In Silico Drug Design: Repurposing Techniques and Methodologies

By Kunal Roy

In Silico Drug Design: Repurposing Techniques and Methodologies explores the application of computational tools that can be utilized for this approach. The book covers theoretical background and methodologies of chem-bioinformatic techniques and network modeling and discusses the various applied strategies to systematically retrieve, integrate and analyze datasets from diverse sources. Other topics include in silico drug design methods, computational workflows for drug repurposing, and network-based in silico screening for drug efficacy. With contributions from experts in the field and the inclusion of practical case studies, this book gives scientists, researchers and R&D professionals in the pharmaceutical industry valuable insights into drug design.

• Discusses the theoretical background and methodologies of useful techniques of cheminformatics and bioinformatics that can be applied for drug repurposing
• Offers case studies relating to the in silico modeling of FDA-approved drugs for the discovery of antifungal, anticancer, antiplatelet agents, and for drug therapies against diseases
• Covers tools and databases that can be utilized to facilitate in silico methods for drug repurposing

• Language: English
• Publisher: Academic Press
• Release date: Feb 12, 2019
• ISBN: 9780128163771

Book preview

general.

Section 1

Introduction

Chapter 1

Drug Repositioning: New Opportunities for Older Drugs

Vladimir Poroikov; Dmitry Druzhilovskiy    Institute of Biomedical Chemistry, Moscow, Russia

Abstract

Identification of new applications of drugs already launched to the pharmaceutical market, which is called drug repositioning, creates possibilities of finding new therapies for unmet medical needs and discovering a more efficacious treatment. Therapeutic switching also enables the replacement of expensive drugs with their cheap analogs and the substitution of safer medications for those with unwanted side effects. Drug repositioning requires much smaller resources in comparison with de novo drug development. Thus drug-repositioning projects are feasible for academic researchers. Experimental investigation of drug action on multiple pharmacological targets is a laborious task; thus the search for new indications for old drugs is carried out using computational methods. Many well-known examples of drug repositioning originated from occasional findings; however, regular computational screening has been recently applied for this purpose. In this case, the total universe of available biomedical and clinical information may be used for the: detection of unexpected side effects and drug interactions with new targets; creation of new regulatory networks specifying particular signs of a disease; identification of pharmaceutical agents that could affect particular phenotypic manifestations of a disease; and identification of new associative linkages by using text-mining techniques. In this article we have taken an in-depth look at the current state, possibilities, and limitations of further progress in the field of drug repositioning.

Keywords

Drugs; Indications; Off-target applications; Drug repositioning; Chemoinformatics; Chemical-biological interactions

Acknowledgment

The work is supported by the RSF-DST Grant No. 16-45-02012-INT/RUS/RSF/12.

1 Introduction

Recently, drug repositioning (also called drug repurposing, drug reprofiling, drug reformulating or drug redirecting) has gained popularity (Fig. 1). The attractiveness of drug repositioning leads to the organization of different thematic conferences (e.g., 7th Annual Drug Repositioning, Repositioning, and Rescue Conference; https://www.drugrepositioningconference.com/index/), occurrence of specialized scientific journals (e.g., Drug Repositioning, Rescue and Repositioning; https://www.liebertpub.com/loi/DRRR), and development of focused portals (e.g., DRP—Drug Repositioning Portal; http://drugrepositioningportal.com/drug-repositioning-news.php), etc.

Fig. 1 The number of publications on drug repositioning indexed in Scopus and Web of Sciences in 2001–2017. Query: drug repositioning OR drug repositioning OR drug reprofiling OR drug reformulating OR drug redirecting.

According to the US National Comprehensive Cancer Network (NCCN), 50%–75% of drugs have been used off-label in the United States (Jin & Wong, 2014). The National Center for Advancing Translational Sciences (NCATS, NIH) has invested $575 million in drug rescuing and repositioning (Jin & Wong, 2014). The Center for World Health & Medicine (CWHM, NIH) has initiated the creation of a screening platform for the development of drugs for rare/neglected diseases (Huang et al., 2011).

Ashburn and Thor defined drug repositioning as The process of finding new uses outside the scope of the original medical indication for existing drugs (Ashburn & Thor, 2004). In contrast to the discovery of new medicines, drug repositioning is less time-consuming, requires a significantly lower amount of financial expenses and is associated with a reduced risk of unfavorable results (Fig. 2).

Fig. 2 Drug repositioning in comparison with the discovery of new chemical entities and development of me-too-drugs: time, cost, and risk estimates.

A significant reduction in the cost, time, and risk estimates presented in Fig. 2 may be achieved only if a drug is proposed for a repurposed indication in the same formulation, doses, and route of administration as the initial indication. Thus drug repositioning may differ from off-label drug use, which often suggests the utilization of medicine in different doses, particularly in pediatric practice (Wittich, Burkle, & Lanierb, 2012).

Currently, there are several hundred drugs that are utilized in medical practice both for initial and repurposed indications. For example, acetylsalicylic acid (Aspirin) was launched as a nonsteroidal antiinflammatory agent in 1897, and as an antithrombotic agent in 1956; Zidovudine was launched as an anticancer agent in 1964 and as an anti-HIV agent in 1987; etc. Information about many repurposed drugs may be found at the Drug Repositioning Portal (http://drugrepositioningportal.com/drug-repositioning-news.php) as well as in published literature. However, to be sure that a particular drug exhibits the repositioned indication mentioned in a particular publication, it is necessary to carry out an extensive informational search and carefully analyze the results. For instance, Ceftriaxone had been suggested (Yacila & Sari, 2014) to be useful for the treatment of amyotrophic lateral sclerosis, but its efficacy was not confirmed in the subsequent clinical trials (Cudkowicz et al., 2014).

Now, to check the results obtained in drug repositioning projects, it is necessary to analyze the information presented at the US FDA official website (https://www.fda.gov/Drugs/) and clinical trials database (https://www.clinicaltrials.gov/).

Many new indications for old drugs were discovered by serendipity due to some additional pharmacotherapeutic effect observed in clinical or preclinical studies. Let us consider the reasons why the potentially useful pharmacotherapeutic actions of the launched drugs were not identified during their first introduction into medical practice.

2 The Fundamentals for Drug Repositioning

Schematically, the process of new drug discovery may be described as the progressive identification of a pathological condition associated with a particular disease, including the identification of a target, the impact on which can lead to normalization of this pathological condition, and the ligand, which can interact with this target. Such an approach corresponds to the magic bullet concept developed in 1900 by Paul Ehrlich, a German physician and scientist awarded the Nobel Prize (Winau, Westphal, & Winau, 2004).

However, it was shown that usually the magic bullet interacts with several or even many targets modulating their function, which, in some cases, may be favorable for treatment of a particular disorder. Thus the concerted action of pharmaceutical agents on multiple targets was called the magic shotguns (Roth, Sheffler, & Kroeze, 2004). Such multitargeted ligands could be designed intentionally from information about the molecular mechanisms of binding (Morphy, Kay, & Rankovic, 2004).

Molecular biological, biochemical, and cellular studies conclude that the interaction of the ligand with a particular target does not always result in the desirable pharmacotherapeutic effect. Owing to the numerous negative feedbacks, blockade of a specific node in one part of the signal transduction regulatory network may cause the activation of alternative parts of the network, which may reduce or even prevent the desirable pharmacotherapeutic effect (Hornberg, Bruggeman, Westerhoff, & Lankelma, 2006). For instance, loss of expression of ubiquitous transcription factor CREB in the human cancer cell line H295R leads to upregulation of CREMtau, which compensates the CREB deficiency to maintain CRE regulation by cAMP (Groussin, Massias, Bertagna, & Bertherat, 2000).

Taking into account the complexity of modulation of the biological/pathological processes mentioned above, the impact of network pharmacology was recognized as a crucial factor of drug-target-effect relationships (Hopkins, 2007). Thus a relatively simple concept disease-target-ligand (Fig. 3) is replaced by a more complicated paradigm disease-pathway-target-ligand (Fig. 4).

Fig. 3 Simplistic magic bullet concept as the basis for new drugs discovery.

Fig. 4 A network pharmacology paradigm as the basis for new drugs discovery.

Currently, in many cases, the study of the action of drugs begins with the analysis of biological activity in biochemical or cellular tests in vitro; then their specific action must be confirmed in animal experiments. After confirmation of their specific activity in vivo, the safety of lead compounds is investigated in preclinical studies. If the safety and efficacy of particular compounds are substantiated, they become drug-candidates and are further investigated in clinical trials. Preclinical studies of drugs safety are carried out following the good laboratory practice (GLP) standards, and clinical studies are carried out in accordance with the good clinical practice (GCP) standards.

In contrast to the latest stages of the drug research and development process conducted according to the GLP and GCP protocols, the early stages of drug discovery are often performed in exploratory studies using the nonstandardized assays. Taking into account the difference in terms and conditions of experiments, it is sometimes rather difficult to compare the results obtained. Moreover, the conclusion that the compound under study is shown to be inactive in a specific assay may be drawn due to the imperfection of this particular analysis for this ligand-target interaction. Also, the results obtained for one kind of experimental animals may be not transferable to other laboratory species and, finally, to the human. A good example illustrating such a situation is the CS-514 molecule, which was extracted from fungi Penicillium citrinum by Sankyo Pharma in 1970. Since its potent inhibitory activity against hydroxymethylglutaril-CoA (HMG-CoA) reductase was shown in vitro, it was suggested as a promising lead compound for the treatment of hypercholesterolemia and hyperlipidemia. However, CS-514 was found to be inactive during in vivo testing in mice and rats, which might be considered as a reason for termination of the project. Fortunately, the company decided to study CS-514 in hens, and it was found to be active. Then its activity was confirmed in rabbits, dogs and, finally, in humans (Fig. 5). In 1989 pravastatin sodium was registered as a medicine for treatment of familial hypercholesterolemia and hyperlipidemia. In 2005 Pravachol (pravastatin sodium) became a blockbuster in the United States with annual sales of 1.3 billion dollars (Yoshino et al., 1986).

Fig. 5 The history of pravastatin discovery by Sankyo Pharma.

Thus it is necessary to be very cautious in considering a particular drug-like compound as inactive because there are different aspects that might lead to the wrong conclusions (Lipinski & Hopkins, 2004). Taking into account the concerns mentioned above, the concepts presented in Figs. 3 and 4 should be complicated (Fig. 6).

Fig. 6 Relationships disease-pathway-target-ligand should be considered keeping in mind the different experimental terms and conditions (TC).

It is worth reminding ourselves of the statement of Ivan Pavlov, another recipient of the Nobel Prize, made in 1894: On the vast territory of medical knowledge pharmacology seems, one may say the border, where there is a particularly lively exchange of services between the natural scientific basis of medicine, physiology, and medical knowledge - therapy, and where therefore particularly felt the mutual usefulness of one knowledge to another. Pharmacology, studying animal drug action by using physiological methods, improving therapy, puts it on a rational solid ground; on the other hand, the treatment indication, subjected to laboratory analysis, often leads to the discovery of the such physiological phenomena that would remain undetected for a long time with pure physiological study (Pavlov, 1894).

Therefore investigating the biological action of drug-like substances, one should keep in mind several important issues:

•Most of the ligands may exhibit multitargeted action.

•Knowledge about the functioning of biological systems in normal and pathological states is incomplete.

•Any in vitro or in vivo assays/experimental model provides only partial information regarding the pharmacological potential of the compound under study.

Due to continual progress in the development of experimental techniques resulting in a better understanding of the pathological mechanisms and their normalization by ligands-targets-pathway interactions, new useful features of old drugs, which are currently used in medical practice or removed from the market for any reason, may be discovered.

3 Different Approaches to the Development of New Indications for Old Drugs

An overview of the existing approaches to drug repositioning is given in Fig. 7. They include the detection of unexpected side effects, the establishment of drug interactions with new targets, the construction of new regulatory networks specifying particular signs of disease, the identification of pharmaceutical agents that could affect individual phenotypic manifestations of a disease, and the identification of new associative linkages by the application of text-mining. In many cases, two or more of the listed approaches are applied in combination, thus allowing for the exploration of the total universe of available biomedical and clinical information.

Fig. 7 Different approaches to drug repositioning.

Some examples demonstrating the possibilities of the methods mentioned previously are given in the following sections.

3.1 Serendipity and Text Mining

Thalidomide was initially introduced into clinical practice in 1957 as an over-the-counter sedative medicine, particularly recommended for the treatment of morning sickness in pregnant women. It was withdrawn from the market in the early 60s due to its teratogenic action, which caused severe birth defects in thousands of children (Kim & Scialli, 2011). It is necessary to mention that thalidomide safety had been tested in rodents according to the existing standards at that time; however, neither in mice nor rats was a teratogenic effect of the drug identified. Based on this observation, testing of teratogenic action in other species (guinea pig, rabbit) has been added as a requirement to the appropriate protocols for teratogenicity assessment. Despite its harmful effects, FDA approved thalidomide for the treatment of leprosy in 1998 and multiple myeloma in 2003 (https://integrity.clarivate.com/). However, the distribution of thalidomide is regulated by a particular System for Thalidomide Education and Prescribing Safety (S.T.E.P.S.) program. These additional useful applications of thalidomide have been found due to the serendipity.

Text mining based on the so-called ABC Swanson's rule allowed the identification of some new possible pharmacotherapeutic effects of thalidomide. Swanson considered association studies in medical literature as a potential source for new discoveries: If concepts A and B are reported to be related to one set of publications and concepts B and C are reported to be related to another set, then A and C might be indirectly related to each other (Swanson, 1990). Based on this rule, Marc Weeber and coauthors analyzed the available medical literature and concluded that thalidomide might be useful for the treatment of acute pancreatitis, chronic hepatitis C, Helicobacter pylori-induced gastritis, and myasthenia gravis (Weeber et al., 2003). Our informational search in PubMed at least partially supported these hypotheses, e.g., in the case of acute pancreatitis (Lv, Fan, Li, Meng, & Liu, 2015; Lv, Li, Ji, Li, & Fan, 2014), hepatitis C (Milazzo et al., 2006; Pardo-Yules et al., 2011), and myasthenia gravis (Crain, McIntosh, Gordon, Pestronk, & Drachman, 1989).

Successful examples of text mining application for drug repositioning lead to systematic studies in this field (Baker, Ekins, Williams, & Tropsha, 2018; Capuzzi et al., 2018). Alex Tropsha and coauthors developed a web server for mining drug-target-disease relationships (Capuzzi et al., 2018) and applied it for the analysis of over 25 million papers in PubMed (Baker et al., 2018). As a result, they found that more than 60% of drugs or drug-candidates have been studied for two or several diseases. About 200 drugs have been tried in more than 300 diseases, which in the majority of cases were rather close to the initial therapeutic applications; however, some unexpected therapeutic areas have been identified as well (Baker et al., 2018).

3.2 Observation of Unexpected Side Effects

A classic example of drug repositioning based on unexpected side effects is sildenafil (Viagra), which was initially investigated as a remedy for the treatment of a painful heart condition called angina. During the clinical trials, it was found that this medicine could be applied efficiently for the treatment of erectile dysfunction, and it is widely used now for this indication (Osterloh, 2004).

Currently, information about the side effects of various drugs is available from the SIDER resource (http://sideeffects.embl.de/). SIDER 4.1 contains data on 1430 drugs, 5868 side effects, and 139,756 drug-side effect pairs. Weida Tong and coauthors proposed a phenome-guided approach to drug repositioning (Bisgin et al., 2014). The authors assumed that all known phenotypes in the human population are characterized by a combination of indications and side effects. Based on the developed latent Dirichlet allocation (LDA) model the authors were able to predict 70% of pairs of probable significance. For further validation of the LDA model, they carried out the prediction of indications. The approved indications for six drugs not listed in SIDER were predicted successfully. Also, for 908 drugs, some alternative indications were predicted; information from the scientific literature supports some of these findings. The authors came to the conclusion that the phenome can be further analyzed to discover novel associations between the launched drugs and their therapeutic uses.

To provide the possibility for systematic analysis of drug-side effect associations, a particular knowledge base has been created by the review of 119,085,682 MEDLINE sentences and their parse trees (Xu & Wang, 2014). The authors identified 38,871 drug-side effect pairs, most of which have not been mentioned in FDA drug labels. The extracted drug side effects correlated positively with drug targets, metabolism, and indications.

Recently, a novel approach has been proposed based on the analysis of information presented in social media, to identify new indications for drugs used in clinical practice (Nugent, Plachouras, & Leidner, 2016; Rastegar-Mojarad, Liu, & Nambisan, 2016). The authors (Rastegar-Mojarad et al., 2016) developed a dictionary-based system and applied a machine-learning method to analyze several thousand diseases mentioned in 64,616 patients' comments. Ten common patterns used by patients to report any beneficial effects or uses of medication were identified. The conclusion of this preliminary study regarding the potential benefits of social media for drug repositioning was quite obvious (Rastegar-Mojarad et al., 2016).

3.3 Detection of a New Role for the Existing Targets

As an example of such an approach to drug repositioning, let us consider 5 alpha-reductase inhibitor finasteride, which was approved by Merck Sharp & Dohme and Sigma-Tau for the treatment of benign prostatic hyperplasia (BPH) under the trade name Proscar in 1992 (https://integrity.clarivate.com/). In 1998 Merck & Co. launched finasteride under the trade name Propecia for stimulation of hair growth in patients with mild-to-moderate androgenic alopecia (https://integrity.clarivate.com/). 5-Alpha-reductase is a microsomal enzyme that converts testosterone into dihydrotestosterone and progesterone or corticosterone into their 5-alpha-3-oxosteroids. Overproduction of androgens (particularly DHT) is associated with benign prostatic hyperplasia; thus inhibition of this enzyme is useful in the treatment of BPH. On the other hand, overproduction of androgens leads to a hair loss problem, and drugs with an antiandrogen effect like finasteride may be useful for alopecia treatment. It is interesting to note that Propecia contains a fivefold lower dose of finasteride comparing to Proscar, and in the appropriate drug label, it is emphasized that Propecia cannot be used for the treatment of BPH (https://www.fda.gov/Drugs/).

Another example of drug repositioning due to the identification of a new role for the existing targets is crizotinib, a dual inhibitor of hepatocyte growth factor receptor (c-Met/HGFR) kinase and anaplastic lymphoma kinase (ALK) (https://integrity.clarivate.com/). In 2011 this drug was launched for the treatment of patients with ALK-positive advanced or metastatic nonsmall cell lung cancer (NSCLC). In 2016 the FDA and EMA approved crizotinib for the treatment of patients with ROS1-positive advanced NSCLC (https://integrity.clarivate.com/).

In general, to reveal a new role of known targets, which may result in drug repositioning, it is necessary to apply a systems pharmacology approach. This approach is also utilized for the discovery of new pathways associated with a particular disease. More information about systems pharmacology methods for detecting disease-pathway-target relationships may be found in recent reviews (Fotis, Antoranz, Hatziavramidis, Sakellaropoulos, & Alexopoulos, 2018; Kibble et al., 2015; Zhang, Bai, Wang, & Xiao, 2016).

3.4 Identification of New Drug-Target Interactions

As we have already mentioned, most of the known pharmaceuticals exhibit pleiotropic pharmacological effects due to interaction with several or even many targets. There are many millions of previously synthesized drug-like compounds (http://www.chemnavigator.com/), and many more such molecules could be obtained (https://cactus.nci.nih.gov/download/savi_download/). Testing of such a huge number of molecules against many thousands of targets (Oprea et al., 2018) is not a feasible task from both economical and practical points of view. Thus computational estimates become the methods-of-the-choice for launching priorities in drug-target pairs to be tested. Both target-based and ligand-based drug design approaches are used for this purpose (Bezhentsev et al., 2017; Lagunin et al., 2014).

Target-based methods are based on molecular docking, which is applied for inverse virtual screening (Lee, Lee, & Kim, 2016; Luo et al., 2016; Xu, Huang, & Zou, 2018). Some examples demonstrating the potential of this approach for drug repositioning are considered below.

For instance, Vilar and coauthors (Vilar et al., 2016) proposed to calculate the so-called target interaction profile fingerprints (TIPFs), which may be used to generate a hypothesis regarding new putative drug-target interactions. Predicted interaction with monoamine oxidase B of carbonic anhydrase inhibitor ethoxzolamide, which is used for the treatment of glaucoma, was confirmed by experiment (IC50 = 25 μM). Also, several drugs and drug-candidates including lapatinib, SB-202190, RO-316233, GW786460X and indirubin-3′-monoxime were predicted to interact with cyclooxygenase-1. It was shown that SB-202190 and RO-316233 have IC50 values comparable with those for reference drugs diclofenac and indomethacin at the same experimental conditions (24 and 25 μM, respectively). Moderate COX-1 inhibitory activity was shown for lapatinib and indirubin-3′-monoxime as well.

Another example (Hamdoun, Jung, & Efferth, 2017) demonstrated how molecular docking allowed the identification of the probable binding of anthelmintic Niclosamide to the ATP-binding site of glutathione synthetase (calculated scoring function value about − 9.40 kcal/mol). The experiment confirmed this prediction: it was found that the binding constant between niclosamide and recombinant human glutathione synthetase is about 5.64 μM. Anticancer activity of niclosamide was established in the cellular assay, and the drug exhibited potent activity against the multidrug-resistant CEM/ADR5000 leukemia cells.

Despite a successful demonstration of molecular docking potential for detection of new targets of the existing pharmaceuticals, some issues have complicated its broad application for drug repositioning (Chen, 2015; Schomburg & Rarey, 2014). The most substantial limitation is that this method requires knowledge of the 3D structure of the target protein, which is not always available, particularly for membrane-bound proteins or large protein complexes. Second, there is no clear correlation between the calculated scoring function and experimentally determined binding energies. Third, time-consuming calculations are necessary to estimate the possibilities of drug binding to many target proteins. Finally, in the direct comparative study, a better performance of ligand-based methods in estimating biological activity profiles for drug-like molecules was found (Druzhilovskiy et al., 2016).

Ligand-based drug-design methods require data on the structure and activity of drug-like compounds, which may be used for the creation of (Q)SAR or pharmacophore models (Bezhentsev et al., 2017; Lagunin et al., 2014). This information may be obtained from several freely available databases (e.g., PubChem (https://pubchem.ncbi.nlm.nih.gov/), ChEMBL (https://www.ebi.ac.uk/chembl/), DrugBank (https://www.drugbank.ca/), etc.) and commercially available web resources (e.g., Integrity; https://integrity.clarivate.com/). For instance, ChEMBL 23th version contains 2,101,843 records on compounds (1,735,442 unique structures), 11,538 records on targets, and 14,675,320 records on biological activity. This substantial information is to be used as a training set for developing multiple (Q)SAR models, which may be applied for the prediction of the biological activity of drug profiles. These computational tools have been recently described in the literature (Cheng, Zhou, Li, Liu, & Tang, 2012; Hu, Lounkine, & Bajorath, 2014; Huang et al., 2017; Pogodin, Lagunin, Filimonov, & Poroikov, 2015; Shaikh, Sharma, & Garg, 2016) and some of them are freely available via the Internet (e.g., http://prediction.charite.de; http://www.cbligand.org/TargetHunter/; http://potentia.cbs.dtu.dk/ChemProt/; http://www.swisstargetprediction.ch/; http://sea.bkslab.org/).

The first freely available web service that predicted many kinds of biological activity based on a structural formula of the drug-like compound is PASS Online (Lagunin, Stepanchikova, Filimonov, & Poroikov, 2000; http://www.way2drug.com/passonline/). Currently, PASS (prediction of activity spectra for substances) predicts over 4000 kinds of biological activity with an average accuracy of about 95% and is widely used by more than 19,000 researchers from more than 100 countries to optimize the synthesis and biological testing of the studied organic molecules (Filimonov et al., 2014, 2018). In 2001 PASS predictions regarding new pharmacotherapeutic effects of eight from the Top200 drug list were published (Poroikov, Akimov, Shabelnikova, & Filimonov, 2001). It was suggested that sertraline might be used for cocaine dependency treatment amlodipine may be used as an antineoplastic enhancer, ramipril could be used for the treatment of arthritis, and oxaprozin may be applied as an interleukin 1 antagonist. A recent analysis of the published literature confirmed those predictions (Murtazalieva, Druzhilovskiy, Goel, Sastry, & Poroikov, 2017). Special computational experiments were performed to compare the accuracy of the predicted initial and repurposed indications of 50 well-known repositioned drugs and 12 recently patented medicines (Murtazalieva et al., 2017). It was shown that PASS Online performance exceeds those from several other freely available web services predicting biological activity profiles (http://prediction.charite.de; http://www.cbligand.org/TargetHunter/; http://potentia.cbs.dtu.dk/ChemProt/; http://www.swisstargetprediction.ch/; http://sea.bkslab.org/).

3.5 Drug Repositioning for Specific Disease Phenotypes

Recent progress in genomics, transcriptomics, proteomics, metabolomics, and other OMIC sciences resulted in the proposition of the concept of P4 medicine (Galas & Hood, 2009; Hood, Rowen, Galas, & Aitchison, 2008). It is expected that the development of P4 (personalized, predictive, preventive, and participatory) medicine will result in the large-scale integration of complementary skills, technologies, computational tools, patient records, and samples and analysis of societal issues, which will be beneficial for patients, and society in general (Xu, Li, & Wang, 2013).

Since a particular phenotype makes an essential contribution to a disease, the systematic study of phenotypic relationships among human diseases and integration of disease phenotypic data with the existing genetic and ‘omics’ data will allow for elucidation of the disease genetic mechanisms and development of effective drug therapies without requiring detailed knowledge of the exact relationships among genes, which often are not clearly understood (Jadamba & Shin, 2016).

Therefore several approaches have been proposed for the utilization of phenotypic information aimed at drug repositioning (Chen, Gao, Wang, & Xu, 2016; Jadamba & Shin, 2016; Oh et al., 2017; Xu et al., 2013). For instance, Xu and coauthors (Xu et al., 2013) built a disease-phenotype knowledge base by extracting information about disease-manifestation relationships from the literature. Chen and coauthors performed a combined analysis of information about human glioblastoma genomics and mouse glioblastoma phenotypes (Chen et al., 2016). Investigation of FDA-approved drugs for candidates that share similar mouse phenotype profiles with glioblastoma allowed for the prioritization of the existing drugs for the treatment of disease with a particular phenotype.

However, taking into account the considerable variability of disease and patient genotypes and phenotypes, it seems that the identification of compounds modulating specific disease phenotypes is still in its infancy, and many efforts are needed to achieve some clinically useful results.

4 Conclusions

Currently drug repositioning is a very hot topic that is studied by many researchers worldwide. Since drug repositioning requires significantly less time and financial expenses, it provides some possibilities for drug discovery in academia (Oprea et al., 2011; Shamas-Din & Schimmer, 2015; Strovel et al., 2012). In particular, the integration of data on the existing drugs with the predictive computational services on the freely available Drug Repositioning Platform (http://www.way2drug.com/dr) opens up new directions for further research in this field.

In general, investigations on drug repositioning are based on the integration of the existing information on diseases, pathways, targets, and ligands. Considering variability of terms and conditions under which experimental and clinical data are obtained, this task seems to be highly challenging. Therefore significant and concerted community efforts are necessary for efficient usage of the existing biomedical and clinical information and extraction of knowledge from this information, which may help to provide better repositioning of the current drugs. Many studies directed at the annotation, curation, and integration of information about chemical-biological interactions are currently underway (https://pubchem.ncbi.nlm.nih.gov/; https://www.ebi.ac.uk/chembl/; https://www.drugbank.ca/; http://www.way2drug.com/dr; Arvidsson, Sandberg, & Forsberg-Nilsson, 2016; Nguyen et al., 2017; Ostaszewski et al., 2018; Williams et al., 2012). They open new opportunities in the field of drug repositioning, which will, eventually, result in the development of more safe and effective medicines in the future.

References

Arvidsson P.I., Sandberg K., Forsberg-Nilsson K. Open for collaboration: an academic platform for drug discovery and development at SciLifeLab. Drug Discovery Today. 2016;21:1690–1698.

Ashburn T.T., Thor K.B. Drug repositioning: identifying and developing new uses for existing drugs. Nature Reviews Drug Discovery. 2004;3:673–683.

Baker N.C., Ekins S., Williams A.J., Tropsha A. A bibliometric review of drug repositioning. Drug Discovery Today. 2018;23:661–672.

Bezhentsev V.M., Druzhilovskiy D.S., Ivanov S.M., Filimonov D.A., Sastry G.N., Poroikov V.V. Web resources for discovery and development of new medicines. Pharmaceutical Chemistry Journal. 2017;51:91–99.

Bisgin H., Liu Z., Fang H., Kelly R., Xu X., Tong W. A phenome-guided drug repositioning through a latent variable model. BMC Bioinformatics. 2014;15:267.

Capuzzi S.J., Thornton T.E., Liu K., Baker N., Lam W.I., O'Banion C.P.,… Tropsha A. Chemotext: a publicly available web server for mining drug-target-disease relationships in PubMed. Journal of Chemical Information and Modeling. 2018;58:212–218.

Chen Y.C. Beware of docking!. Trends in Pharmacological Sciences. 2015;36:78–95.

Chen Y., Gao Z., Wang B., Xu R. Towards precision medicine-based therapies for glioblastoma: interrogating human disease genomics and mouse phenotypes. BMC Genomics. 2016;17(Suppl. 7):516.

Cheng F., Zhou Y., Li W., Liu G., Tang Y. Prediction of chemical-protein interactions network with weighted network-based inference method. PLoS One. 2012;7:e41064.

Crain E., McIntosh K.R., Gordon G., Pestronk A., Drachman D.B. The effect of thalidomide on experimental autoimmune myasthenia gravis. Journal of Autoimmunity. 1989;2:197–202.

Cudkowicz M.E., Titus S., Kearney M., Yu H., Sherman A., Schoenfeld D.… On behalf of the Ceftriaxone Study Investigators. Efficacy and safety of ceftriaxone for amyotrophic lateral sclerosis: results of a multi-stage, randomised, double-blind, placebo-controlled, phase 3 study. Lancet Neurology. 2014;13:1083–1091.

Druzhilovskiy D.S., Rudik A.V., Filimonov D.A., Lagunin A.A., Gloriozova T.A., Poroikov V.V. Online resources for the prediction of biological activity of organic compounds. Russian Chemical Bulletin, International Edition. 2016;65:384–393.

Filimonov D.A., Druzhilovskiy D.S., Lagunin A.A., Gloriozova T.A., Rudik A.V., Dmitriev A.V.,… Poroikov V.V. Computer-aided prediction of biological activity spectra for chemical compounds: opportunities and limitations. Biomedical Chemistry: Research and Methods. 2018;1:e00004.

Filimonov D.A., Lagunin A.A., Gloriozova T.A., Rudik A.V., Druzhilovskiy D.S., Pogodin P.V., Poroikov V.V. Prediction of the biological activity spectra of organic compounds using the PASS online web resource. Chemistry of Heterocyclic Compounds. 2014;50:444–457.

Fotis C., Antoranz A., Hatziavramidis D., Sakellaropoulos T., Alexopoulos L.G. Network-based technologies for early drug discovery. Drug Discovery Today. 2018;23:626–635.

Galas D.J., Hood L. Systems biology and emerging technologies will catalyze the transition from reactive medicine to predictive, personalized. preventive and participatory (P4) medicine. Interdisciplinary Bio Central. 2009;1:1–4.

Groussin L., Massias J.F., Bertagna X., Bertherat J. Loss of expression of the ubiquitous transcription factor cAMP response element-binding protein (CREB) and compensatory overexpression of the activator CREMtau in the human adrenocortical cancer cell line H295R. The Journal of Clinical Endocrinology & Metabolism. 2000;85:345–354.

Hamdoun S., Jung P., Efferth T. Drug repositioning of the anthelmintic niclosamide to treat multidrug-resistant leukemia. Frontiers in Pharmacology. 2017;8:110.

Hood L., Rowen L., Galas D.J., Aitchison J.D. Systems biology at the Institute for Systems Biology. Briefings in Functional Genomics & Proteomics. 2008;7:239–248.

Hopkins A.L. Network pharmacology. Nature Biotechnology. 2007;25:1110–1111.

Hornberg J.J., Bruggeman F.J., Westerhoff H.V., Lankelma J. Cancer: a systems biology disease. Biosystems. 2006;83:81–90.

Hu Y., Lounkine E., Bajorath J. Many approved drugs have bioactive analogs with different target annotations. The AAPS Journal. 2014;16:847–859.

Huang T., Mi H., Lin C.Y., Zhao L., Zhong L.L., Liu F.B.… For MZRW Group. MOST: most-similar ligand based approach to target prediction. BMC Bioinformatics. 2017;18:165.

Huang R., Southall N., Wang Y., Yasgar A., Shinn P., Jadhav A.,… Austin C.P. The NCGC pharmaceutical collection: a comprehensive resource of clinically approved drugs enabling repositioning and chemical genomics. Science Translational Medicine. 2011;27:80ps16.

Jadamba E., Shin M. A Systematic framework for drug repositioning from integrated omics and drug phenotype profiles using pathway-drug network. Biomedical Research International. 2016;2016:7147039.

Jin G., Wong S.T. Toward better drug repositioning: prioritizing and integrating existing methods into efficient pipelines. Drug Discovery Today. 2014;19:637–644.

Kibble M., Saarinen N., Tang J., Wennerberg K., Mäkelä S., Aittokallio T. Network pharmacology applications to map the unexplored target space and therapeutic potential of natural products. Natural Product Reports. 2015;32:1249–1266.

Kim J.H., Scialli A.R. Thalidomide: the tragedy of birth defects and the effective treatment of disease. Toxicological Sciences. 2011;122:1–6.

Lagunin A.A., Goel R.K., Gawande D.Y., Priynka P., Gloriozova T.A., Dmitriev A.V.,… Poroikov V.V. Chemo- and bioinformatics resources for in silico drug discovery from medicinal plants beyond their traditional use: a critical review. Natural Product Reports. 2014;31:1585–1611.

Lagunin A., Stepanchikova A., Filimonov D., Poroikov V. PASS: prediction of activity spectra for biologically active substances. Bioinformatics. 2000;16:747–748.

Lee A., Lee K., Kim D. Using reverse docking for target identification and its applications for drug discovery. Expert Opinions in Drug Discovery. 2016;11:707–715.

Lipinski C., Hopkins A. Navigating chemical space for biology and medicine. Nature. 2004;432:855–861.

Luo H., Zhang P., Cao X.H., Du D., Ye H., Huang H.,… Yang L. DPDR-CPI, a server that predicts drug positioning and drug repositioning via chemical-protein interactome. Scientific Reports. 2016;6:35996.

Lv P., Fan L.J., Li H.Y., Meng Q.S., Liu J. Protective effect of thalidomide on liver injury in rats with acute pancreatitis via inhibition of oxidative stress. Annals of Clinical and Laboratory Science. 2015;45:508–514.

Lv P., Li H.Y., Ji S.S., Li W., Fan L.J. Thalidomide alleviates acute pancreatitis-associated lung injury via down-regulation of NFκB induced TNF-α. Pathology, Research and Practice. 2014;210:558–564.

Milazzo L., Biasin M., Gatti N., Piacentini L., Niero F., Zanone Poma B.,… Riva A. Thalidomide in the treatment of chronic hepatitis C unresponsive to alfa-interferon and ribavirin. American Journal of Gastroenterology. 2006;101:399–402.

Morphy R., Kay C., Rankovic Z. From magic bullets to designed multiple ligands. Drug Discovery Today. 2004;9:641–651.

Murtazalieva K.A., Druzhilovskiy D.S., Goel R.K., Sastry G.N., Poroikov V.V. How good are publicly available web services that predict bioactivity profiles for drug repositioning?. SAR and QSAR in Environmental Research. 2017;28:843–862.

Nguyen D.T., Mathias S., Bologa C., Brunak S., Fernandez N., Gaulton A.… Oprea, T.I.& Guha, R. PHAROS: collating protein information to shed light on the druggable genome. Nucleic Acids Research. 2017;45(D1):D995–D1002.

Nugent T., Plachouras V., Leidner J.L. Computational drug repositioning based on side-effects mined from social media. Peer Journal of Computer Science. 2016;2:e46.

Oh M., Ahn J., Lee T., Jang G., Park C., Yoon Y. Drug voyager: a computational platform for exploring unintended drug action. BMC Bioinformatics. 2017;18:131.

Oprea T.I., Bauman J.E., Bologa C.G., Buranda T., Chigaev A., Edwards B.S.,… Sklar L.A. Drug repositioning from an academic perspective. Drug Discovery Today: Therapeutic Strategies. 2011;8:61–69.

Oprea T.I., Bologa C.G., Brunak S., Campbell A., Gan G.N., Gaulton A.,… Zahoránszky-Köhalmi G. Unexplored therapeutic opportunities in the human genome. Nature Reviews Drug Discovery. 2018;17:317–332.

Ostaszewski M., Gebel S., Kuperstein I., Mazein A., Zinovyev A., Dogrusoz U.,… Schneiderm R. Community-driven roadmap for integrated disease maps. Briefings in Bioinformatics. 2018;23:bby024.

Osterloh I.H. The discovery and development of Viagra® (sildenafil citrate). In: Dunzendorfer U., ed. Sildenafil. Milestones in drug therapy MDT. Basel: Birkhäuser; 2004.

Pardo-Yules B., Gallego-Durán R., Eslam M., García-Collado C., Grande L., Paradas C.,… Romero-Gómez M. Thalidomide with peginterferon alfa-2b and ribavirin in the treatment of non-responders genotype 1 chronic hepatitis C patients: proof of concept. Revista Española de Enfermedades Digestivas. 2011;103:619–625.

Pavlov I.P. On the incompleteness of the modern physiological analysis of drugs action. Saint Petersburg: Typography of the Ministry of Railways; 1894.

Pogodin P.V., Lagunin A.A., Filimonov D.A., Poroikov V.V. PASS targets: ligand-based multi-target computational system based on public data and naïve bayes approach. SAR and QSAR in Environmental Research. 2015;26:783–793.

Poroikov V., Akimov D., Shabelnikova E., Filimonov D. Top 200 medicines: can new actions be discovered through computer-aided prediction?. SAR and QSAR in Environmental Research. 2001;12:327–344.

Rastegar-Mojarad M., Liu H., Nambisan P. Using social media data to identify potential candidates for drug repositioning: a feasibility study. JMIR Research Protocols. 2016;5:e121.

Roth B.L., Sheffler D.J., Kroeze W.K. Magic shotguns versus magic bullets: selectively non-selective drugs for mood disorders and schizophrenia. Nature Reviews Drug Discovery. 2004;3:353–359.

Schomburg K.T., Rarey M. What is the potential of structure-based target prediction methods?. Future Medicinal Chemistry. 2014;6:1987–1989.

Shaikh N., Sharma M., Garg P. An improved approach for predicting drug-target interaction: proteochemometrics to molecular docking. Molecular BioSystems. 2016;12:1006–1014.

Shamas-Din A., Schimmer A.D. Drug discovery in academia. Experimental Hematology. 2015;43:713–717.

Strovel J., Sittampalamm S., Coussens N.P., Hughes M., Inglese J., Kurtz A.,… Weir S. Early drug discovery and development guidelines: For academic researchers, collaborators, and start-up companies. Retrieved July 7, 2018, from: http://www.ncbi.nlm.nih.gov/books/NBK92015/. 2012.

Swanson D.R. Medical literature as a potential source of new knowledge. Bulletin of the Medical Library Association. 1990;78:29–37.

Vilar S., Quezada E., Uriarte E., Costanzi S., Borges F., Viña D., Hripcsak G. Computational drug target screening through protein interaction profiles. Scientific Reports. 2016;6:36969.

Weeber M., Vos R., Klein H., De Jong-Van Den Berg L.T., Aronson A.R., Molema G. Generating hypotheses by discovering implicit associations in the literature: a case report of a search for new potential therapeutic uses for thalidomide. Journal of the American Medical Informatics Association. 2003;10:252–259.

Williams A.J., Harland L., Groth P., Pettifer S., Chichester C., Willighagen E.L.,… Mons B. Open PHACTS: semantic interoperability for drug discovery. Drug Discovery Today. 2012;17:1188–1198.

Winau F., Westphal O., Winau R. Paul Ehrlich—in search of the magic bullet. Microbes and Infection. 2004;6:786–789.

Wittich C.M., Burkle C.M., Lanierb W.L. Ten common questions (and their answers) about off-label drug use. Mayo Clinic Proceedings. 2012;87:982–990.

Xu X., Huang M., Zou X. Docking-based inverse virtual screening: methods, applications, and challenges. Biophysical Reports. 2018;4:1–16.

Xu R., Li L., Wang Q. Towards building a disease-phenotype knowledge base: extracting disease-manifestation relationship from literature. Bioinformatics. 2013;29:2186–2194.

Xu R., Wang Q. Automatic construction of a large-scale and accurate drug-side-effect association knowledge base from biomedical literature. Journal of Biomedical Informatics. 2014;51:191–199.

Yacila G., Sari Y. Potential therapeutic drugs and methods for the treatment of amyotrophic lateral sclerosis. Current Medicinal Chemistry. 2014;21:3583–3593.

Yoshino G., Kazumi T., Kasama T., Iwatani I., Iwai M., Inui A.,… Baba S. Effect of CS-514, an inhibitor of 3-hydroxy-3-methylglutaryl coenzyme A reductase, on lipoprotein and apolipoprotein in plasma of hypercholesterolemic diabetics. Diabetes Research and Clinical Practice. 1986;2:179–181.

Zhang W., Bai Y., Wang Y., Xiao W. Polypharmacology in drug discovery: a review from systems pharmacology perspective. Current Pharmaceutical Design. 2016;22:3171–3181.

Chapter 2

Computational Drug Design Methods—Current and Future Perspectives

Fernando D. Prieto-Martínez⁎; Edgar López-López†; K. Eurídice Juárez-Mercado⁎; José L. Medina-Franco⁎    ⁎ Department of Pharmacy, School of Chemistry, National Autonomous University of Mexico, Mexico City, Mexico

† Medicinal Chemistry Laboratory, University of Veracruz, Veracruz, Mexico

Abstract

Computer-aided drug design (CADD) comprises a broad range of theoretical and computational approaches that are part of modern drug discovery. CADD methods have made key contributions to the development of drugs that are in clinical use or in clinical trials. Such methods have emerged and evolved along with experimental approaches used in drug design. In this chapter we discuss the major CADD methods and examples of recent applications to drugs that have advanced in clinical trials or that have been approved for clinical use. We also comment on representative trends in current drug discovery that are shaping the development of novel methods, such as computer-aided drug repurposing. Similarly we present emerging concepts and technologies in molecular modeling and chemoinformatics. Furthermore, this chapter discusses the authors’ point of view of the challenges of traditional and novel CADD methods to increase their positive impact in drug discovery.

Keywords

Artificial intelligence; Big data; Chemical space; Chemoinformatics; Deep learning; Molecular modeling; Polypharmacology; SmART; Target fishing; Virtual screening

Acknowledgments

This work was supported by the Programa de Apoyo a Proyectos para la Innovación y Mejoramiento de la Enseñanza (PAPIME) grant PE200118, Programa de Apoyo a Proyectos de Investigación e Innovación Tecnológica (PAPIIT) grant IA203718 and National Council of Science and Technology (CONACyT), Mexico grant number 282785. FD P-M acknowledges the PhD scholarship from CONACyT no. 660465/576637. KE J-M thanks CONACyT for the support (No. 16231).

1 Introduction

The generation of, search for, and experimental evaluation of new molecules with increased potency and selectivity is booming. This is possible with the aid of techniques such as combinatorial chemistry and high-throughput screening (HTS). However, these techniques also generate a large number of false positives, so even with this evaluation capacity it is necessary to further filter screening data sets. Computer-aided drug design (CADD) is a discipline that collects multiple chemical-molecular and quantum strategies with the aim of discovering, designing, and developing therapeutic chemical agents. Many CADD approaches are based on structure-activity relationships (SAR). The main objectives of CADD are part of a multidisciplinary work for the improvement of bioactive molecules, the development of therapeutic alternatives, and the understanding of biological events at the molecular level.

In general, the drug-discovery process includes three key stages: (1) the discovery phase, in which the goal is the identification of relevant molecular targets and active molecules or hits; (2) the development phase, where the compounds are evaluated using in vitro and in vivo models (this phase includes various stages: preclinical, clinical I, II, and III); and (3) the registry phase that will enable distribution on the market and the clinical use of drugs. Recent estimates indicate that the average cost of the preclinical phase is 3.4 million dollars, increasing to 8.6 and 21.4 million dollars respectively in clinical phases II and III (Martin, Hutchens, Hawkins, & Radnov, 2017).

The role of CADD in the drug-discovery process lies mainly in the discovery phase, where a primary goal is to filter and select compounds for experimental synthesis or testing. It is expected that this filtering will reduce the time and costs involved in drug development. In addition, CADD enables the possibility of systematically identifying novel potential uses for drugs already approved for other indications. As discussed later in this chapter and this book, this strategy is called drug repurposing. Computational calculations have played a significant role in the investigation of molecules that are currently in clinical use. For example, CADD has made notable contributions to the treatment of acquired immunodeficiency syndrome, influenza virus infections, glaucoma, and patients with nonsmall-cell lung cancer (Medina-Franco, Martínez-Mayorga, Juárez-Gordiano, & Castillo, 2007; Talele, Khedkar, & Rigby, 2010).

With new technological advances and the application of CADD techniques it is possible to solve complex problems in the pharmaceutical area. Recent related reviews have been published (Prieto-Martínez & Medina-Franco, 2018a, 2018b; Talevi, 2018). For instance, Saldívar-González, Prieto-Martínez, and Medina-Franco (2017) commented on the need to further increase these technologies and augment the multidisciplinary investigation of new drugs. Other reviews highlight that CADD represents a systematic manner to merge basic and applied science (Das, 2017). For authors such as Usha, Shanmugarajan, Goyal, Kumar, and Middha (2018), CADD includes a collection of pharmacological, pharmacodynamic, and in silico toxicity predictions, which are useful to identify or filter out active or toxic molecules, respectively.

Currently, the development of new computational techniques is allowing a more comprehensive and detailed study of compounds of clinical interest, such as the application of artificial intelligence, big data, chemical space, chemoinformatics, deep learning, molecular modeling, polypharmacology, structure multiple-activity relationships (SmART), target fishing, and virtual screening. These concepts are further commented on the sections later in this chapter.

The goal of this chapter is to provide a general introduction of CADD, covering their principal methods, recent successful applications in the development of compounds that currently on the market, and major challenges. The chapter is organized in six major sections. After this introduction, an overview of current methods used in CADD and their major applications is presented. Theoretical frameworks in CADD are covered briefly. Section 3 mentions recent successful examples of CADD. Section 4 discusses trending concepts and topics in the area. Section 5 covers the major challenges involved in the development and application of computational approaches. Section 6 presents summary conclusions.

2 Overview of Current Approaches Used in Computer-Aided Drug Design

During the past 30 years the increase in computational power and the availability of chemogenomic data have allowed computational chemistry methods to become an indispensable part in drug discovery. To date, several marketed drugs, for example, imatinib, zanamivir, and nelfinavir, and several clinical candidates, have been discovered or optimized with the aid of molecular modeling techniques. Fig. 1 outlines a summary of the CADD process with concepts and methods discussed in this chapter.

Fig. 1 Schematic overview of a representative computer-aided drug design process.

The concept of big data impacts our everyday life, and the area of CADD is not an exception. Through current computational processors it is possible to collect, evaluate, and analyze molecular characteristics in a massive, systematic, and logical manner.

One can make use of the data of each compound to analyze them from different perspectives. In this sense one of the key questions to ask would be: What do I want to analyze? Theoretical chemistry, chemoinformatics, and machine learning (Varnek & Baskin, 2011) provide methods to guide the answers to this question. Using the principles of each major discipline, we can evaluate nearly any kind of similarity between molecules. It is through these disciplines that chemistry, biology, and pharmaceutical sciences converge.

For drug-discovery applications, homology modeling makes use of structural information available to generate 3D models of biological targets that have not been crystallized (Sliwoski, Kothiwale, Meiler, & Lowe, 2014). Thus homology modeling is a useful tool to explore and guide, for instance, the structure-based design of novel therapeutic targets or difficult to crystallize, as is the case of calcium channels and some epigenetic and protein complexes.

Molecular docking is one of the most used techniques to study 3D ligand-target interactions. Comprehensive reviews of docking have been published recently. One of the main purposes of docking is the generation of models that reveal the possible conformations and, thereafter, evaluate which of them are energetically more viable. However, its application is not limited to characterizing ligand-interactions to increase the potency of active molecules. Docking is also valuable to evaluate specificity and drug resistance, based on 3D structure-property relationships (Pagadala, Syed, & Tuszynski, 2017).

De novo design is another group of CADD techniques. De novo design can be roughly compared to a puzzle, where atoms or small fragments are fitted into the 3D structure of a binding site. After which these small fragments need to be connected through linkers. One of the challenges may be to find the suitable linkers that can allow molecules to be synthesized in the laboratory. Therefore key questions are: How can we assemble the candidate compounds? How do we evaluate their potential quality? How do we sample the search space effectively? A comprehensive review of de novo design has been covered in the literature (Schneider & Fechner, 2005).

With the advent of technological alternatives for the massive evaluation of compounds or fragments, it has become possible to emphasize the identification of structurally simple hits that can be optimized and generate more powerful ligands. Fragment-based screening is based on the fact that a relatively small number of fragments can represent a large fraction of the chemical space. However, for various technical reasons, including the low affinity of the hits fragments and the biophysical methods used for their discovery (e.g., nuclear magnetic resonance, surface plasmon resonance, isothermal calorimetry), fragment-based screening has been limited mainly to in vitro tests with purified proteins (Parker et al., 2017).

It is noteworthy that several CADD techniques make large simplifications of the systems and assume, for the sake of speed, that macromolecules are rigid. A classic example is rigid docking. However, sometimes it is compulsory to consider the dynamics in a model, for example, the binding of small compounds to highly flexibly targets or the simulation of the binding of two macromolecules. To this end molecular dynamic techniques are employed that focus on the use of statistical mechanics, quantum chemistry, and the properties of the electric potential (force field) (Ganesan, Coote, & Barakat, 2017).

The recovery and analysis of chemical information for any type of application in the physical or biological sciences enter the spectrum of chemoinformatics, as well as the relevant computational approaches for the maximized exploration of pharmaceutically relevant compounds. Some of its most used chemoinformatic approaches are quantitative structure-activity relationship (QSAR) and molecular similarity methods (Leach & Gillet, 2007). Overall, QSAR approaches allow the improvement of the pharmacological characteristics of a certain scaffold (basic structure), mediating the determination of the key interactions for a given target, that is, it refines the conformational, spatial, and electronic characteristics of a series of compounds.

Molecular similarity is, in principle, a method simpler than QSAR. It is founded in the premise that similar compounds will have a similar activity. One of the main applications of similarity searching is filtering compounds from existing databases, such as a database of compounds approved by the Food and Drug Administration (FDA). Similarity searching of approved drugs is a technique in drug repurposing (Bajorath, 2017).

2.1 Classification of Computer-Aided Drug Design Methods

As described in this chapter, CADD includes multiple approaches to answer questions of a biological-pharmaceutical nature. In general, CADD methods can be classified in three major groups: structure-based, ligand-based, and hybrid methods.

2.1.1 Structure-Based Methods

Structure-based methods depend on the 3D information of the molecular target. Prominent examples of these methods are docking and molecular dynamics (MD) (see also Section 4.5). Applications of structure-based methods include characterization of binding sites, elucidation of the mechanism of action of active molecules at the molecular level, and assessment of the kinetics and thermodynamics involved in ligand-target recognition (Śledź & Caflisch, 2018).

2.1.2 Ligand-Based Methods

Ligand-based methods are based on the information of the chemical structures of a set of ligands with known biological activity. One of the main goals of these methods is identifying bioactive compounds or improving the activity of active molecules. Typical examples of ligand-based methods are similarity searching and QSAR modeling (Siju et al., 2017).

2.1.3 Hybrid Methods and Methods Based on End-Points

When the structure of the target is known as well as the structure of active molecules, it is possible to use hybrid or combined methods, i.e., a combination of structure-based and ligand-based methods. An example is certain methods of pharmacophore modeling. Other examples are in silico approaches to predict bioactivity based on the biological profile of compounds tested vs. one or multiple targets (Yongye & Medina-Franco, 2012).

2.2 Main Applications of Computer-Aided Drug Design

CADD has two major applications: identify novel potential active compounds, for example, hit identification, and optimize the bioactivity or ADMETox profile of active compounds, for example, assist in hit-to-lead process.

2.2.1 Hit Finding

A common general CADD approach to identify hit compounds is virtual screening. This technique can be compared to a filtering process: starting from a usually large number of compounds, structure, ligand, or hybrid approaches are used to select a reduced number of molecules (Siju et al., 2017). The working hypothesis is that the reduced number of compounds have increased probabilities to be active. Of course, experimental validation of selected compounds is mandatory. After the experimental evaluation has been conducted a second round, or more, of filtering steps is performed. In the second, or more, iteration the experimental information of the previous iteration should be considered in the selection of the molecules. Virtual screening is also applicable to filter potential biological targets for a given small molecule. The later process is called inverse virtual screening or target fishing (Yuriev, Holien, & Ramsland, 2015).

2.2.2 Lead Optimization

A number of structure-, ligand-, or hybrid-based methods can be used to improve the potency or reduce side effects of active molecules. Notably, issues with absorption, distribution, metabolism, and excretion (ADME) properties may hamper the development of compounds in the clinic. QSAR and machine learning approaches (see Section 4.4 in this chapter) have been employed to address not only ADME but also toxicity and potency (Caldwell & Yan, 2014).

3 Case Studies: Successful Applications of Computer-Aided Drug Design

Nowadays, chemoinformatics and molecular modeling methods are useful in several scientific areas. These approaches are becoming key components in the development of new drugs. Despite the fact the computational results applied to pharmaceutical and medicinal chemistry problems are not 100% accurate, CADD represents an efficient way to help save time and costs as compared to using only experimental approaches. Often two or more methods are used in research projects. This is because CADD complement each other, helping to predict more efficiently active compounds. Table 1 summarizes representative CADD based on 3D structures. The table includes a brief description of their common use with actual approaches in the chemical-biological area. Table 2 summarizes two of the more common methods of CADD based on 2D structures with an example of a recent application.

Table 1

Table 2

Several chemoinformatics and CADD methods in general, as exemplified in Tables 1 and 2 (and other sections of this chapter), are being employed to develop drugs that are currently on the market. Some examples are oxymorphone, saqunavir, zanamivir, dorzolamide, and norfloxacin. Table 3 summarizes the information on these drugs and their chemical structures are shown in Fig. 2. Of note, molecular similarity is not stated in Table 3 because this method was employed during the first stages of the design, therefore it is not considered as the principle method for drug discovery. However, molecular similarity principles are commonly used in CADD.

Table 3

Fig. 2 Chemical structures of approved drugs developed with computer-aided drug design.

4 Trending Concepts and Technologies

The landscape of drug discovery and development is changing constantly. In this section, we aim to present a concise yet substantial picture of the current trends and emerging concepts on CADD. Table 4 summarizes the main concepts discussed hereafter.

Table 4

4.1 Big Data

The term big data has been mystified. Nowadays in the era of information and the Internet, the quantity of data generated has increased exponentially. Recently it was estimated that the volume of total stored data stacks to nearly two zettabytes, with projections of this figure doubling every 2 years (Akoka, Comyn-Wattiau, & Laoufi, 2017). Therefore mining this information offers a myriad of possibilities to enhance competitivity and productivity. But to effectively use big data, one must dive directly in and not just dip one's toes in the shores of this vast ocean. The problem with big data is not just its volume, but its complexity, leading to debate as to the true role and efficiency of statistical thinking in such an arena (Secchi, 2018). Additionally, it often happens that such broad data contains errors, duplicates, and missing values due to overcollection. Hence, preprocessing and curation of data are mandatory to correctly assess the quality of information and avoid any potential bias (Cox, Kartsonaki, & Keogh, 2018). However, data curation may be done differently by different research groups based on experience or previous reports.

The need for a unified or canonical protocol for data curation arises to justify the quality of any given study. Several studies have questioned the role of quality in research due to a major focus on impact (Bornmann, 2012), with others suggesting guidelines towards an objective assessment of the true impact and quality of a given work (Mårtensson, Fors, Wallin, Zander, & Nilsson, 2016). Thus the current paradigm on research stays with a less-is-more approach.

Big data has always played a significant role in medicinal chemistry, sometimes indirectly. Methods like combinatorial chemistry and HTS produce large amounts of data over brief periods of time. Previously this was seen like a new dawn for medicinal chemistry, i.e., the ability to process large amounts of new data could reduce the time invested in the drug-development cycle. Consider the following example: human immunodeficiency virus (HIV), a global pandemic for almost 40 years with an estimated of 37 million people infected, and only 57% of patients receiving antiviral therapies (WHO, 2018). Over the years several studies have focused on the inhibition of the viral reverse transcriptase and/or integrase (Cabrera, Hernández, Chávez, & Medina-Franco, 2018; Ghosh, Osswald, & Prato, 2016). While this approach has proven effective enough, it still has several drawbacks, such as viral resistance and poor bioavailability. During the 1990s, studies on the HIV's entry mechanisms showed the role of CD4+ cells and CCR5 chemokine. Chemokines activity is related to their G-coupled receptors (GPCRs), in CCR5’s case it is a C-C receptor with 75% homology to CCR2 (Barmania & Pepper, 2013). Once CCR5 had been stablished as an interesting and druggable novel target to tackle HIV, several