http://thomsonreuters.com/world-drug-index (accessed 20.3.15). (42) Pharmapendium. http://www.elsevier.com/online-tools/pharmapendium (accessed 20.3.15). (43) Toxicity Reference Database. http://actor.epa.gov/toxrefdb (accessed 20.3.15). (44) DrugMatrix. http://ntp.niehs.nih.gov/drugmatrix/index.html (accessed 20.3.15). (45) Mestres, J., Gregori-Puigjané, E., Valverde, S., and Solé, R. V. (2008) Data completeness: the Achilles heel of drug-target networks. Nat. Biotechnol. 26, 983-984. (46) Mestres, J., Gregori-Puigjané, E., Valverde, S., and Solé, R. V. (2009) The topology of drug-target interaction networks: implicit dependence on drug properties and target families. Mol. BioSys. 5, 1051-1057. (47) Vogt, I., and Mestres, J. (2010) Drug-target networks. Mol. Inf. 29, 10-14. (48) Peters, J. U. (2013) Polypharmacology: foe or friend? J. Med. Chem. 56, 8955-8971. (49) Jalencas, X., and Mestres, J. (2013) On the origins of drug polypharmacology. Med. Chem. Commun. 4, 80-87. (50) Anighoro, A., Bajorath, J., and Rastelli, G. (2014) Polypharmacology: challenges and opportunities in drug discovery. J. Med. Chem. 57, 7874-7887. (51) Whitebread, S., Hamon, J., Bojanic, D., and Urban, L. (2005) In vitro safety pharmacology profiling: an essential tool for successful drug development. Drug Discov. Today 10, 1421-1433. (52) Bender, A., Scheiber, J., Glick, M., Davies, J. W., Azzaoui, K., Hamon, J., Urban, L., Whitebread, S., and Jenkins, J. L. (2007) Analysis of pharmacology data and the prediction of adverse drug reaction and off-target effects from chemical structure. ChemMedChem 2, 861-873. (53) Hamon, J., Whitebread, S., Techer-Etienne, V., Le Coq, H., Azzaoui, K., and Urban, L. (2009) In vitro safety pharmacology profiling: what else beyond hERG? Future Med. Chem. 1, 645-665.


21


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 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 54 55 56 57 58 59 60

Page 22 of 33

(54) Garcia-Serna, R., and Mestres, J. (2010) Anticipating drug side effects by comparative pharmacology. Expert Opin. Drug Metab. Toxicol. 6, 1253-1263. (55) Mestres, J., Seifert, S. A., and Oprea, T. I. (2011) Linking pharmacology to clinical reports: cyclobenzaprine and its possible association with serotonin syndrome. Clin. Pharmacol. Ther. 90, 662-665. (56) Lounkine, E., Keiser, M. J., Whitebread, S., Mikhailov, D., Hamon, J., Jenkins, J. L., Lavan, P., Weber, E., Doak, A. K., Côté, S., Shoichet, B. K., and Urban, L. (2012) Large-scale prediction and testing of drug activity on side-effect targets. Nature 486, 361-367. (57) Evans, W. E., and McLeod, H. L. (2003) Pharmacogenomics – drug disposition, drug targets, and side effects. N. Engl. J. Med. 348, 538-549. (58) Rieder, M. J. (1993) Immunopharmacology and adverse drug reactions. J. Clin. Pharmacol. 33, 316-323. (59) Park, B. K., Kitteringham, N. R., Maggs, J. L., Pirmohamed, M., and Williams, D. P. (2005) The role of metabolic activation in drug-induced hepatotoxicity. Annu. Rev. Pharmacol. Toxicol. 45, 177-202. (60) Uetrecht, J. (2008) Idiosyncratic drug reactions: past, present, and future. Chem. Res. Toxicol. 21, 84-92. (61) Uetrecht, J, and Naisbitt, D. J. (2013) Idiosyncratic adverse drug reactions: current concepts. Pharmacol. Rev. 65, 779-808. (62) Wang, Y., Suzek, T., Zhang, J., Wang, J., He, S., Cheng, T., Shoemaker, B. A., Gindulyte, A., and Bryant, S. H. (2014) PubChem BioAssay: 2014 update. Nucleic Acids Res. 42, D1075-D1082. (63) Bento, A. P., Gaulton, A., Hersey, A., Bellis, L. J., Chambers, J., Davies, M., Krüger, F. A., Light, Y., Mak, L., McGlinchey, S., Nowotka, M., Papadatos, G., Santos, R., and Overington, J. P. (2014) The ChEMBL bioactivity database: an update. Nucleic Acids Res. 42, D1083-D1090. (64) Hu, Y., and Bajorath, J. (2012) Extending the activity cliff concept: structural categorization of activity cliffs and systematic identification of different types of cliffs in the ChEMBL database. J. Chem. Inf. Model. 52, 18061811. (65) Liu, T., Lin, Y., Wen, X., Jorissen, R. N., and Gilson, M. K. (2007) BindingDB: a web-accessible database of experimentally determined protein-ligand binding affinities. Nucleic Acids Res. 35, D198-D201. (66) Wassermann, A. M., and Bajorath, J. (2011) BindingDB and ChEMBL: online compound databases for drug discovery. Expert Opin. Drug Discov. 6, 683-687. (67) Law, V., Knox, C., Djoumbou, Y., Jewison, T., Guo, A. C., Liu, Y., Maciejewski, A., Arndt, D., Wilson, M., Neveu, V., Tang, A., Gabriel, G., Ly, C., Adamjee, S., Dame, Z. T., Han, B., Zhou, Y., and Wishart, D. S. (2014) DrugBank 4.0: shedding new light on drug metabolism. Nucleic Acids Res. 42, D1091-D1097. (68) Harmar, A. J., Hills, R. A., Rosser, E. M., Jones, M., Buneman, O. P., Dunbar, D. R., Greenhill, S. D., Hale, V. A., Sharman, J. L., Bonner, T. I., Catterall, W. A., Davenport, A. P., Delagrange, P., Dollery, C. T., Foord, S. M., Gutman, G. A., Laudet, V., Neubig, R. R., Ohlstein, E. H., Olsen, R. W., Peters, J., Pin, J. P., Ruffolo, R. R., Searls, D. B., Wright, M. W., and Spedding, M. (2009) IUPHAR-DB: the IUPHAR database of G protein-couple receptors and ion channels. Nucleic Acids Res. 37, D680-D685.


22

Page 23 of 33

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 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 54 55 56 57 58 59 60


(69) Davis, A. P., Grondin, C. J., Lennon-Hopkins, K., Saraceni-Richards, C., Sciaky, D., King, B. L., Wiegers, T. C., and Mattingly, C. J. (2015) The Comparative Toxicogenomics Database’s 10th year anniversary: update 2015. Nucleic Acids Res. 43, D914-D920. (70) PDSP: National Institute of Mental Health's Psychoactive Drug Screening Program. http://pdspdb.unc.edu (accessed 20.3.15). (71) Grover, M. P., Ballouz, S., Mohanasundaram, K. A., George, R. A., Sherman, C. D. H., Crowley, T. M., and Wouters, M. A. (2014) Identification of novel therapeutics for complex diseases from genome-wide association data. BMC Med. Gen. 7(S1), S8. (72) Woo, J. H., Shimoni, Y., Yang, W. S., Subramaniam, P., Iyer, A., Nicoletti, P., Rodríguez-Martínez, M., López, G., Mattioli, M., Realubit, R., Karan, C., Stockwell, B. R., Bansal, M., and Califano, A. (2015) Elucidating compound mechanism of action by network perturbation analysis. Cell 162, 441-451. (73) Vinken, M., Whelan, M., and Rogiers, V. (2014) Adverse outcome pathways: hype or hope? Arch. Toxicol. 88, 1-2. (74) Croft, D., Mundo, A. F., Haw, R., Milacic, M., Weiser, J., Wu, G., Caudy, M., Garapati, P., Gillespie, M., Kamdar, M. R., Jassal, B., Jupe, S., Matthews, L., May, B., Palatnik, S., Rothfels, K., Shamovsky, V., Song, H., Williams, M., Birney, E., Hermjakob, H., Stein, L., and D'Eustachio, P. (2014) The Reactome pathway knowledgebase. Nucleic Acids Res. 42, D472-D477. (75) Cerami, E. G., Gross, B. E., Demir, E., Rodchenkov, I., Babur, O., Anwar, N., Schultz, N., Bader, G. D., and Sander, C. (2011) Pathway Commons, a web resource for biological pathway data. Nucleic Acids Res. 39, D685D690. (76) Kelder, T., van Iersel, M. P., Hanspers, K., Kutmon, M., Conklin, B. R., Evelo, C. T., and Pico, A. R. (2012) WikiPathways: building research communities on biological pathways. Nucleic Acids Res. 40, D1301-D1307. (77) Whirl-Carrillo, M., McDonagh, E. M., Hebert, J. M., Gong, L., Sangkuhl, K., Thorn, C. F., Altman, R. B., and Klein, T. E. (2012) Pharmacogenomics knowledge for personalized medicine. Clin. Pharmacol. Ther. 92, 414-417. (78) Qin, C., Zhang, C., Zhu, F., Xu, F., Chen, S. Y., Zhang, P., Li, Y. H., Yang, S. Y., Wei, Y. Q., Tao, L., and Chen, Y. Z. (2014) Therapeutic target database update 2014: a resource for targeted therapeutics. Nucleic Acids Res. 42, D1118-D1123. (79) AOP-KB. http://aopkb.org. (accessed 20.3.15). (80) Kotera, M., Hirakawa, M., Tokimatsu, T., Goto, S., and Kanehisa, M. (2012) The KEGG databases and tools facilitating omics analysis: latest developments involving human diseases and pharmaceuticals. Methods Mol. Biol. 802, 19-39. (81) Kuhn, M., Al Banchaabouchi, M., Campillos, M., Jensen, L. J., Gross, C., Gavin, A. C., and Bork, P. (2013) Systematic identification of proteins that elicit drug side effects. Mol. Syst. Biol. 9, 663. (82) Shepherd, G. (2006) Treatment of poisoning caused by beta-adrenergic and calcium-channel blockers. Am. J. Health Syst. Pharm. 63, 1828-1835. (83) Ahmed, R., and Branley, H. M. (2009) Reversible bronchospasm with the cardio-selective beta-blocker celiprolol in a non-asthmatic subject. Respir. Med. CME 2, 141-143.


23


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 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 54 55 56 57 58 59 60

Page 24 of 33

(84) Rampe, D., and Brown, A. M. (2013) A history of the role of the hERG channel in cardiac risk assessment. J. Pharmacol. Toxicol. Methods 68, 13-22. (85) Roth, B. L. (2007) Drugs and valvular heart disease. N. Engl. J. Med. 356, 6-9. (86) Davies, M., Dedman, N., Hersey, A., Papadatos, G., Hall, M. D., Cucurull-Sanchez, L., Jeffrey, P., Hasan, S., Eddershaw, P. J., and Overington, J. P. (2015) ADME SARfari: comparative genomics of drug metabolizing systems. Bioinformatics 31, 1695-1697. (87) Briansó, F., Carrascosa, M. C., Oprea, T. I., and Mestres, J. (2011) Cross-pharmacology of G protein-coupled receptors. Curr. Top. Med. Chem. 11, 1956-1963. (88) Kirchmair, J., Williamson, M. J., Tyzack, J. D., Tau, L., Bond, P. J., Bender, A., and Glen, R. C. (2012) Computational prediction of metabolism: sites, products, SAR, P450 enzyme dynamics, and mechanisms. J. Chem. Inf. Model. 52, 617-648. (89) Piechota, P., Cronin, M. T. D., Hewitt, M., Madden, J. C. (2013) Pragmatic approaches to using computational methods to predict xenobiotic metabolism. J. Chem. Inf. Model. 53, 1282-1293. (90) Kirchmair, J., Göller, A. H., Lang, D., Kunze, J., Testa, B., Wilson, I. D., Glen, R. C., and Schneider, G. (2015) Predicting drug metabolism: experiment and/or computation? Nat. Rev. Drug Discov. 14, 387-404. (91) Williams, D. P., and Park, B. K. (2003) Idiosyncratic toxicity: the role of toxicophores and bioactivation. Drug Discov. Today 8, 1044-1050. (92) Kazius, J., McGuire, R., and Bursi, R. (2005) Derivation and validation of toxicophores for mutagenicity prediction. J. Med. Chem. 48, 312-320. (93) Kalgutkar, A. S., Fate, G., Didiuk, M. T., and Bauman, J. (2008) Toxicophores, reactive metabolites and drug safety: when is it a cause for concern? Expert Rev. Clin. Pharmacol. 1, 515-531. (94) Scheiber, J., Jenkins, J. L., Sukuru, S. C., Bender, A., Mikhailov, D., Milik, M., Azzaoui, K., Whitebread, S., Hamon, J., Urban, L., Glick, M., and Davies, J. W. (2009) Mapping adverse drug reactions in chemical space. J. Med. Chem. 52, 3103-3107. (95) Ellison, C. M., Enoch, S. J., and Cronin, M. T. D. (2011) A review of the use of in silico methods to predict the chemistry of molecular initiating events related to drug toxicity. Expert Opin. Drug Metab. Toxicol. 7, 1481-1495. (96) Przybylak, K. R., and Cronin, M. T. D. (2011) In silico studies of the relationship between chemical structure and drug induced phospholipidosis. Mol. Inf. 30, 415-429. (97) Sullivan, K. M., Manuppello, J. R., and Willett, C. E. (2014) Building on a solid foundation: SAR and QSAR as a fundamental strategy to reduce animal testing. SAR QSAR Environ. Res. 25, 357-365. http://qsartoolbox.org (accessed 20.3.15). (98) Bhatia, S., Schultz, T., Roberts, D., Shen, J., Kromidas, L., and Marie Api, A. (2015) Comparison of Cramer classification between ToxTree, the OECD QSAR Toolbox and expert judgment. Regul. Toxicol. Pharmacol. 71, 52-62. (99) Patlewicz, G., Jeliazkova, N., Safford, R. J., Worth, A. P., and Aleksiev, B. (2008) An evaluation of the implementation of the Cramer classification scheme in the Toxtree software. SAR QSAR Environ. Res. 19, 495-524. http://toxtree.sourceforge.net (accessed 20.3.15).


24

Page 25 of 33

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 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 54 55 56 57 58 59 60


(100) Hardy, B., Douglas, N., Helma, C., Rautenberg, M., Jeliazkova, N., Jeliazkov, V., Nikolova, I., Benigni, R., Tcheremenskaia, O., Kramer, S., Girschick, T., Buchwald, F., Wicker, J., Karwath, A., Gütlein, M., Maunz, A., Sarimveis, H., Melagraki, G., Afantitis, A., Sopasakis, P., Gallagher, D., Poroikov, V., Filimonov, D., Zakharov, A., Lagunin, A., Gloriozova, T., Novikov, S., Skvortsova, N., Druzhilovsky, D., Chawla, S., Ghosh, I., Ray, S., Patel, H., and Escher, S. (2010) Collaborative development of predictive toxicology applications. J. Chemoinf. 2, 7. http://opentox.org (accessed 20.3.15). (101) Sushko, I., Salmina, E., Potemkin, V. A., Poda, G., and Tetko, I. V. (2012) ToxAlerts: A web server of structural alerts for toxic chemicals and compounds with potential adverse reactions. J. Chem. Inf. Model. 52, 23102316. (102) Sushko, I., Novotarskyi, S., Körner, R., Pandey, A. K., Rupp, M., Teetz, W., Brandmaier, S., Abdelaziz, A., Prokopenko, V. V., Tanchuk, V. Y., Todeschini, R., Varnek, A., Marcou, G., Ertl, P., Potemkin, V., Grishina, M., Gasteiger, J., Schwab, C., Baskin, I. I., Palyulin, V. A., Radchenko, E. V., Welsh, W. J., Kholodovych, V., Chekmarev, D., Cherkasov, A., Aires-de-Sousa, J., Zhang, Q. Y., Bender, A., Nigsch, F., Patiny, L., Williams, A., Tkachenko, V., and Tetko, I. V. (2011) Online chemical modeling environment (OCHEM): Web platform for data storage, model development and publishing of chemical information. J. Comput. Aided. Mol. Design 25, 533-554. http://ochem.eu (accessed 20.3.15). (103) TOPKAT. http://accelrys.com/products/discovery-studio/admet.html (accessed 20.3.15). (104) Mombelli, E. (2008) An evaluation of the predictive ability of the QSAR software packages DEREK, HAZARDEXPERT, and TOPKAT to describe chemically-induced skin irritation. Altern. Lab. Anim. 36, 15-24. (105) Snyder, R. D., Pearl, G. S., Mandakas, G., Choy, W. N., Goodsaid, F., and Rosenblum, I. Y. (2004) Assessment of the sensitivity of the computational programs DEREK, TOPKAT, and MCASE in the prediction of the genotoxicity of pharmaceutical molecules. Environ. Mol. Mutagen. 43, 143-158. (106) Cariello, N. F., Wilson, J. D., Britt, B. H., Wedd, D. J., Burlinson, B., and Gombar, V. (2002) Comparison of the computer programs DEREK and TOPKAT to predict bacterial mutagenicity. Mutagenesis 17, 321-329. (107) Roberts, G., Myatt, G. J., Johnson, W. P., Cross, K. P., and Blower, P. E., Jr. (2000) LeadScope: software for exploring

large

sets

of

screening

data.

J.

Chem.

Inf.

Comput.

Sci.

40,

1302-1314.

http://www.leadscope.com/model_appliers (accessed 20.3.15). (108) HazardExpertPro. http://www.compudrug.com/hazardexpertpro (accessed 20.3.15). (109) Marchant, C. A., Briggs, K. A., and Long, A. (2008) In silico tools for sharing data and knowledge on toxicity and

metabolism: Derek for

Windows,

Meteor,

and

Vitic.

Toxicol.

Mech. Methods 18,

189-206.

http://www.lhasalimited.org/products/derek-nexus.htm (accessed 20.3.15). (110) Saiakhov, R. D., and Klopman, G. (2008) MultiCASE Expert Systems and the REACH initiative. Toxicol. Mech. Methods 18, 159-175. http://www.multicase.com (accessed 20.3.15). (111) Campillos, M., Kuhn, M., Gavin, A. C., Jensen, L. J., and Bork, P. (2008) Drug target identification using side-effect similarity. Science 321, 263-266. (112) Simmons, D. L., Botting, R. M., and Hla, T. (2004) Cyclooxygenase isozymes: the biology of prostaglandin synthesis and inhibition. Pharmacol. Rev. 56, 387-437.


25


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 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 54 55 56 57 58 59 60

Page 26 of 33

(113) Koutsoukas, A., Simms, B., Kirchmair, J., Bond, P. J., Whitmore, A. V., Zimmer, S., Young, M. P., Jenkins, J. L., Glick, M., Glen, R. C., and Bender, A. (2011) From in silico target prediction to multi-target drug design: current databases, methods, and applications. J. Proteomics 74, 2554-2574. (114) Schmidt, F., Matter, H., Hessler, G., and Czich, A. (2014) Predictive in silico off-target profiling in drug discovery. Future Med Chem. 6, 295-317. (115) Vidal, D., Garcia-Serna, R., and Mestres, J. (2011) Ligand-based approaches to in silico pharmacology. Methods Mol. Biol. 672, 489-502. (116) CTlink software, version 2015 (Chemotargets SL, Barcelona). http://www.chemotargets.com (accessed 20.3.15). (117) Thanacoody, H. K., and Thomas, S. H. (2005) Tricyclic antidepressant poisoning: cardiovascular toxicity. Toxicol. Rev. 24, 205-214. (118) Hou, V. Y., and Hirshman, C. A., and Emala, C.W. (1998) Neuromuscular relaxants as antagonists for M2 and M3 muscarinic receptors. Anesthesiology 88, 744-750. (119) Søndergaard, M.T., Sorensen, A. B., Skov, L. L., Kjaer-Sorensen, K., Bauer, M. C., Nyegaard, M., Linse, S., Oxvig, C., and Overgaard, M. T. (2015) Calmodulin mutations causing catecholaminergic polymorphic ventricular tachycardia confer opposing functional and biophysical molecular changes. FEBS J. 282, 803-816. (120) Ciudin, A., Hernandez, C., and Simó, R. (2012) Update on cardiovascular safety of PPARgamma agonists and relevance to medicinal chemistry and clinical pharmacology. Curr. Top. Med. Chem. 12, 585-604. (121) Kröger, E., Berkers, M., Carmichael, P. H., Souverein, P., van Marum, R., and Egberts, T. (2012) Use of rivastigmine or galantamine and risk of adverse cardiac events: a database study from the Netherlands. Am. J. Geriatr. Pharmacother. 10, 373-380. (122) Keefe, D.L. (2001) Anthracycline-induced cardiomyopathy. Semin. Oncol. 28, 2-7. (123) English, B. A., Appalsamy, M., Diedrich, A., Ruggiero, A.M., Lund, D., Wright, J., Keller, N. R., Louderback, K. M., Robertson, D., and Blakely, R. D. (2010) Tachycardia, reduced vagal capacity, and agedependent ventricular dysfunction arising from diminished expression of the presynaptic choline transporter. Am. J. Physiol. Heart. Circ. Physiol. 299, H799–H810. (124) Zhu, Q.M., Lesnick, J.D., Jasper, J.R., MacLennan, S.J., Dillon, M.P., Eglen, R. M., and Blue, D. R. (1999) Cardiovascular effects of rilmenidine, moxonidine and clonidine in conscious wild-type and D79N α2Aadrenoceptor transgenic mice. Br. J. Pharmacol. 126, 1522-1530. (125) Golomb, E., Nyska, A., and Schwalb, H. (2009) Occult cardiotoxicity-toxic effects on cardiac ischemic tolerance. Toxicol. Pathol. 37, 572-593 (126) Mary-Krause, M., Cotte, L., Simon, A., Partisani, M., and Costagliola, D., Clinical Epidemiology Group from the French Hospital Database. (2003) Increased risk of myocardial infarction with duration of protease inhibitor therapy in HIV-infected men. AIDS 17, 2479-2486. (127) Scheiber, J., Chen, B., Milik, M., Sukuru, S. C., Bender, A., Mikhailov, D., Whitebread, S., Hamon, J., Azzaoui, K., Urban, L., Glick, M., Davies, J. W., and Jenkins, J. L. (2009) Gaining insight into off-target mediated


26

Page 27 of 33

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 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 54 55 56 57 58 59 60


effects of drug candidates with a comprehensive systems chemical biology analysis. J. Chem. Inf. Model. 49, 308317. (128) Duran-Frigola, M., and Aloy, P. (2013) Analysis of chemical and biological features yields mechanistic insights into drug side effects. Chem. Biol. 20, 594-603. (129) Duran-Frigola, M., Rossell, D., and Aloy, P. (2014) A chemo-centric view of human health and disease. Nat. Commun. 5, 5676. (130) Juan-Blanco, T., Duran-Frigola, M., and Aloy, P. (2015) IntSide: a web server for the chemical and biological examination of drug side effects. Bioinformatics 31, 612-613. (131) Pauwels, E., Stoven, V., and Yamanishi, Y. (2011) Predicting drug side-effect profiles: a chemical fragmentbased approach. BMC Bioinformatics 12, 169. (132) Sturla, S. J., Boobis, A. R., FitzGerald, R. E., Hoeng, J., Kavlock, R. J., Schirmer, K., Whelan, M., Wilks, M. F., and Peitsch, M. C. (2014) Systems toxicology: from basic research to risk assessment. Chem. Res. Toxicol. 27, 314-329. (133) Mizutani, S., Pauwels, E., Stoven, V., Goto, S., and Yamanishi, Y. (2012) Relating drug-protein interaction network with drug side effects. Bioinformatics 28, 522-528. (134) Van Kampen, J. M., and Stoessl, A. J. (2000) Dopamine D1A receptor function in a rodent model of tardive dyskinesia. Neuroscience 101, 629-635. (135) Kalgutkar, A. S., and Dalvie, D. (2015) Predicting toxicities of reactive metabolite-positive drug candidates. Annu. Rev. Pharmacol. Toxicol. 55, 35-54. (136) Hert, J., Irwin, J. J., Laggner, C., Keiser, M. J., and Shoichet, B. K. (2009) Quantigying biogenic bias in screening libraries. Nat. Chem. Biol. 5, 479-483. (137) Pacurariu, A. C., Coloma, P. M., van Haren, A., Genov, G., Sturkenboom, M. C. J. M., and Straus, S. M. J. M. (2014) Drug Saf. 37, 1059-1066. (138) Schuemie, M. J., Coloma, P. M., Straatman, H., Herings, R. M., Trifirò, G., Matthews, J. N., Prieto-Merino, D., Molokhia, M., Pedersen, L., Gini, R., Innocenti, F., Mazzaglia, G., Picelli, G., Scotti, L., van der Lei, J., and Sturkenboom, M. C. (2012) Using electronic health care records for drug safety signal detection: a comparative evaluation of statistical methods. Med. Care 50, 890-897. (139) Obach, R. S., Kalgutkar, A. S., Soglia, J. R., and Zhao, S. X. (2008) Can in vitro metabolism-dependent covalent binding data in liver microsomes distinguish hepatotoxic from nonhepatotoxic drugs? An analysis of 18 drugs with consideration of intrinsic clearance and daily dose. Chem. Res. Tox. 21, 1814-1822. (140) Papoian, T., Chiu, H.-J., Elayan, I., Jagadeesh, G., Khan, I., Laniyonu, A. A., Li, C. X., Saulnier, M., Simpson, N., and Yang, B. (2015) Nat. Rev. Drug Discov. 14, 294-296. (141) Antolín, A. A., and Mestres, J. (2015) Distant polypharmacology among MLP chemical probes. ACS Chem. Biol. 10, 395-400.


27


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 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 54 55 56 57 58 59 60

Page 28 of 33

Headshots and biographies

Ricard Garcia-Serna received his BA degree in Biology from the University of Barcelona in 2001. He then obtained a technical degree in Computer Science before attending graduate school at the Pompeu Fabra University where in 2010 he earned his PhD in Biomedicine under the the supervision of Dr. Jordi Mestres. He then joined Chemotargets, where he is currently responsible for developing new systems approaches to drug safety risk assessment.

David Vidal received his PhD in Organic Chemistry from the University of Barcelona in 2006. Under the supervision of Prof. Miquel Pons, he specialized in the design and development of modeling techniques and algorithms for drug discovery. After a short post-doctoral stage in Germany, he joined Chemotargets in 2007, where he has contributed to the development of new methods to predict off-target pharmacology and participated in different drug discovery projects, mainly focused on hit identification and hit to lead optimization. His scientific interests are focused on the interface between chemistry and biology and, particularly, the understanding of the physico-chemical basis of molecular recognition.


28

Page 29 of 33

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 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 54 55 56 57 58 59 60


Nikita Remez received his BA degree in Biotechnology from the Universitat Autònoma de Barcelona in 2008. As a final degree project, he explored various methods for the computational prediction of gene functions under the supervision of Prof. Antonio Gomez. In 2010, he received a Master degree in Bioinformatics from the Universitat Pompeu Fabra under the supervision of Dr. Jordi Mestres. In 2011, he received a Talent Empresa fellowship from the Government of Catalonia to carry out his PhD thesis within Chemotargets on novel computational approaches to systems drug discovery.

Jordi Mestres holds a PhD in Computational Chemistry from the University of Girona. After a postdoctoral stay at Pharmacia&Upjohn (USA), in 1997 he joined the Molecular Design & Informatics department at N.V. Organon (The Netherlands) and in 2000 he was appointed Head of Computational Medicinal Chemistry at Organon Laboratories (Scotland). In 2003, he took on his current position as Head of Systems Pharmacology at the IMIM Hospital del Mar Medical Research Institute in Barcelona. He is also Associate Professor at the University Pompeu Fabra. In 2006, he founded Chemotargets as a spin-off company from his group. His most recent research interests focus on developing novel systems approaches to gain insights on how small molecules interfere with the human metabolome. He is the author of over 130 publications, 8 patents among them.


29


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 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 54 55 56 57 58 59 60

Page 30 of 33

Figure 1. Some relevant data sources for building computational approaches to large scale predictive drug safety


30

Page 31 of 33

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 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 54 55 56 57 58 59 60


Figure 2. (a) Currently experimentally known affinities (left) and additionally predicted interactions (right) for a set of drugs linked to cardiotoxicity; (b) the list of most relevant targets linked to cardiotoxicity events using only known affinities (left) or adding predicted affinities to those known (right); the numerical scale represents the difference between the percentage of drugs being active on a certain target and associated with cardiotoxicity, and those that, while being active on that same target, are not linked (yet) to cardiotoxicity


31


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 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 54 55 56 57 58 59 60

Page 32 of 33

a)

b)

Dyskinesia +

Dyskinesia −

Figure 3. Some safety events linked to amitriptyline. (a) hazard identification: examples of fragments, targets, and pathways significantly found in drugs linked to rash, palpitations, dyskinesia, and miosis that are also found in amitriptyline, as well as the most similar drug linked to the respective safety events; (b) risk assessment: relative relevance of those fragments, drugs, targets, and pathways in drugs linked to dyskinesia (left) and not associated with it (right)


32

Page 33 of 33

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 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 54 55 56 57 58 59 60


TOC Graphics


33

Large-Scale Predictive Drug Safety: From Structural Alerts to

Recommend Documents