A Comprehensive and Scientifically Accurate Pharmaceutical

Knowledge Ontology based on Multi-source Data

Pengfei Wang

, Yiqing Mao

, Wei Song

, Wenting Jiang

, Yang Liu

, Liumeng Zheng

Bin Ma

, Qingqing Sun

and Sheng Liu

Beijing MedPeer Information Technology Co., Ltd., Beijing, China

Keywords: Pharmaceutical Knowledge, Ontology, Drug, Knowledge Graph.

Abstract: Recently, knowledge graphs have been applied by large pharmaceutical companies to improve the efficiency

of drug discovery. Specifically, knowledge graphs based on drug ontology have been used for many purposes.

Current drug ontologies have different scopes, but mainly focus on the description of basic drug information.

Here, we describe a comprehensive pharmaceutical knowledge ontology, including information of active

ingredients, indications, inactive ingredients, drugs, clinical trials, organs and tissues, literature, patents,

targets, therapeutics, and biomolecules. Using multiple data sources, we apply a seven-step method for

ontology modelling using Protégé software. A comprehensive pharmaceutical knowledge ontology model is

established to complete the knowledge representation of drug information. By means of ontology theory, the

pharmaceutical knowledge is modelled, standardized and networked, so as to clarify the knowledge structure

and quickly acquire related knowledge and logical relationships. In the future, knowledge graphs based on

this ontology could be helpful to deal with the dispersion, heterogeneity, redundancy and fragmentation of

medical big data, to share and integrate pharmaceutical data, and to provide a set of solutions for the

networked development of pharmaceutical knowledge.

1 INTRODUCTION

Drug discovery is a complex process with a

development cycle of 10-15 years and an average

research and development cost of $2.6 billion

(Wouters et al., 2020). In recent years, knowledge

graphs have been applied by large pharmaceutical

companies to improve the efficiency of drug

discovery; for example, AstraZeneca is applying

BenevolentAI for drug development for chronic

kidney disease (CKD) and idiopathic pulmonary

fibrosis (IPF), showing the application and

development prospect of knowledge graph

Contributed equally to this work.

https://orcid.org/0000-0003-0956-6556

https://orcid.org/0000-0002-4596-5303

https://orcid.org/0000-0002-0900-7220

https://orcid.org/0000-0003-0679-2275

https://orcid.org/0000-0003-3280-8445

https://orcid.org/0000-0003-0235-3419

https://orcid.org/0000-0002-2442-4735

https://orcid.org/0000-0002-1054-6440

Address correspondence to: Sheng Liu

technology in such tasks. Knowledge graphs based on

drug ontology have been used for many purposes,

such as comparative effectiveness research (Hanna et

al., 2013), adverse drug reactions (Cai et al., 2015;

Hur et al., 2018), and clinical data warehousing

(Podchiyska et al., 2010). RxNorm was created to

address a lack of standardization of drug names and

to make them interoperable by integrating drug terms

into a reference system (Nelson et al., 2020). RxNorm

currently integrates terminology information from

most drug knowledge base vendors (e.g., First

DataBank, Multum, Micromedex, Gold Standard), as

well as drug ingredients from standard terminology

168

Wang, P., Mao, Y., Song, W., Jiang, W., Liu, Y., Zheng, L., Ma, B., Sun, Q. and Liu, S.

A Comprehensive and Scientiﬁcally Accurate Pharmaceutical Knowledge Ontology based on Multi-source Data.

DOI: 10.5220/0011012100003123

In Proceedings of the 15th International Joint Conference on Biomedical Engineering Systems and Technologies (BIOSTEC 2022) - Volume 3: BIOINFORMATICS, pages 168-175

ISBN: 978-989-758-552-4; ISSN: 2184-4305

(e.g., SNOMED CT, MeSH). RxNorm focuses on

drug names and codes; however, clinical information

and administrative information are out of scope

(Bodenreider et al., 2018). Based on the RxNorm

drug terminology and the Chemical Entities of

Biological Interest ontology (ChEBI), Hanna et al.

(2013) built Drug Ontology (DrOn), a modular,

extensible body of drugs, ingredients, and biological

activities, originally created to enable comparative

effectiveness. OCRx is a Canadian drug ontology

system built to provide a normalized and standardized

description of drugs authorized to be marketed in

Canada. OCRx is focused on clinical drug description

(i.e., substance, strength, route of administration,

pharmaceutical form) (Nikiema et al., 2021). Sharp

(2017) created a drug-indication database (DID), a

database of structured drug-indication relations

intended to facilitate building practical,

comprehensive, and integrated drug ontologies.

Current drug ontologies have different scopes, for

example, RxNorm and OCRx describe drugs

available in the U.S. and Canada, respectively, and

mainly focusing on the description of basic drug

information. However, the current lack of

comprehensive pharmaceutical knowledge ontology

covering drug targets, adverse drug reactions, clinical

trials, patents, literature and other classes limits the

application of drug ontology.

In this paper, we describe a comprehensive drug

knowledge ontology, including active ingredients,

indications, inactive ingredients, drugs, clinical trials,

organ and tissue targets, literature, patents,

therapeutics, and biomolecules. We use ontology

theory to model, standardize and network

pharmaceutical knowledge, which is convenient for

clarifying the knowledge structure, quickly obtaining

relevant knowledge and logical relationships, and

helping to deal with the dispersion, heterogeneity,

redundancy and fragmentation of medical big data.

The purpose of this ontology is to aid the sharing and

interaction of pharmacy data and provides a set of

solutions for the development of pharmacy

knowledge network.

2 METHODS

Ontology theory has been widely used for knowledge

presentation; common building methods are seven-

step method, Skeletal Methodology, IDEF5, and

METHONTOLOGY. Based on a comprehensive

analysis of the existing approaches to the construction

of drug ontology, we applied a seven-step method

developed by Stanford University School of

Medicine to construct a pharmaceutical knowledge

ontology using Protégé software (Musen et al., 2015).

The construction process of pharmaceutical

knowledge ontology is shown in Figure 1. Ontology

modelling is based on hierarchical and structured

biomedical thesauri such as MeSH and ICD,

combined with authoritative data sources such as

Drugbank and PubChem, to develop a standardized

glossary and glossary of pharmaceutical ontologies to

unify and integrate multi-source data and facilitate

inter-term relationships. The next step is to refer to

the results of the analysis of the web site data, record

the conceptual level, format, and data type of the

entity in the data source for the pharmacy

terms/entities identified in the data. For relevant text

of an entity that cannot be directly identified, entity

name recognition is carried out through the biological

named entity recognition tool and the standardized

terminology. After the entity is identified, the

relationship between entities is recognized by the

machine learning method, summarizing the

properties and inter-entity relations of the new

entities, and classifying the entities to determine the

hierarchical structure of the data. Finally, experts

examine and verify the entities, properties and

relationship information, and then incorporate newly

identified entities and relationships into existing

models. The final realization uses the standard

terminology of ontology, annotates the entity, and

completes the knowledge representation.

2.1 Data Sources

Data sources are the basis of pharmaceutical

knowledge ontology modelling. Data acquired from

multiple databases and platforms is listed in Table 1.

MeSHDrugbank ICD PubChem

ATC FDA

ClinicalTrials

Registry

Data Source

ObjectpropertiesClasses Dataproperties

Knowledge

extraction

DrugOntology

Knowledge

Presentation

Figure 1: Construction process of pharmaceutical

knowledge ontology.

A Comprehensive and Scientiﬁcally Accurate Pharmaceutical Knowledge Ontology based on Multi-source Data

169

The following section provides an overview of the

multiple data sources used here.

Medical Subject Headings (MeSH) is an

authoritative thesaurus compiled by The United

States National Library of Medicine (NLM). MeSH

is a standardized and expandable dynamic thesaurus

of medical concepts and is used for indexing,

cataloging, and searching of biomedical and health-

related information. It includes the subject headings

appearing in MEDLINE/PubMed, the NLM Catalog,

and other NLM databases.

The International Classification of Diseases and

Related Health Problems (ICD) is a tool for

recording, reporting, and grouping conditions and

factors that influence health. It contains categories for

diseases, health-related conditions, and external

causes of illness or death. The ICD is used to translate

disease diagnoses into an alphanumeric code, which

allows data storage, retrieval, and analysis.

DrugBank is a comprehensive, free-to-access

online database containing information on drugs and

drug targets (Wishart et al., 2006). The latest release

(version 5.1.8, released 2021-01-03) contains 14,589

entries. 5,263 non-redundant protein (i.e., enzyme,

transporter, carrier) sequences are linked to these

entries. Each entry contains more than 200 data fields

with half of the information devoted to drug/chemical

data and the other half devoted to drug target or

protein data (Wishart et al., 2018).

PubChem is an open chemistry database by the

National Institutes of Health (NIH). It contains both

small and larger molecules such as nucleotides,

carbohydrates, lipids, peptides, and chemically

modified macromolecules. PubChem collects

information on chemical structures, identifiers,

chemical and physical properties, biological

activities, patents, health, safety, and toxicity (Kim et

al., 2019).

The Anatomical Therapeutic Chemical (ATC)

system is the official drug classification of the World

Health Organization (WHO). In the ATC system,

active substances are classified in a five-level

hierarchy.

China Medical Information Platform is an expert-

certified information platform for all types of medical

information, including disease and symptom queries,

drug introduction, drug instructions, hospital queries,

and expert queries.

Clinical Pathways were released by the National

Health Commission of the People's Republic of China

and contains clinical pathways for 224 disease species

in 19 disciplines.

U.S. Clinical Trials is a web-based resource

maintained by the NLM and NIH. Each record

presents summary information about a study protocol

and includes information such as disease or condition,

intervention, title, description, and study design. The

European Union Clinical Trials Register, Chinese

Clinical Trial Registry, China Drug Trials, and WHO

International Clinical Trials Registry Platform are

similar platforms for the registration of clinical trials.

Table 1: Data sources for ontology modelling.

No. Data Source URL of Data Source

1 MeSH

https://www.nlm.nih.gov

/databases/download/me

sh.html

2 ICD-10

https://icd.who.int/brows

e10/2019/en

3 ICD-11

https://icd.who.int/brows

e11/l-m/en

4 DrugBank

https://go.drugbank.com/

releases/latest

5 PubChem

https://ftp.ncbi.nlm.nih.g

ov/

ubchem

6 ATC

https://www.whocc.no/at

c_ddd_index

China Medical

Information Platform

https://www.dayi.org.cn/

8 Clinical Pathways

http://www.nhc.gov.cn/y

zygj

9 US Clinical Trials

https://www.clinicaltrial

s.gov/ct2/resources/dow

nloa

European Union

Clinical Trials Register

https://www.clinicaltrial

sregister.eu/ctr-search/

Chinese Clinical Trial

Registr

http://www.chictr.org.cn

/searchproj.aspx

China Drug Trials

http://www.chinadrugtri

als.or

.cn

International Clinical

Trials Re

istr

Platfor

https://trialsearch.who.in

14 FDA Orange Book

https://www.accessdata.f

da.gov/scripts/cder/ob/in

dex.cf

15 Japanese Orange Book

http://www.jp-orange

book.gr.jp/cgi-bin/

search_h/search_e.cgi

List of Reference

Preparations for Generic

Dru

(

China

)

https://www.nmpa.gov.c

n/xxgk/ggtg/qtggtg/inde

x.html

FDA Inactive

Ingredients Database

https://www.fda.gov/dru

gs/drug-approvals-and-

databases/inactive-

ingredients-database-

downloa

The publication Approved Drug Products with

Therapeutic Equivalence Evaluations (commonly

known as the Orange Book) identifies drug products

BIOINFORMATICS 2022 - 13th International Conference on Bioinformatics Models, Methods and Algorithms

170

approved on the basis of safety and effectiveness by

the Food and Drug Administration (FDA) under the

Federal Food, Drug, and Cosmetic Act and related

patent and exclusivity information. It contains

information regarding active ingredient, proprietary

name, applicant, application number, dosage form,

route of administration or patent number of approved

drug products. The Japanese Orange Book is a similar

guide published in Japan.

The List of Reference Preparations for Generic

Drugs published by the National Medical Products

Administration of China is the basis for registration

application and consistency evaluation of generic

drugs in China. Inquiry contents include generic

name, English name or trade name, license holder,

specifications, dosage form, remarks, sources and

other information.

The Inactive Ingredients Database provides

information on inactive ingredients present for FDA-

approved drug products. It contains information on

route of administration, dosage and dosage form,

CAS Number, UNII, potency amount and potency

unit on inactive ingredients.

2.2 Modelling of Classes

Based on hierarchical and structured biomedical

lexicon such as MeSH and ICD, combined with

relevant terms from authoritative data sources such as

Drugbank and PubChem, we developed a

standardized glossary and term annotations using a

pharmaceutical knowledge ontology.

13 concepts were extracted as first-level semantic

types (classes), named Drug, Indication, Therapy,

ClinicalTrial, Product, Food, InactiveIngredient,

Patent, Reference, Organ, Targets, Biomolecule and

Organism. Each class was extracted from one or more

of the corresponding data sources, shown in Table 2

(data source numbers are from Table 1).

Table 2: Extraction of the classes from data sources.

No. Class Data Source No.

1 Drug 1, 4, 5, 6

2 Indication 1, 2, 3, 4, 7

3 Thera

1, 8

4 ClinicalTrial 9, 10, 11, 12, 13

5 Product 4, 14, 15, 16

6 InactiveIngredient 17

7 Foo

8 Patent 4

9 Reference 4

10 Or

an 7

11 Targets 4

12 Biomolecule 4

13 Organism 1, 4

2.3 Modelling of Object Properties

and Data Properties

Sentence and word segmentation were executed on

data collected from multi-source heterogeneous

sources to generate pre-processed data. Name

information of a plurality of entities defined in the

concept layer was extracted from the pre-processed

data by the named entity recognition, and the name

information was contained in the entity statement of

the data. The relationship extraction method based on

the depth neural network identifies the relationships

between different entities defined in the concept layer

from the entity statements.

Specifically, with words as the basic units, the

entity statements were characterized by word and

position vectors, and the vector splicing results

corresponding to the entity statements are obtained.

These results were input into a deep neural network,

and the relationship classification were distinguished,

which comprises an attention model, a fully

connected neural network layer and a convolutional

neural network model. Focused on the deep neural

network, the entity sentences with relationships are

identified. Syntax was analyzed, and related words

were extracted according to the dependency relation

among entities, thus obtaining the cause-effect tuples,

which include a relationship between different

entities.

The property information of each entity was

modelled as data properties, and the relationship

between the entities is modelled as object properties.

Figure 2: Classes of pharmaceutical knowledge ontology.

A Comprehensive and Scientiﬁcally Accurate Pharmaceutical Knowledge Ontology based on Multi-source Data

171

3 RESULTS

3.1 Classes

After data cleansing and parsing, with the help of

experts, we extracted 13 classes from multiple data

source species. These classes are listed in Table 2.

With further analysis of the data sources, a total of 11

second-level semantic types (i.e., subclasses), were

identified in three classes: Targets, Biomolecules and

Organism. Finally, the classes and subclasses were

built in Protégé, shown as Figure 2.

3.2 Object Properties and Data

Properties

Using analysis of data sources, and verification by

experts, 11 object properties and 5 subproperties, as

well as 39 data properties and 43 subproperties were

determined. The detailed object properties and data

properties are listed in Tables 3 and 4, respectively.

Object properties are used to describe the

relationships between pharmaceutical knowledge

entities, each of which defines the subject and scope

of application, as shown in Table 3. Drugs, for

example, includes the following relationships: drug-

literature-citation, drug-clinical trial-effective effect,

drug-indication-active effect, drug-drug-interaction,

drug-drug-active ingredient, drug-target-effective

effect, drug-biomolecular-benign/adverse effect. In

addition, each entity includes the same basic

properties: hasName, hasDescription, hasEntityClass,

hasSynonyms, hasSource, and hasID.

3.3 Model of the Pharmaceutical

Knowledge Ontology

As with the classes, object properties and data

properties were determined. The data level, scope,

type and definition of each type of entity were

formulated to achieve a structured, standardized and

normalized description of pharmaceutical entities.

The ontology model of pharmaceutical knowledge

based on Protégé is shown in Figure 3. Relationships

between classes are represented by lines with arrows.

At this point, the basic architecture of the

ontology is complete, and entities can be added into

the ontology as individuals in Protégé manually, or by

using data association between multi-source data and

ontology model by means of technology.

Table 3: Object properties and subproperties defining the relationships between pharmaceutical entities.

No. Object property Subproperty Domains Ranges

1 isAIRelationOn

2 isCitationOf

Patent

Reference

Biomolecule

ClinicalTrial

Drug

InactiveIngredient

Indication

Product

Targets

3 isClinicalTrialOn

ClinicalTrial

Drug

Indication

Product

Therap

4 isDrugActionOn

Drug

Product

Indication

Thera

5 isIndicationLocationIn

Indication Organ

6 isInteractionWith

Drug

Biomolecule

Drug

7 isProductIngredientOf

isProductActiveIngredientOf Drug Product

isProductInactiveIngredientOf Inactive Ingredient Product

isProductAdverseReaction

Product

ClinicalTrial

Indication

Therap

9 isProductReferenceOf

Product Product

10 isTherapyOn

Therapy Indication

11 isTargetActionsOn

isTargetActionOn Drug Targets

isTargetOrganismOf Drug Organ

isTargetPharmacologicalActionOn Drug Targets

BIOINFORMATICS 2022 - 13th International Conference on Bioinformatics Models, Methods and Algorithms

172

Figure 3: Model of the pharmaceutical knowledge ontology.

4 DISCUSSION

Pharmaceutical knowledge ontology helps describe

and organize pharmaceutical information, and forms

a network of pharmaceutical knowledge about

pharmaceutical terms and the relationships between

them. Combined with computer technology,

pharmaceutical-related data can be shared and

exchanged in the network. Through the

standardization of terms in pharmaceutical

knowledge ontology, metadata from different data

sets can be unified to eliminate heterogeneity and

realize the integration of pharmaceutical data. At the

same time, through the relationship between the

standardized terms in the ontology, metadata in the

data set can also construct the semantic association

and realize the index of the metadata content, in order

to achieve deeper level conformity, annotation,

analysis and mining of the original data.

Knowledge base is a structured, operable and

organized cluster in knowledge engineering. With the

common demand of solving problems in a particular

domain, it is a set of interrelated knowledge slices that

are stored, organized, managed, and used in computer

memory in a certain knowledge representation

manner.

Ontology provides a basic architecture for the

establishment of knowledge base. It describes the

domain with a set of concepts and terms, and obtains

the essential conceptual structure of the domain.

Ontology constitutes the core of the domain

knowledge representation system. The knowledge

base uses these terms to represent information.

Ontology-based knowledge base can help users to

acquire knowledge most suitable to their needs

through these relationships and properties, thus

avoiding irrelevant information during knowledge

acquisition. Pharmaceutical knowledge ontology can

help realize the standardized description and

structured organization of pharmaceutical knowledge

and information, promote efficient use of

pharmaceutical data, and lay a foundation for

knowledge graph.

In building this ontology model, we have reached

a milestone. However, there are still much more to do

in the future. To reach the goal of knowledge graph,

further than the ontology model, we need to optimize

the data model and realize the data mapping from

ontology to relational data and graph database.

5 CONCLUSIONS

Based on the basic concepts and knowledge system

of pharmacy, using the idea of ontology modelling,

and referring to existing pharmaceutical, biological

and medical knowledge ontology models and

resource databases, we sort out the concept, scope,

classification, hierarchy and structure of

pharmaceutical knowledge ontology. We have built a

basic ontology model, which can be gradually

integrated and updated by analysing and sorting the

data content of authoritative global data sources.

Finally, a relatively complete pharmaceutical

knowledge ontology model was established to

complete the knowledge representation of drug

information. With this tool, we can not only show

pharmaceutical entities and their relationships, and

systematically describe pharmaceutical knowledge,

but also can integrate and supplement existing

biomedical ontologies, in order to further realize the

standardization, normalization and structurization of

pharmaceutical data in the form of knowledge graph.

ACKNOWLEDGEMENTS

This work was conducted using Protégé, which is

supported by grant GM10331601 from the National

Institute of General Medical Sciences of the United

States National Institutes of Health.

A Comprehensive and Scientiﬁcally Accurate Pharmaceutical Knowledge Ontology based on Multi-source Data

173

Table 4: Data properties and subproperties defining the basic attributes of pharmaceutical entities.

No. Data property Subproperty

1 hasApproved

2 hasCategory

3 hasClinicalTrial

hasClinicalTrialDrug

hasClinicalTrialIndication

hasClinicalTrialStatus

hasClinicalTrialThera

4 hasCountryName

5 hasDates

6 hasDeleted

7 hasDescription

8 hasDescription

hasDetailDescription

hasSummar

9 hasDrugChemicalIdentifier

hasDrugCas

hasDrugInChI

hasDrugInChIKey

hasDrugIupacName

hasDru

Unii

10 hasDrugPharmacology

11 hasDrugDissolution

12 hasDrugProperty

13 hasDrugSpectra

14 hasEntityClass

15 hasFunctions

16 hasGene

17 hasNames

hasGenericName

hasName

hasNonProprietaryName

hasPreferredName

hasS

non

18 hasID

19 hasIndicationDiagnosis

hasIndicationCause

hasIndicationCheck

hasIndicationChiefComplaint

hasIndicationDiagnosis

hasIndicationDiagnosisBasis

hasIndicationS

20 hasIndicationHospitalDepartment

21 hasIndicationPathway

22 hasIndicationSite

23 hasLink

24 hasFunctions

hasMainFunction

hasSpecificFunction

25 hasMedicalInsuranced

26 hasOrganisationName

27 hasOriginalID

28 hasProductAdverseReaction

29 hasProductCompany

hasProductApplicantHolde

hasProductDistributor

hasProductLabeller

hasProductManufacturer

hasProductPacka

30 hasProductBrand

31 hasProductCompany

32 hasProductInfo

hasProductDosage

hasProductRoute

hasProductStrength

33 hasProductInstruction hasProductSpecification

BIOINFORMATICS 2022 - 13th International Conference on Bioinformatics Models, Methods and Algorithms

174

Table 4: Data properties and subproperties defining the basic attributes of pharmaceutical entities (cont.).

No. Data property Subproperty

34 hasTitle

hasPublicTitle

hasScientificTitle

35 hasReferenceInfo

hasReferenceDoi

hasReferenceExternalDatabaseID

hasReferenceFile

hasReferenceLink

hasReferencePatentID

hasReferencePmi

36 hasRegistered

37 hasSource

38 hasDates

hasStartDate

hasU

dateTime

39 hasType

REFERENCES

Wouters, O. J., McKee, M., & Luyten, J. (2020). Estimated

Research and Development Investment Needed to

Bring a New Medicine to Market, 2009-

2018. JAMA, 323(9), 844–853.

Hanna, J., Joseph, E., Brochhausen, M., & Hogan, W. R.

(2013). Building a drug ontology based on RxNorm and

other sources. Journal of biomedical semantics, 4(1),

44.

Hur, J., Özgür, A., & He, Y. (2018). Ontology-based

literature mining and class effect analysis of adverse

drug reactions associated with neuropathy-inducing

drugs. Journal of biomedical semantics, 9(1), 17.

Cai, M. C., Xu, Q., Pan, Y. J., Pan, W., Ji, N., Li, Y. B., Jin,

H. J., Liu, K., & Ji, Z. L. (2015). ADReCS: an ontology

database for aiding standardization and hierarchical

classification of adverse drug reaction terms. Nucleic

acids research, 43(Database issue), D907–D913.

Podchiyska, T., Hernandez, P., Ferris, T., Weber, S., &

Lowe, H. J. (2010). Managing Medical Vocabulary

Updates in a Clinical Data Warehouse: An RxNorm

Case Study. AMIA ... Annual Symposium proceedings.

AMIA Symposium, 2010, 477–481.

Nelson, S. J., Zeng, K., Kilbourne, J., Powell, T., & Moore,

R. (2011). Normalized names for clinical drugs:

RxNorm at 6 years. Journal of the American Medical

Informatics Association: JAMIA, 18(4), 441–448.

Bodenreider, O., Cornet, R., & Vreeman, D. J. (2018).

Recent Developments in Clinical Terminologies -

SNOMED CT, LOINC, and RxNorm. Yearbook of

medical informatics, 27(1), 129–139.

Bona, J. P., Brochhausen, M., & Hogan, W. R. (2019).

Enhancing the drug ontology with semantically-rich

representations of National Drug Codes and RxNorm

unique concept identifiers. BMC bioinformatics, 20

(Suppl 21), 708.

Nikiema, J. N., Liang, M. Q., Després, P., & Motulsky, A.

(2021). OCRx: Canadian Drug Ontology. Studies in

health technology and informatics, 281, 367–371.

Sharp M. E. (2017). Toward a comprehensive drug

ontology: extraction of drug-indication relations from

diverse information sources. Journal of biomedical

semantics, 8(1), 2.

Musen, M. A., & Protégé Team (2015). The Protégé

Project: A Look Back and a Look Forward. AI

matters, 1(4), 4–12.

Wishart, D. S., Knox, C., Guo, A. C., Shrivastava, S.,

Hassanali, M., Stothard, P., Chang, Z., & Woolsey, J.

(2006). DrugBank: a comprehensive resource for in

silico drug discovery and exploration. Nucleic acids

research, 34(Database issue), D668–D672.

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., Assempour, N., Iynkkaran, I., Liu, Y.,

Maciejewski, A., Gale, N., Wilson, A., Chin, L.,

Cummings, R., Le, D., Pon, A., … Wilson, M. (2018).

DrugBank 5.0: a major update to the DrugBank

database for 2018. Nucleic acids research, 46(D1),

D1074–D1082.

Kim, S., Chen, J., Cheng, T., Gindulyte, A., He, J., He, S.,

Li, Q., Shoemaker, B. A., Thiessen, P. A., Yu, B.,

Zaslavsky, L., Zhang, J., & Bolton, E. E. (2019).

PubChem in 2021: new data content and improved web

interfaces. Nucleic Acids Res., 49(D1), D1388–D1395.

A Comprehensive and Scientiﬁcally Accurate Pharmaceutical Knowledge Ontology based on Multi-source Data

175