Article The Methodological Trends of Traditional Herbal Medicine Employing Network

Won-Yung Lee 1, Choong-Yeol Lee 1, Youn-Sub Kim 2,* and Chang-Eop Kim 1,*

1 Department of Physiology, College of Korean Medicine, Gachon University, Seongnam 13120, Korea 2 Department of Anatomy-Pointology, College of Korean Medicine, Gachon University, Seongnam 13120, Korea * Correspondence: [email protected] (Y.-S.K.); [email protected] (C.-E.K.); Tel.: +82-31-750-5416 (Y.-S.K.); +82-31-750-5416 (C.-E.K.)  Received: 18 June 2019; Accepted: 6 August 2019; Published: 13 August 2019 

Abstract: Natural products, including traditional herbal medicine (THM), are known to exert their therapeutic effects by acting on multiple targets, so researchers have employed network pharmacology methods to decipher the potential mechanisms of THM. To conduct THM-network pharmacology (THM-NP) studies, researchers have employed different tools and databases for constructing and analyzing herb–compound–target networks. In this study, we attempted to capture the methodological trends in THM-NP research. We identified the tools and databases employed to conduct THM-NP studies and visualized their combinatorial patterns. We also constructed co-author and affiliation networks to further understand how the methodologies are employed among researchers. The results showed that the number of THM-NP studies and employed databases/tools have been dramatically increased in the last decade, and there are characteristic patterns in combining methods of each analysis step in THM-NP studies. Overall, the Traditional Chinese Medicine Database and Analysis Platform (TCMSP) was the most frequently employed network pharmacology database in THM-NP studies. Among the processes involved in THM-NP research, the methodology for constructing a compound–target network has shown the greatest change over time. In summary, our analysis describes comprehensive methodological trends and current ideas in research design for network pharmacology researchers.

Keywords: network pharmacology; traditional herbal medicine; methodological trend

1. Introduction Traditional herbal medicine (THM) has maintained the health of Asian people for thousands of years and built a unique medical system based on empirically accumulated knowledge. Billions of people around the world are taking THM daily, and the drug development field considers THM to be a source of inspiration [1,2]. The research indicates that THM’s therapeutic effects may involve various biomolecules [3]. However, due to the complexity of THM and the limitations of experimental applications, specific mechanisms of action have been fully elucidated for only a few THMs [4]. This is a major obstacle to THM’s modernization and wider application to modern healthcare. Network pharmacology has emerged as a promising approach to accelerate drug development and elucidate the mechanisms of action of multiple target components [5]. It understands disease as a perturbation of interconnected complex biological networks and identifies the mechanisms of drug action by [6,7]. The conceptual elements of network pharmacology were derived from , which can address both the connectivity and the interdependence of individual components [8]. The core idea of network pharmacology is well suited for analyzing the

Biomolecules 2019, 9, 362; doi:10.3390/biom9080362 www.mdpi.com/journal/biomolecules Biomolecules 2019, 9, 362 2 of 15

Biomolecules 2019, 9, x FOR PEER REVIEW 2 of 15 multi-targeted agents, so network pharmacology methods may be appropriate for identifying the multi-targeted agents, so network pharmacology methods may be appropriate for identifying the complex mechanisms of THM. complex mechanisms of THM. In the last decade, researchers have employed network pharmacology methods to elucidate In the last decade, researchers have employed network pharmacology methods to elucidate the the potential targets and toxicity of THMs [9]. THM-network pharmacology (THM-NP) studies are potential targets and toxicity of THMs [9]. THM-network pharmacology (THM-NP) studies are conducted by constructing an herb–compound–target (H-C-T) network by integrating the herbal conducted by constructing an herb–compound–target (H-C-T) network by integrating the herbal constituent data and drug-target interactions (DTIs) information. Then, the target network is analyzed constituent data and drug-target interactions (DTIs) information. Then, the target network is to interpret related biological functions, pathways, and diseases (Figure1). Since there are no gold analyzed to interpret related biological functions, pathways, and diseases (Figure 1). Since there are standard methods for THM-NP studies yet, researchers have developed and applied various tools and no gold standard methods for THM-NP studies yet, researchers have developed and applied various databases for each step. tools and databases for each step.

Figure 1. The general framework of network pharmacology analysis of herbal medicine. Figure 1. The general framework of network pharmacology analysis of herbal medicine. Several studies have described THM-NP researches by summarizing network pharmacology Several studies have described THM-NP researches by summarizing network pharmacology databases for THM and illustrating several representative applications [10–15]. Although these studies databases for THM and illustrating several representative applications [10–15]. Although these contributed to a better understanding of THM-NP studies, they were limited in providing quantitative studies contributed to a better understanding of THM-NP studies, they were limited in providing information on the frequencies, variations, and combinatorial patterns of the employed methods in quantitative information on the frequencies, variations, and combinatorial patterns of the employed THM-NP studies. In this study, we systematically attempted to capture the methodological trends methods in THM-NP studies. In this study, we systematically attempted to capture the of the THM-NP research field. We identified the tools and databases employed to conduct THM-NP methodological trends of the THM-NP research field. We identified the tools and databases studies and visualized their frequency, variation, and combinatorial pattern in a step-by-step manner. employed to conduct THM-NP studies and visualized their frequency, variation, and combinatorial The THM-NP studies were identified by searching PubMed and then preprocessed. We also constructed pattern in a step-by-step manner. The THM-NP studies were identified by searching PubMed and and analyzed the co-author and affiliation networks to identify how the diverse methods for THM-NP then preprocessed. We also constructed and analyzed the co-author and affiliation networks to studies are employed and shared among researchers in the field. We believe that analyzing the identify how the diverse methods for THM-NP studies are employed and shared among researchers methodological trends will provide a comprehensive understanding and valuable insights into the in the field. We believe that analyzing the methodological trends will provide a comprehensive THM-NP research field. understanding and valuable insights into the THM-NP research field. Biomolecules 2019, 9, 362 3 of 15

2. Materials and Methods

2.1. Search Strategy The literature search was performed in PubMed (https://www.ncbi.nlm.nih.gov/pubmed/) from January 2000 to December 2018. The search language was restricted to English. The search terms used were NP-related terms (“network pharmacology” OR “network analysis” OR “system-level” OR “systems-level” OR “systems pharmacology” OR “systems biology” OR “”) in [title] AND THM-related terms (“oriental medicine” OR “traditional medicine” OR “traditional Asian medicine” OR “Chinese medicine” OR “Kampo medicine” OR “Korean medicine”) in [title/abstract]. The search range of THM-related terms was extended to [title/abstract] since the titles of THM studies generally contain only the name of herbs or herbal formulae that are difficult to search.

2.2. Inclusion Criteria We considered a THM-NP study as the original article that analyzed a THM’s mode of action through the construction of a compound-target network. Full-text articles from the literature search were checked to determine their eligibility. There was no restriction regarding in vivo, in vitro, and in silico studies. THM was considered as (1) extract(s) from a single herb; (2) preparation(s) containing multiple herbs; (3) proprietary herbal product(s); and (4) molecule(s) derived from a single herb.

2.3. Study Selection and Data Extraction Two authors (W.Y. Lee and C.E. Kim) independently examined titles, abstracts, and journals to select eligible THM-NP studies. When articles were duplicated, only the most recent information was included. Then the full text of potentially relevant studies was retrieved. Two authors (W.Y. Lee and C.E. Kim) independently examined the full-text records to determine which studies met the inclusion criteria. Disagreements about the study selection were resolved by rechecking whether the studies met our criteria for inclusion. Authors extracted the following data from the included THM-NP studies: authors, affiliations, publication years, tools, and databases. Synonyms for tools and databases were merged and counted under a single keyword. DTpre and SysDT [16] were considered to be the same method as Traditional Chinese Medicine Systems Pharmacology Database and Analysis Platform (TCMSP, http://lsp.nwu. edu.cn/tcmsp.php)[17] since these methods were originally developed and implemented in TCMSP.

2.4. Categorizing Drug-Target Interaction Methods To capture the trends in the methods for constructing compound-target networks, we categorized DTI methods into four groups by their hypothesis and which information was used: the chemogenomic approach, docking simulation approach, ligand-based approach, and others [18–20]. The chemogenomic approach predicts potential compound–target pairs similar to validated compound-target pairs. This method is based on the assumption that a compound–target pair is likely to interact with high similarity to a validated compound-target interaction in terms of chemo–physical properties [21]. The docking simulation approach predicts the binding conformation of small-molecule ligands to the appropriate binding site of the target using 3D structural information on the compounds and protein targets [22]. The key hypothesis of this approach is that compounds with a high binding affinity at the binding site are likely to interact with the target [23]. The ligand-based approach predicts interactions by comparing a new ligand to known proteins’ ligands based on the hypothesis that similar molecules usually bind to similar proteins [24]. DTI methods that do not belong to the above categories were assigned to the category “others”, such as data mining techniques, high-throughput screening, and databases that integrate drug-target interaction information from heterogeneous sources. Biomolecules 2019, 9, x FOR PEER REVIEW 4 of 15 Biomolecules 2019, 9, 362 4 of 15 2.5. Construction of the Co-Author Network and Affiliation Network 2.5. ConstructionThe author ofnetwork the Co-Author and affiliation Network andnetwork Affiliation were Network constructed to identify the methodological characteristicsThe author of network corresponding and affi liationauthors network and affi wereliations. constructed The nodes to identifyin each thenetwork methodological represent authorscharacteristics or affiliations, of corresponding and the edges authors represent and affi liations.co-occurrences The nodes of authors in each or network affiliations represent in THM-NP authors studies.or affiliations, The frequencies and the edges of employed represent co-occurrencesDTI and drug ofavailability authors or methods affiliations were in THM-NPcounted for studies. each Thecorresponding frequencies author of employed or affiliation. DTI and These drug availabilitymethodologies methods were weremapped counted to the for author each corresponding network and theauthor affiliation or affiliation. network. These Cytosc methodologiesape 3.7.1 (http://www.cytoscape.or were mapped to the authorg/) was network used and to thevisualize affiliation the networksnetwork. Cytoscape[25]. 3.7.1 (http://www.cytoscape.org/) was used to visualize the networks [25].

3. Results Results

3.1. Description Description of of the Search We initiallyinitially foundfound 233 233 potentially potentially relevant relevant articles articles from from PubMed. PubMed. The searchThe search was conductedwas conducted using usingcombined combined keywords keywords consisting consisting of THM-related of THM-rela termsted and terms NP-related and NP-related terms. Another terms. 15 Another potentially 15 potentiallyrelevant articles relevant were articles included were by searchingincluded by references searching in otherreferences THM-NP in other studies THM-NP or review studies articles. or Titles,review abstracts, articles. Titles, and journal abstracts, names and were journal screened, names and were 167 screened, studies were and considered 167 studies potentially were considered eligible potentiallyfor inclusion. eligible Of these, for inclusion. 20 articles Of were these, excluded 20 articles after were screening excluded the full after texts. screening Finally, the a total full oftexts. 147 THM-NPFinally, a total studies of 147 were THM-NP included studies in our were study included (Figure2 in). our The study included (Figure THM-NP 2). The studies included are THM-NP listed in studiesSupplementary are listed Table in Supplementary S1. Table S1.

Figure 2. TheThe flowchart flowchart of the study selection process.

We identifiedidentified the number of publishedpublished THM-NP studies over time, along with the aaffiliationsffiliations involved in the studies. THM-NP THM-NP studies studies that that met met our our criteria criteria for for inclusion inclusion have have appeared appeared since since 2011. 2011. In 2011, only three aaffiliationsffiliations published two THM-NP studies, but in 2018, the number of aaffiliationsffiliations Biomolecules 2019, 9, 362 5 of 15 Biomolecules 2019, 9, x FOR PEER REVIEW 5 of 15 and studiesstudies increased to 83 andand 52,52, respectivelyrespectively (Figure(Figure3 3).). TheThe increasedincreased numbernumber ofof paperspapers andand ofof aaffiliationsffiliations involved indicates that that the the THM-NP THM-NP research research fields fields have have continuously continuously grown. grown.

Figure 3. Annual publication trends of traditional herbal medicine-network pharmacology (THM-NP)Figure 3. Annual studies. publication trends of traditional herbal medicine-network pharmacology (THM- 3.2. Methodological Trends in Constructing the Herb-CompoundNP) studies. Network 3.2. MethodologicalWe next investigated Trends in the Constructing trends in employedthe Herb-Compound methods Network in THM-NP studies. Commonly used databases and tools are described in Table1 (see Supplementary Table S2 for complete lists). We next investigated the trends in employed methods in THM-NP studies. Commonly used The construction of the herb-compound network is the first step of a THM-NP study. Among databases and tools are described in Table 1 (see Supplementary Table 2 for complete lists).The the databases for herbal medicines, TCMSP was most commonly used to construct an herb-compound construction of the herb-compound network is the first step of a THM-NP study. Among the network. Additionally, some THM-NP researchers used their own experimental results (e.g., Ultra databases for herbal medicines, TCMSP was most commonly used to construct an herb-compound Performance Liquid Chromatography (UPLC) or High-performance liquid chromatography) to identify network. Additionally, some THM-NP researchers used their own experimental results (e.g., Ultra ingredients of the herbal medicines in their studies (Figure4A). Performance Liquid Chromatography (UPLC) or High-performance liquid chromatography) to identify ingredients of the herbal medicines in their studies (Figure 4A). Table 1. The public databases related to traditional herbal medicine-network pharmacology (THM-NP) studies. Table 1. The public databases related to traditional herbal medicine-network pharmacology (THM- Providing Information PMID Name NP) studies.Description Website (Reference) H-C C-T TI Providing A system of pharmacology platforms http: PMID Name Information thatDescription provide information about Website 24735618 TCMSP //lsp.nwu.edu.cn/ ingredients, ADME-related properties, (Reference)[26] H-C C-T TI tcmsp.php ### targets, and diseases of herbal medicines. A system of pharmacology platforms that An integrative database which stores the http://lsp.nw information of herbs, herbal compounds, http://www. provide information about ingredients, 2320387524735618 TCMSPTCMID ○ ○ ○ targets, and their related information megabionet.orgu.edu.cn/tcm/ ADME-related properties, targets, and [17[26]] ### from different resources and through tcmidsp.php/ diseases oftext-mining herbal medicines. method An integrativeA database databa that includesse which the stores information the TCM information ofof herbs, molecular herbal properties compounds, and http:http://www.//tcm.cmu. 21253603 Databasetaiwan substructures, TCM ingredients with edu.tw/ 23203875[27] TCMID ○ #○ ○ targets, and their related information from megabionet. their 2D and 3D structures. [17] different resourcesA web server and for through potential text-mining drug target org/tcmid/ identificationmethod by reversed http://lilab.ecust. 20430828 PharmMapper pharmacophore matching the query edu.cn/ TCM A database that includes the information of [28] # compound against an in-house pharmmapperhttp://tcm.c/ 21253603 Databasetai ○ molecular propertiespharmacophore and substructures, model database TCM mu.edu.tw/ [27] wan ingredients with their 2D and 3D structures. Biomolecules 2019, 9, 362 6 of 15

Table 1. Cont.

Providing Information PMID Name Description Website (Reference) H-C C-T TI A database that integrates disparate data http: 18084021 STITCH sources of interactions between proteins //stitch.embl.de/ [29] # and small molecules A database that provides information about the therapeutic targets in the http://xin.cz3.nus. 11752352 TTD literature, targeted disease condition, edu.sg/group/ttd/ [30] ## and the corresponding drugs/ligands ttd.asp directed at each of these targets. A computational tool that relates proteins and chemicals based on the http: 17287757 SEA set-wise chemical similarity among their //sea.bkslab.org/ [31] # ligands. A comprehensive and fully curated http://lifecenter. 21097881 HIT database for herbal ingredients with sgst.cn/hit/ [32] ## protein target information A unique bioinformatics and cheminformatics resource that combines https://www. 16381955 Drugbank detailed drug data with comprehensive drugbank.ca/ [33] ## drug target information A database resource for understanding high-level functions and utilities of the https://www. KEGG 9847135 [34] from molecular-level genome.jp// # information http: The world’s largest source of 18792943 ontology //geneontology. information on the functions of [35] # org/ A comprehensive and authoritative https://www. 11752252 OMIM compendium of human genes and omim.org/ [36] # genetic phenotypes A database for the aggregation, curation, integration, and dissemination of https://www. 11752281 PharmGkb knowledge regarding the impact of pharmgkb.org/ [37] # human genetic variation on drug response A searchable and integrated database of human genes that provides concise https://www. 12424129 Genecards genomic related information, on all genecards.org/ [38] # known and predicted human genes. H-C, herb-compound network construction; C-T, compound-target network construction; TI, target interpretation. TCMSP, Traditional Chinese Medicine Systems Pharmacology Database; TCMID, Traditional Chinese Medicine Integrated Database; STITCH, Search Tool for Interactions of Chemicals; TTD, Therapeutic Target Database; SEA, Similarity Ensemble Approach; HIT, Herb Ingredients’ Targets; KEGG, Kyoto Encyclopedia of Genes and Genomes database; OMIM, Online Mendelian Inheritance in Man; PharmGkb, The Pharmacogenetics Knowledge Base. BiomoleculesBiomolecules2019 2019, 9,, 9 362, x FOR PEER REVIEW 77 of of 15 15

Figure 4. Methodological trends in the construction of the herb–compound network. (A) Frequency ofFigure databases 4. Methodological and tools used trends to identify in the constituents construction of of herbal the herb–compound medicine. * Analytical network. techniques (A) Frequency to identifyof databases the ingredients and tools ofused herbal to identify medicines; constituents Ultra Performance of herbal medicine. Liquid Chromatography * Analytical techniques (UPLC), to High-performanceidentify the ingredients liquid of chromatography herbal medicines; (HPLC) Ultra Performance (B) The application Liquid Chromatography rate of the drug (UPLC), availability High- methodperformance by year. liquid Note chro thatmatography Obioavail (OB) (HPLC) and drug-likeness(B) The application (DL) arerateamong of the drug the most availability commonly method used by drugyear. availability Note that assessmentObioavail (OB) methods and indrug-likeness THM-NP studies. (DL) are THM-NP among studiesthe most in commonly 2011 were excludedused drug fromavailability the visualization assessment due methods to the low in THM-NP frequency studies. (n = 2). THM-NP TCMSP, Traditional studies in 2011 Chinese were Medicine excluded Systems from the Pharmacologyvisualization Database;due to the TCMID, low frequency Traditional (n Chinese = 2). TCMSP, Medicine Traditiona Integratedl Chinese Database. Medicine Systems Pharmacology Database; TCMID, Traditional Chinese Medicine Integrated Database. Because information on the absorption, distribution, , and excretion (ADME) properties of herbalBecause medicines information in humans on are the lacking, absorption, researchers dist haveribution, employed metabolism, evaluation and methods excretion or machine(ADME) learningproperties tools of to predictherbal thosemedicines properties. in humans We counted are thelacking, number researchers of THM-NP have studies employed that evaluated evaluation the drugmethods availability or machine of herbal learning ingredients. tools to Wepredict found thos thate properties. approximately We halfcounted of (72 the/147, number 49.0%) of THM-NP THM-NP studiesstudies evaluated that evaluated the drug the availabilitydrug availability of herbal of herbal ingredients, ingredients. and the We majority found that of theapproximately studies (54/ 72,half 75.0%)of (72/147, employed 49.0%) Obioavail THM-NP and studies drug-likeness evaluated in the combination drug availability (Figure 4ofB). herbal Obioavail ingredients, is an in and silico the modelmajority that of predicts the studies the fraction (54/72, of75.0%) an administered employed Obioavail dose of a drugand drug-likeness that reaches the in systemiccombination circulation (Figure unchanged4B). Obioavail [39]. is Drug-likeness an in silico model measures that thepredicts structural the fraction similarity of betweenan administered herbal ingredients dose of a drug and thethat drugsreaches inthe the Drugbank systemic databasecirculation (http: unchanged//www.drugbank.ca [39]. Drug-likeness/) using the measures Tanimoto the coe structuralfficient [40 similarity]. They arebetween applied herbal to screen ingredients ADME-favorable and the drugs compounds in the andDrugbank pharmacologically database (http://www.drugbank.ca/) suitable compounds in herbalusing medicines, the Tanimoto respectively. coefficient [40]. They are applied to screen ADME-favorable compounds and pharmacologically suitable compounds in herbal medicines, respectively. 3.3. Methodological Trends for Constructing Compound-Target Networks 3.3.We Methodological next attempted Trends for to Constructing determine theCompound-Target frequency of Networks each DTI method for constructing compound–targetWe next attempted (C-T) networks to determine (Note the that frequency some of theof each THM-NP DTI method studies for combined constructing several compound– methods totarget identify (C-T) DTIs. networks Therefore, (Note the that total some frequency of the THM-NP of the DTI studies method combined is greater several than the methods total number to identify of THM-NPDTIs. Therefore, studies). the The total results frequency showed of thatthe DTI TCMSP meth (47od/222, is greater 21.1%) than and the molecular total number docking of THM-NP (44/222, 19.8%)studies). were The the results most frequentlyshowed that used. TCMSP In addition, (47/222, experimental21.1%) and molecular methods, docking such as (44/222, microarrays, 19.8%) were were alsothe appliedmost frequently (Figure5 A).used. It isIn noteworthy addition, experimental that DTI methods methods, of TCMSPsuch as microarrays, have existed were for less also than applied 10 years(Figure since 5A). its It development is noteworthy but that have DTI been methods usedmost of TCMSP frequently have inexisted THM-NP for less studies than [1016 ].years More since than its one-thirddevelopment of THM-NP but have studies been used (54/147, most 36.7%) frequently combined in THM-NP several DTIstudies methods [16]. More for constructing than one-third C-T of networks,THM-NP and studies most (54/147, of them 36.7%) included combined TCMSP several (e.g., TCMSP-molecular DTI methods for constructing docking and TCMSP-STITCH)C-T networks, and (Figuremost of5B). them included TCMSP (e.g., TCMSP-molecular docking and TCMSP-STITCH) (Figure 5B). Biomolecules 2019, 9, 362 8 of 15 Biomolecules 2019, 9, x FOR PEER REVIEW 8 of 15

Figure 5. Methodological trends in the construction of the compound–target network. (A) The Figure 5. Methodological trends in the construction of the compound–target network. (A) The frequencies of databases and tools used to identify targets of herbal ingredients. * databases that frequencies of databases and tools used to identify targets of herbal ingredients. * databases that provide validated drug-target interactions (DTIs); ** databases that provide validated databases that provide validated drug-target interactions (DTIs); ** databases that provide validated databases that provide both validated and predicted DTIs; # Computational tools to predict DTIs. (B) A co-occurrence provide both validated and predicted DTIs; # Computational tools to predict DTIs. (B) A co- pattern of the DTI method. (C) Categories of DTI methods and their composition. The outer circle and occurrence pattern of the DTI method. (C) Categories of DTI methods and their composition. The inner circle represent the DTI methods and their categories, respectively. (D) The annual rate of the outer circle and inner circle represent the DTI methods and their categories, respectively. (D) The groups of DTI methods. Note that THM-NP studies in 2011 were excluded from the visualization due annual rate of the groups of DTI methods. Note that THM-NP studies in 2011 were excluded from the to their low frequency (n = 2). TCMSP, Traditional Chinese Medicine Systems Pharmacology Database; visualization due to their low frequency (n = 2). TCMSP, Traditional Chinese Medicine Systems TTD, Therapeutic Target Database; SEA, Similarity Ensemble Approach; HIT, Herb Ingredients’ Targets; Pharmacology Database; TTD, Therapeutic Target Database; SEA, Similarity Ensemble Approach; WES, Weighted Ensemble Similarity. HIT, Herb Ingredients’ Targets; WES, Weighted Ensemble Similarity. We categorized the methods into four groups: the chemogenomic approach, docking simulation approach,We categorized ligand-based the methods approach, into and four others groups: (Figure the chemogenomic5C, see Materials approach, and Methods docking for simulation details). To approach,identify trendsligand-based in DTI methods,approach, we and counted others (Figure the frequency 5C, see of Materials each DTI and group Methods each yearfor details). (Figure 5ToD). identifyIn the earlytrends stage, in DTI approximately methods, we halfcounted of THM-NP the frequency studies of (4each/9, 44.4%DTI group and 8each/18, 44.4%year (Figure in 2012 5D). and In2013, the early respectively) stage, approximately employed molecular half of dockingTHM-NP simulation, studies (4/9, but 44.4% the proportion and 8/18, of 44.4% molecular in 2012 docking and 2013,simulations respectively) decreased employed gradually molecular and was docking the lowest simulation, (12/77, but 15.9%) the proportion in 2018. of molecular docking simulations decreased gradually and was the lowest (12/77, 15.9%) in 2018.

3.4. Methodological Trends for Target Interpretation We identified the frequency of biomedical databases employed to analyze the biological processes, pathways, and diseases from the targets of herbal medicines (Figure 6). Most THM-NP Biomolecules 2019, 9, 362 9 of 15

3.4. Methodological Trends for Target Interpretation BiomoleculesWe identified 2019, 9, x theFOR frequency PEER REVIEW ofbiomedical databases employed to analyze the biological processes,9 of 15 pathways, and diseases from the targets of herbal medicines (Figure6). Most THM-NP studies studies (94/96, 98.0%) employed single databases to analyze biological functions and pathways, such (94/96, 98.0%) employed single databases to analyze biological functions and pathways, such as Gene as Gene Ontology (GO, http://geneontology.org/) [35] for biological processes or the Kyoto Ontology (GO, http://geneontology.org/)[35] for biological processes or the Kyoto Encyclopedia of Encyclopedia of Genes and Genomes database (KEGG, https://www.genome.jp/kegg/) [41] for Genes and Genomes database (KEGG, https://www.genome.jp/kegg/)[41] for pathways. On the other pathways. On the other hand, more than half of the studies (44/76, 57.9%) integrated several hand, more than half of the studies (44/76, 57.9%) integrated several databases to analyze target-related databases to analyze target-related diseases, such as the Therapeutic Target Database (TTD, diseases, such as the Therapeutic Target Database (TTD, http://bidd.nus.edu.sg/group/cjttd/)[30], http://bidd.nus.edu.sg/group/cjttd/) [30], Online Mendelian Inheritance in Man (OMIM, Online Mendelian Inheritance in Man (OMIM, https://www.omim.org/)[36], and Drugbank [42]. https://www.omim.org/) [36], and Drugbank [42].

Figure 6. Trends in the biomedical databases used for target interpretation. (A) The frequencies of databasesFigure 6. usedTrends for in target the interpretation.biomedical databases (B) Categories used for of target databases interpretation. and their composition. (A) The frequencies The outer of circledatabases and innerused for circle target represent interpretation. the databases (B) Categories used for of targetdatabases interpretation and their composition. and their categories, The outer respectively.circle and inner Note circle that THM-NPrepresent studiesthe databases in 2011 used were for excluded target frominterpretation the visualization and their due categories, to their lowrespectively. frequency Note (n = 2).that KEGG, THM-NP Kyoto studies Encyclopedia in 2011 were of Genes excluded and Genomes; from the TTD,visualization Therapeutic due Targetto their Database;low frequency OMIM, (n Online= 2). KEGG, Mendelian Kyoto Inheritance Encyclopedia in Man; of Gene IPA,s and Ingenuity Genomes; Pathway TTD, Analysis. Therapeutic Target Database; OMIM, Online Mendelian Inheritance in Man; IPA, Ingenuity Pathway Analysis. 3.5. Combinatorial Patterns in Methodologies of THM-NP Studies 3.5. Combinatorial Patterns in Methodologies of THM-NP Studies We identified the combinatorial patterns of each step in THM-NP studies by a Sankey diagram-like representationWe identified (Figure the7 combinatorial). The Sankey patterns diagram of iseach a visualization step in THM-NP tool usedstudies to by depict a Sankey quantitative diagram- informationlike representation about flows (Figure from 7). one The set Sankey to another diagram within is a a network visualization [43]. The tool nodes used into eachdepict layer quantitative (vertical lines)information represent about the methodsflows from of herb–compound one set to another (H-C) within network a network construction, [43]. The compound–target nodes in each (C-T)layer network(vertical construction,lines) represent and the target methods interpretation, of herb–compound respectively. (H-C) The network edges (connected construction, lines) compound– between layerstarget indicate (C-T) network that these construction, methods are and used target together interpretation, in the same respectively. THM-NP The studies. edges (connected lines) betweenThe Sankeylayers indicate diagram-like that these representation methods are shows used together the diversity in the ofsame databases THM-NP and studies. tools used in THM-NP studies and their combination patterns (Figure7). We found that the nodes in the first layer (H-C network construction) tend to be connected to specific nodes in the second layer (C-T network construction), which indicates that the combinatorial pattern between the first and second layer is biased by the methods for H-C network construction. For example, TCMSP in the first layer is mainly connected to TCMSP and molecular docking in the second layer, and Traditional Chinese Medicine Integrated Database (TCMID), UPLC, and literature mining in the first layer are not linked to molecular docking in the second layer. On the other hand, the nodes in the second layer tended to be evenly connected to the nodes in the third layer (target interpretation), which indicates that the combinatorial Biomolecules 2019, 9, 362 10 of 15

pattern between the second layer and third layer are relatively independent of the methods for C-T Biomoleculesnetwork construction.2019, 9, x FOR PEER REVIEW 10 of 15

Figure 7. The combinatorial pattern of tools and databases employed in THM-NP studies. Each layer Figurerepresents 7. The the combinatorial process of network pattern pharmacology of tools and database analysiss employed of herbal medicines,in THM-NP and studies. the components Each layer representsof each layer the representprocess of the network employed pharmacology tools and databases. analysis of Aherbal connecting medicines, line betweenand the components ofindicates each layer that represent the connected the employed tools and tools databases and databases. are used A together connecting in the line same between THM-NP components studies. indicatesThe thickness that the of eachconnected connecting tools and line indicatesdatabasesthe arefrequency used together with in which the same the twoTHM-NP methodologies studies. The are thicknessused together of each in theconnecting THM-NP line studies. indicates TCMSP, the fr Traditionalequency with Chinese which Medicinethe two methodologies Systems Pharmacology are used togetherDatabase; in TCMID, the THM-NP Traditional studie Chineses. TCMSP, Medicine Traditional Integrated Chinese Database; Medicine HIT, Herb Systems Ingredients’ Pharmacology Targets; Database;KEGG, Kyoto TCMID, Encyclopedia Traditional of Chinese Genes and Medicine Genomes; Integr TTD,ated Therapeutic Database; TargetHIT, Herb Database; Ingredients’ OMIM, Targets; Online KEGG,Mendelian Kyoto Inheritance Encyclopedia in Man. of Genes and Genomes; TTD, Therapeutic Target Database; OMIM, 3.6. Co-AuthorOnline Mendelian Network Inheritance and Affiliation in Man. Network

TheTo furtherSankey understanddiagram-like how repres theentation methodologies shows the of diversity THM-NP of studies databases are and employed tools used among in THM-NPresearchers, studies we constructedand their combination a co-author patterns network (Figure and 7). an We affi foundliation that network the nodes that in were the first mapped layer (H-Cwith drugnetwork availability construction) and DTI tend methods. to be connected The nodes to inspecific each network nodes in denote the second the author layer and(C-T a networkffiliation, construction),and the edges which indicate indicates that two thatof the them combinatorial appear on pattern the same between paper. the The first methods and second of DTI layer and is biaseddrug-availability by the methods used byfor theH-C corresponding network construction. author and For affi example,liation are TCMSP represented in the byfirst the layer pie is chart mainly and connectedthe outline, to respectively. TCMSP and molecular docking in the second layer, and Traditional Chinese Medicine IntegratedIn the Database author network, (TCMID), Yonghua UPLC, Wang and (nliterature= 18) and mining Shao Li in (n the= 8) first appeared layer mostare not frequently linked to as molecularthe corresponding docking in author the second (Figure layer.8). More On the than othe a thirdr hand, of correspondingthe nodes in the authors second (52 layer/147) tended combined to be evenlyDTI methods, connected such to as the the nodes chemogenomic in the third approach, layer (target docking interpretation), simulation approach, which indicates and ligand-based that the combinatorialapproach. Approximately pattern between half ofthe the second corresponding layer and authorsthird layer (69/ 147)are relatively employed independent evaluation tools of the to methodsscreen for for compounds C-T network with construction. favorable pharmacokinetic properties, and most of them (50/69) used Obioavail and drug-likeness in combination. 3.6. Co-Author Network and Affiliation Network To further understand how the methodologies of THM-NP studies are employed among researchers, we constructed a co-author network and an affiliation network that were mapped with drug availability and DTI methods. The nodes in each network denote the author and affiliation, and the edges indicate that two of them appear on the same paper. The methods of DTI and drug- Biomolecules 2019, 9, x FOR PEER REVIEW 11 of 15

availability used by the corresponding author and affiliation are represented by the pie chart and the outline, respectively. In the author network, Yonghua Wang (n = 18) and Shao Li (n = 8) appeared most frequently as the corresponding author (Figure 8). More than a third of corresponding authors (52/147) combined DTI methods, such as the chemogenomic approach, docking simulation approach, and ligand-based approach. Approximately half of the corresponding authors (69/147) employed evaluation tools to screen for compounds with favorable pharmacokinetic properties, and most of them (50/69) used BiomoleculesObioavail2019 and, 9 drug-likeness, 362 in combination. 11 of 15

Figure 8. The co-author network of THM-NP studies. Circles represent corresponding authors, and squares represent non-corresponding authors. The size of the circles and squares reflect the number ofFigure occurrences 8. The co-author in the THM-NP network studies. of THM-NP Nodes studies. that appeared Circles fewerrepresent than corresponding three times were authors, removed. and Thesquares box torepresent the right non-corresponding of the network represents authors. the The index size forof the the circles pie chart and and squares the outline reflect of the the number circle. of occurrences in the THM-NP studies. Nodes that appeared fewer than three times were removed. WeThe alsobox to constructed the right of andthe network visualized represents the affi liationthe index network for the (Supplementarypie chart and the outline Figure of S1). the Northwestcircle. A&F University (n = 22) and China Academy of Chinese Medical Sciences (n = 17) appeared most frequently.We also Most constructed affiliations combinedand visualized various the DTI affi methodsliation (68network/143) and (Supplementary employed drug-availability Figure S1). methodsNorthwest (86 A&F/143). University (n = 22) and China Academy of Chinese Medical Sciences (n = 17) appeared most frequently. Most affiliations combined various DTI methods (68/143) and employed 4. Discussion drug-availability methods (86/143). In this study, we successfully identified the complex methodological trends of THM-NP research fields4. Discussion by analyzing the frequency of the employed methods in THM-NP studies over time and visualizingIn this thestudy, combinatorial we successfully patterns identified between the them. complex Our methodological results showed that trends the of number THM-NP of THM-NP research studiesfields by and analyzing employed the databases frequency/tools of havethe employed been dramatically methods increasedin THM-NP in the studies last decade. over time We alsoand foundvisualizing characteristic the combinatorial patterns existpatterns in combining between th methodsem. Our ofresults each analysisshowed that step the in THM-NPnumber of studies. THM- Finally,NP studies we showedand employed how the databases/tools diverse methods have for been THM-NP dramatically studies areincreased employed in the and last shared decade. among We researchersalso found incharacteristic the field by analyzingpatterns exist the co-authorshipin combining andmethods affiliation of each networks. analysis step in THM-NP studies.Among Finally, the networkwe showed pharmacology how the diverse databases, methods TCMSP for THM-NP was the most studies frequently are employed employed and database shared foramong constructing researchers herb-compound-target in the field by analyzing networks. the co-authorship This database and was affiliation developed networks. in 2014 and has been predominantlyAmong the employed network among pharmacology THM-NP studiesdatabases, [17 ].TCMSP TCMSP was provides the most a network frequently pharmacological employed analysisdatabase of for 499 constructing medicinal herb-compound-target herbs registered in the ne Chinesetworks. pharmacopeia This database alongwas developed with information in 2014 and on ADMEhas been properties, predominantly such as employed bioavailability, among drug-likeness, THM-NP studies and P450, [17]. in TCMSP a one-step provides manner. a Recently,network otherpharmacological network pharmacology analysis of 499 databases, medicinal such herbs as BATMAN-TCM registered in the (http: Chinese//bionet.ncpsb.org pharmacopeia/batman-tcm along with) andinformation TCM-Mesh on (http:ADME//mesh.tcm.microbioinformatics.org properties, such as bioavailability,/), were drug-likeness, developed [ 44and,45 ].P450, They in are a expectedone-step to further facilitate THM-NP research fields by providing network pharmacological analysis for more than 5000 medicinal herbs. Among the processes used in THM-NP research, the methodology for constructing a C-T network has shown the greatest change over time. In the early stages of THM-NP research, DTI methods for identifying targets of herbal ingredients relied on molecular docking simulations, which require high computational resources (Figure5D). With the advancement of DTI prediction methods, several methodologies have been applied to THM-NP research fields that can efficiently identify the multiple targets of multiple ingredients in herbal medicines. First, the development and application of machine Biomolecules 2019, 9, 362 12 of 15 learning techniques and network-based methods enabled large-scale prediction of the targets of herbal medicines in terms of efficient computational costs [17,44,46,47]. Second, increased computational power made it possible to comprehensively explore potential targets of the compounds using the pharmacophore model [48]. Last, the development of databases that integrate disparate data sources provides comprehensive and high-quality information on DTIs [49]. Furthermore, recently developed DTI prediction models based on deep learning showed higher performance than other state-of-the-art models [50,51]. Such innovation in the machine learning field is expected to facilitate the development of the THM-NP research field. To conduct THM-NP research, various tools and databases are combined in each phase of a study (Figure7). We found that the methods for H-C network construction tend to be linked with specific methods for C-T network construction. This result indicates that there might be a preferred combinatorial pattern when choosing the methods for constructing H-C-T network. On the other hand, the combinatorial patterns between the methods for C-T network construction and target interpretation are relatively independent when compared to the previous step. This result indicates that the databases used for target interpretation tend to be chosen for the purpose of the study, while each method was preferred by different researchers in the previous steps. Further studies are needed to evaluate the reliability of network pharmacological analysis by evaluating the consistency between predicted results according to the methodologies of THM-NP studies. There are some limitations to our study that should be noted. First, we identified potentially relevant articles using combined keywords consisting of THM-related terms and NP-related terms. Although we carefully selected search terms, we cannot guarantee that our search strategy can fully identify THM-NP studies. Second, we found potentially relevant articles only in PubMed. It is one of the largest electronic database in the world. However, there are other databases which may include other potential THM-NP studies, such as Embase, China Knowledge Resource Integrated Database (CNKI), Research Information Sharing Service (RISS), and Japan Science Technology Information Aggregator (J-stage). Last, we limited the search range of our study to English literature, which might have introduced some bias. In spite of these limitations, our results will help to improve the understanding of the methodological trends of the THM-NP research fields.

5. Conclusions In conclusion, we investigated the methodological trends in THM-NP studies. Our results provide researchers with the current status of which methodologies are used in THM-NP studies and how they are applied.

Supplementary Materials: The following are available online at http://www.mdpi.com/2218-273X/9/8/362/s1. Supplementary Figure S1. The affiliation network of THM-NP studies; Supplementary Table S1. Selected THM-NP studies; Supplementary Table S2. Providing information and their types of methods employed in THM-NP studies. Author Contributions: Conceptualization, Y.-S.K. and C.-E.K.; Data curation, W.-Y.L.; Formal analysis, W.-Y.L. and C.-Y.L.; Visualization, W.-Y.L.; Writing—original draft, W.-Y.L.; Writing—review and editing, C.-Y.L., Y.-S.K., and C.-E.K. Funding: This work was supported by the National Research Foundation of Korea (NRF) grant funded by the Korea government (MSIP; Ministry of Science, ICT & Future Planning) (No. 2017-0129). This work was also supported by the Gachon University research fund of 2019 (GCU-2019-0025). Conflicts of Interest: The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript; or in the decision to publish the results. Biomolecules 2019, 9, 362 13 of 15

References

1. Kong, D.X.; Li, X.J.; Zhang, H.Y. Where is the hope for ? Let history tell the future. Drug Discov. Today 2009, 14, 115–119. [CrossRef][PubMed] 2. Verpoorte, R.; Crommelin, D.; Danhof, M.; Gilissen, L.J.; Schuitmaker, H.; van der Greef, J.; Witkamp, R.F. Commentary: “A systems view on the future of medicine: Inspiration from Chinese medicine?”. J. Ethnopharmacol. 2009, 121, 479–481. [CrossRef][PubMed] 3. Zheng, J.; Wu, M.; Wang, H.; Li, S.; Wang, X.; Li, Y.; Wang, D.; Li, S. Network Pharmacology to Unveil the Biological Basis of Health-Strengthening Herbal Medicine in Cancer Treatment. Cancers (Basel) 2018, 10, 461. [CrossRef][PubMed] 4. Ma, X.H.; Zheng, C.J.; Han, L.Y.; Xie, B.; Jia, J.; Cao, Z.W.; Li, Y.X.; Chen, Y.Z. Synergistic therapeutic actions of herbal ingredients and their mechanisms from molecular interaction and network perspectives. Drug Discov. Today 2009, 14, 579–588. [CrossRef][PubMed] 5. Berg, E.L. Systems biology in drug discovery and development. Drug Discov. Today 2014, 19, 113–125. [CrossRef][PubMed] 6. Barabási, A.-L.; Gulbahce, N.; Loscalzo, J. Network medicine: A network-based approach to human disease. Nat. Rev. Genet. 2011, 12, 56–68. [CrossRef][PubMed] 7. Hopkins, A.L. Network pharmacology: The next paradigm in drug discovery. Nat. Chem. Biol. 2008, 4, 682–690. [CrossRef] 8. Van der Greef, J. Perspective: All systems go. Nature 2011, 480, S87. [CrossRef] 9. Li, S.; Fan, T.-P.; Jia, W.; Lu, A.; Zhang, W. Network Pharmacology in Traditional Chinese Medicine. Evid. Based Complement. Altern. Med. 2014, 2014, 138460. [CrossRef] 10. Li, S.; Zhang, B. Traditional Chinese medicine network pharmacology: Theory, methodology and application. Chin. J. Nat. Med. 2013, 11, 110–120. [CrossRef] 11. Zhang, R.; Zhu, X.; Bai, H.; Ning, K. Network Pharmacology Databases for Traditional Chinese Medicine: Review and Assessment. Front. Pharmacol. 2019, 10, 123. [CrossRef][PubMed] 12. Yang, M.; Chen, J.-L.; Xu, L.-W.; Ji, G. Navigating Traditional Chinese Medicine Network Pharmacology and Computational Tools. Evid. Based Complement. Altern. Med. 2013, 2013, 731969. [CrossRef][PubMed] 13. 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. Nat. Prod. Rep. 2015, 32, 1249–1266. [CrossRef][PubMed] 14. Hao, D.C.; Xiao, P.G. Network pharmacology: A rosetta stone for traditional chinese medicine. Drug Dev. Res. 2014, 75, 299–312. [CrossRef][PubMed] 15. Liu, Y.; Ai, N.; Keys, A.; Fan, X.; Chen, M. Network Pharmacology for Traditional Chinese Medicine Research: Methodologies and Applications. Chin. Herb. Med. 2015, 7, 18–26. [CrossRef] 16. Yu, H.; Chen, J.; Xu, X.; Li, Y.; Zhao, H.; Fang, Y.; Li, X.; Zhou, W.; Wang, W.; Wang, Y. A systematic prediction of multiple drug-target interactions from chemical, genomic, and pharmacological data. PLoS ONE 2012, 7, e37608. [CrossRef][PubMed] 17. Ru, J.; Li, P.; Wang, J.; Zhou, W.; Li, B.; Huang, C.; Li, P.; Guo, Z.; Tao, W.; Yang, Y.; et al. TCMSP: A database of systems pharmacology for drug discovery from herbal medicines. J. Cheminform. 2014, 6, 13. [CrossRef] 18. Katsila, T.; Spyroulias, G.A.; Patrinos, G.P.; Matsoukas, M.T. Computational approaches in target identification and drug discovery. Comput. Struct. Biotechnol. J. 2016, 14, 177–184. [CrossRef] 19. Macalino, S.J.Y.; Gosu, V.; Hong, S.; Choi, S. Role of computer-aided in modern drug discovery. Arch. Pharm. Res. 2015, 38, 1686–1701. [CrossRef] 20. Yang, S.Y. Pharmacophore modeling and applications in drug discovery: Challenges and recent advances. Drug Discov. Today 2010, 15, 444–450. [CrossRef] 21. Ding, H.; Takigawa, I.; Mamitsuka, H.; Zhu, S. Similarity-basedmachine learning methods for predicting drug-target interactions: A brief review. Brief. Bioinform. 2013, 15, 734–747. [CrossRef][PubMed] 22. Lengauer, T.; Rarey, M. Computational methods for biomolecular docking. Curr. Opin. Struct. Biol. 1996, 6, 402–406. [CrossRef] 23. Cherfils, J.; Duquerroy, S.; Janin, J. Protein-protein recognition analyzed by docking simulation. Proteins Struct. Funct. Bioinform. 1991, 11, 271–280. [CrossRef][PubMed] Biomolecules 2019, 9, 362 14 of 15

24. Vidal, D.; Garcia-Serna, R.; Mestres, J. Ligand-Based Approaches to In Silico Pharmacology. Methods Mol. Biol. 2010, 672, 489–502. 25. Shannon, P.; Markiel, A.; Ozier, O.; Baliga, N.S.; Wang, J.T.; Ramage, D.; Amin, N.; Schwikowski, B.; Ideker, T. Cytoscape: A software environment for integrated models of biomolecular interaction networks. Genome Res. 2003, 13, 2498–2504. [CrossRef] 26. Xue, R.; Fang, Z.; Zhang, M.; Yi, Z.; Wen, C.; Shi, T. TCMID: Traditional Chinese medicine integrative database for herb molecular mechanism analysis. Nucleic Acids Res. 2013, 41, 1089–1095. [CrossRef] 27. Chen, C.Y.C. TCM Database@Taiwan: The world’s largest traditional Chinese medicine database for drug screening In Silico. PLoS ONE 2011, 6, e15939. [CrossRef] 28. Liu, X.; Ouyang, S.; Yu, B.; Liu, Y.; Huang, K.; Gong, J.; Zheng, S.; Li, Z.; Li, H.; Jiang, H. PharmMapper server: A web server for potential drug target identification using pharmacophore mapping approach. Nucleic Acids Res. 2010, 38, 5–7. [CrossRef] 29. Kuhn, M.; von Mering, C.; Campillos, M.; Jensen, L.J.; Bork, P. STITCH: Interaction networks of chemicals and proteins. Nucleic Acids Res. 2008, 36, 684–688. [CrossRef] 30. Chen, X.; Ji, Z.L.; Chen, Y.Z. TTD: Therapeutic Target Database. Nucleic Acids Res. 2002, 30, 412–415. [CrossRef] 31. Keiser, M.J.; Roth, B.L.; Armbruster, B.N.; Ernsberger, P.; Irwin, J.J.; Shoichet, B.K. Relating protein pharmacology by ligand chemistry. Nat. Biotechnol. 2007, 25, 197–206. [CrossRef] 32. Ye, H.; Ye, L.; Kang, H.; Zhang, D.; Tao, L.; Tang, K.; Liu, X.; Zhu, R.; Liu, Q.; Chen, Y.Z.; et al. HIT: Linking herbal active ingredients to targets. Nucleic Acids Res. 2011, 39, 1055–1059. [CrossRef] 33. Wishart, D.S. DrugBank: A comprehensive resource for in silico drug discovery and exploration. Nucleic Acids Res. 2005, 34, D668–D672. [CrossRef] 34. Ogata, H.; Goto, S.; Sato, K.; Fujibuchi, W.; Bono, H.; Kanehisa, M. KEGG: Kyoto encyclopedia of genes and genomes. Nucleic Acids Res. 1999, 28, 27–30. [CrossRef] 35. Harris, M.A.; Clark, J.; Ireland, A.; Lomax, J.; Ashburner, M.; Foulger, R.; Eilbeck, K.; Lewis, S.; Marshall, B.; Mungall, C.; et al. The Gene Ontology (GO) database and informatics resource. Nucleic Acids Res. 2004, 32, D258–D261. 36. Hamosh, A.; Scott, A.F.; Amberger, J.S.; Bocchini, C.A.; McKusick, V.A. Online Mendelian Inheritance in Man (OMIM), a knowledgebase of human genes and genetic disorders. Nucleic Acids Res. 2005, 33, D514–D517. [CrossRef] 37. Hewett, M.; Oliver, D.E.; Rubin, D.L.; Easton, K.L.; Stuart, J.M.; Altman, R.B.; Klein, T.E. PharmGKB: The pharmacogenetics knowledge base. Nucleic Acids Res. 2002, 30, 163–165. [CrossRef] 38. Safran, M.; Solomon, I.; Shmueli, O.; Lapidot, M.; Shen-Orr, S.; Adato, A.; Ben-Dor, U.; Esterman, N.; Rosen, N.; Peter, I.; et al. GeneCardsTM 2002: Towards a complete, object-oriented, human gene compendium. Bioinformatics 2002, 18, 1542–1543. [CrossRef] 39. Wang,X.; Xu, X.; Tao,W.; Li, Y.; Wang,Y.; Yang,L. A Systems Biology Approach to Uncovering Pharmacological Synergy in Herbal Medicines with Applications to Cardiovascular Disease. Evid. Based Complement. Altern. Med. 2012, 2012, 519031. [CrossRef] 40. Vistoli, G.; Pedretti, A.; Testa, B. Assessing drug-likeness—What are we missing? Drug Discov. Today 2008, 13, 285–294. [CrossRef] 41. Kanehisa, M.; Furumichi, M.; Tanabe, M.; Sato, Y.; Morishima, K. KEGG: New perspectives on genomes, pathways, diseases and drugs. Nucleic Acids Res. 2017, 45, D353–D361. [CrossRef] 42. Wishart, D.S.; Feunang, Y.D.; Guo, A.C.; Lo, E.J.; Marcu, A.; Grant, J.R.; Sajed, T.; Johnson, D.; Li, C.; Sayeeda, Z.; et al. DrugBank 5.0: A major update to the DrugBank database for 2018. Nucleic Acids Res. 2018, 46, D1074–D1082. [CrossRef] 43. Soundararajan, K.; Ho, H.K.; Su, B. Sankey diagram framework for energy and exergy flows. Appl. Energy 2014, 136, 1035–1042. [CrossRef] 44. Liu, Z.; Guo, F.; Wang, Y.; Li, C.; Zhang, X.; Li, H.; Diao, L.; Gu, J.; Wang, W.; Li, D.; et al. BATMAN-TCM: A Bioinformatics Analysis Tool for Molecular mechANism of Traditional Chinese Medicine. Sci. Rep. 2016, 6, 21146. [CrossRef] 45. Zhang, R.Z.; Yu, S.J.; Bai, H.; Ning, K. TCM-Mesh: The database and analytical system for network pharmacology analysis for TCM preparations. Sci. Rep. 2017, 7, 2821. [CrossRef] Biomolecules 2019, 9, 362 15 of 15

46. Zhao, S.; Li, S. Network-based relating pharmacological and genomic spaces for drug target identification. PLoS ONE 2010, 5, e11764. [CrossRef] 47. Wu, Z.; Lu, W.; Wu, D.; Luo, A.; Bian, H.; Li, J.; Li, W.; Liu, G.; Huang, J.; Cheng, F.; et al. In silico prediction of chemical mechanism of action via an improved network-based inference method. Br. J. Pharmacol. 2016, 173, 3372–3385. [CrossRef] 48. Wang, X.; Shen, Y.; Wang, S.; Li, S.; Zhang, W.; Liu, X.; Lai, L.; Pei, J.; Li, H. PharmMapper 2017 update: A web server for potential drug target identification with a comprehensive target pharmacophore database. Nucleic Acids Res. 2017, 45, W356–W360. [CrossRef] 49. Szklarczyk, D.; Santos, A.; Von Mering, C.; Jensen, L.J.; Bork, P.; Kuhn, M. STITCH 5: Augmenting protein-chemical interaction networks with tissue and affinity data. Nucleic Acids Res. 2016, 44, D380–D384. [CrossRef] 50. Bahi, M.; Batouche, M. Deep semi-supervised learning for DTI prediction using large datasets and H2O-spark platform. In Proceedings of the 2018 International Conference on Intelligent Systems and Computer Vision (ISCV), Fez, Morocco, 2–4 April 2018; pp. 1–7. 51. Wen, M.; Zhang, Z.; Niu, S.; Sha, H.; Yang, R.; Yun, Y.; Lu, H. Deep-Learning-Based Drug-Target Interaction Prediction. J. Proteome Res. 2017, 16, 1401–1409. [CrossRef]

© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).