Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2024 Dec 15.
Published in final edited form as: J Chem Inf Model. 2019 Dec 12;59(12):4968–4973. doi: 10.1021/acs.jcim.9b00683

ROBOKOP KG AND KGB: Integrated Knowledge Graphs from Federated Sources

Chris Bizon †,*, Steven Cox †, James Balhoff †, Yaphet Kebede †, Patrick Wang ‡, Kenneth Morton ‡, Karamarie Fecho †, Alexander Tropsha #
PMCID: PMC11646564  NIHMSID: NIHMS2035816  PMID: 31769676

Abstract

A proliferation of data sources has led to the notional existence of an implicit Knowledge Graph (KG) that contains vast amounts of biological knowledge contributed by distributed Application Programming Interfaces (APIs). However, challenges arise when integrating data across multiple APIs due to incompatible semantic types, identifier schemes, and data formats. We present ROBOKOP KG (http://robokopkg.renci.org), which is a KG that was initially built to support the open biomedical question-answering application, ROBOKOP (Reasoning Over Biomedical Objects linked in Knowledge-Oriented Pathways) (http://robokop.renci.org). Additionally, we present the ROBOKOP Knowledge Graph Builder (KGB), which constructs the KG and provides an extensible framework to handle graph query over and integration of federated data sources.

Graphical Abstract

graphic file with name nihms-2035816-f0001.jpg

1. INTRODUCTION

Data sources increasingly expose information via Application Programming Interfaces (APIs) that allow for structured query of biomedical knowledge. APIs may be based on experimental or curated data, collected in a variety of formats, and used to answer questions of interest to biomedical investigators, such as: what genes are involved with type 2 diabetes mellitus? what cell types are located in cardiac tissue? what diseases are associated with exposure to benzene? Answers to these queries define a relationship between the input entity and the output entity. If this relationship is considered to be an edge in a Knowledge Graph (KG),1,2 then a distributed set of such APIs constitutes an implicit, federated KG, which then can be queried.3

Large, arbitrarily complex graph queries, ones that combine calls to multiple APIs and use results from one API as inputs to another, can be performed in principle, but any attempt to do so must deal with several challenges, including inconsistencies in data formats, identifier schemes, and semantic types.

Here, we present ROBOKOP Knowledge Graph Builder (KGB) and ROBOKOP KG, which are, respectively, software code that addresses these challenges and the database that is created by this code. These tools were initially created to support the open question-answering application ROBOKOP (Reasoning Over Biologically Oriented Knowledge Outcome Pathways)4, which is hosted at http://robokop.renci.org and supported by the National Center for Advancing Translational Sciences, as part of the Biomedical Data Translator (‘Translator’) program.5 However, both tools are independently useful. ROBOKOP KGB builds KGs in response to user-specified graph queries by parsing the queries into a series of type-driven API calls, between which the identifiers used in different systems are reconciled with one another in a process called “synonymization”. A general-purpose open biomedical ROBOKOP KG has been created by crawling over a large number of simple questions and combining the results into the single graph hosted at http://robokopkg.renci.org.

2. IMPLEMENTATION

2.1. BioLink Data Model and Type Hierarchy.

ROBOKOP KG is based on a set of semantic types, as defined in the BioLink data model.6 (A simplified version of the ROBOKOP KG database schema is provided as a graphical abstract, with nodes representing entities in the BioLink data model, and edges representing predicates or relationships between connected entities.) This model defines the high-level concepts between which relationships can be made, as well as a series of properties belonging to these concepts. The BioLink data model is hierarchical, with more specific concepts deriving from more general ones; for instance “cellular component” is derived from “anatomical entity”. The model also contains union terms that provide capabilities to, for example, group diseases and phenotypes. Entities are identified with conceptual terms from biomedical ontologies such as Gene Ontology (GO)7. ROBOKOP KG requires all nodes to be an instance of one or more BioLink data model types.

2.2. APIs and Clients.

Individual knowledge sources are defined by APIs that take as input an entity (e.g., a disease) and return information about related entities (e.g., genes), as well as the nature of the relationship between them. Many formats of such APIs exist, including Smart APIs8 and SPARQL or SQL endpoints. ROBOKOP KGB does not require a particular style or format for the APIs that it queries in order to allow for maximum generality in the queryable data sources. However, this approach requires that a client must be coded for each knowledge source. Data for which only flat files or RDF data can be acquired may be exposed using tools such as SmartBag9 or data2services.10 Because results are returned using types represented in the BioLink data model, we are able to create and apply mappings from the types returned from individual APIs to those contained in the BioLink data model.

The ROBOKOP KGB paradigm differs from earlier graph builders11-13 in its online approach. Previous approaches rely on downloading bulk datasets, which are curated and integrated together in bulk. In contrast, ROBOKOP KGB can respond to queries in which the data is fully federated, as long as an API exists to access it.

2.3. Concept Map.

To calculate the services that must be called in order to produce answers, individual clients are annotated with their input and output types, thus forming a graph in which nodes represent types from the BioLink data model and edges represent the APIs that provide relationships between entities of those types. In the creation of this concept map, the BioLink data model allows inference of transitions involving entity superclasses and subclasses. For instance, if an API is annotated as returning entities of type “biological_process”, then the model implies that output entities are also members of the superclass “biological_process_or_activity”. Inferences can also proceed in reverse to identify subclasses.

2.4. Query Specification.

Queries are specified using the Translator KG API standard specification (https://github.com/NCATS-Tangerine/NCATS-ReasonerStdAPI/), which defines a JSON-based representation of the requested graph. A node in the question graph represents an entity type; edges represent a desired relationship between entities of those types. Responses to the query will be in the form of a graph of instances that match the query in topology and entity types. Any node in the query graph may be bound to an identifier, so that any answer will contain only the entity specified at the corresponding location in the answer graph. Edge types can be specified to limit the allowed relationships between entities. Alternatively, if multiple edge types are specified, then edges that match any of the specified types will be returned.

2.5. Identifiers and Synonymization.

APIs are generated by independent data providers and use different identifier systems for the entities they describe. The integration of APIs thus requires a reconciliation of different identifiers that represent the same entity. ROBOKOP KGB creates and maintains a set of equivalent identifiers and stores them in a redis database for rapid retrieval during integration. Equivalent identifier sets are created per entity type. Where possible, we make use of pre-existing equivalences; for example, genes are synonymized using mappings from HGNC14 and diseases are synonymized using equivalentTo and skos#exactMatch relationships from the Monarch Disease Ontology15. Chemicals are iteratively identified starting with the use of InchiKey, as found in UniChem16, then moving to SMILES for chemicals that are not expressible with InchiKey. For chemical substances lacking any structure, database cross-references are used (including the registry relationship in the case of MeSH17). Other types are synonymized using database cross-reference tags. Because such tags are inconsistently applied in different ontologies, specialized rules are crafted to prevent over-eager identification, using pairs of vocabularies such as Human Phenotype Ontology and International Classification of Diseases18. Given a set of equivalent identifiers, one is chosen as the node identifier for the output graph, using the vocabulary defined in the BioLink data model6.

2.6. Traversal.

Query graphs are mapped to the concept transition model by matching query edges into model transitions and, thereby, into data source invocations. The traversal collects the minimal subgraph of federated data that contains every answer to the query. The transition model allows multiple paths per transition; it differs from a strict nondeterministic finite automaton (NFA) in that start and end states are not explicitly modeled, and transitions entailing ontological relationships are implicit. An accepted traversal of the NFA for a question compiles a stack of service invocations, the execution of which invokes services that produce triple instances that ontologically conform to the query pattern. Queries that are not traversals of the NFA are invalid with respect to the grammar and are rejected. ROBOKOP KGB manages results across service calls by resolving identifier synonyms, managing equivalent identifiers in subsequent calls, and caching results in a Redis key-value store in order to improve performance.

2.7. Integrated Database.

The output from a query is written to a Neo4j database that integrates the data returned from all previous queries. Running multiple queries allows both the Neo4j and Redis caches to grow though query answers. Alternatively, a set of queries such as “return every gene-disease relationship” may be asked systematically, thereby effectively crawling all of the data sources, integrating the data, and storing the data in the Neo4j and Redis databases. The output of such a crawling campaign is ROBOKOP KG. ROBOKOP KG consists of data drawn either directly or indirectly from numerous data sources, including DrugBank, 19 DrugCentral,20 Aeolus,21 Comparative Toxicogenomics Database,22 PubChem,23 Panther,24 UniChem,16 ChEMBL,25 Chemical Entities of Biological Interest,26 mychem.info,27 Monarch,28 Monarch Disease Ontology,15 Human Phenotype Ontology,29 GO,30 QuickGO,31 AmiGO,32 Pharos,33 ClinGen,34 ClinVar,35 GWAS Catalog,36 Kyoto Encyclopedia of Genes and Genomes,37 mygene.info,38 myvariant.info,38 ensembl,39 Human Metabolome Database,40 UniProt Knowledgebase,41 and bio2RDF.42 ROBOKOP KG currently contains approximately 500,000 nodes and 12 million edges. A simplified version of the database schema is provided in the graphical abstract.

2.8. Example Queries.

Queries of the ROBOKOP KG can be made using the cypher query language43, which supports queries of arbitrary patterns. The ROBOKOP KG can also be queried via the ROBOKOP application user interface (UI), which provides an additional analytic algorithm that scores and ranks results.

2.8.1. Direct Cypher Queries of the ROBOKOP KG.

Case Example 1: Linear Queries of Entity Relationships.

Simple linear queries, with one or more interior nodes, are a common use case for ROBOKOP KG. An example “two-hop” cypher query of the ROBOKOP KG asks: what cells are part of the brain and related to the biological process of thyroid stimulating hormone secretion? The query itself is structured as a linear chain as follows:

anatomical entity (brain) – cell – biological process or activity (thyroid stimulating hormone secretion)

When executed, ROBOKOP KG returns a single cell type, thyrotroph. Here, the entities at either end of the chain are specified, and the query is intended to find unknown concepts that relate them.

Case Example 2: Mechanistic Identification of Potential Disease Treatments.

Figure 1 provides a graphical representation of a more complex cypher query that aims to determine potential treatments for irritable bowel syndrome (IBS) by exploiting relationships between the specified disease and contributing genes and chemical substances. Here, a user has hypothesized a relationship between chemicals that worsen IBS, disease-related intermediary genes, and chemicals that act on those genes in a manner that is opposite to that of chemicals that worsen IBS. Specifically, the query asks: what chemical substances contribute to IBS that also increase the activity of genes involved in IBS, and are there drugs that decrease the activity of those same genes (and thus may be useful in the treatment of IBS)? In addition, the query includes a constraint in that the chemical-gene pairs must share the same biological process. While this constraint is not strictly necessary to implement the query, it was included in order to reduce the number of spurious answers. The execution of this query in ROBOKOP KG identifies serotonin as a contributing chemical, HTR3A as a shared gene, and serotonin receptor signaling pathway as a biological process that relates serotonin to HTR3A. Sixteen potential drugs are also returned. A review of the associated metadata from Aeolus21 indicates that one of the 16 drugs (alosetron) is an approved treatment for IBS and four drugs (6alpha-methylprednisolone, aripiprazole, clozapine, and dexamethasone) have IBS annotated as an adverse event that occurs less frequently than would occur by chance, thus indicating potential therapeutic activity.

Figure 1.

Figure 1.

Open in a new tab

A graphical representation of a complex cypher query of the ROBOKOP KG. The query (available in the user guide at robokopkg.renci.org) searches for potential treatments for IBS by looking for drugs that act on disease-related genes in a manner that is opposite to that of disease-contributing chemical substances.

Case Example 3: Using Common Disease to Discover Drugs for Repurposing for Rare Disease.

Figure 2 provides a graphical representation of a cypher query that aims to identify drugs that treat a common disease and are repurposing candidates for the treatment of a rare disease. Specifically, the query asks: are there drugs for common diseases that share a gene and related biological process with the rare disease chronic granulomatosis? When the query is executed in ROBOKOP KG, we find that psoriatic arthritis is one common disease that is linked to chronic granulomatosis disease via the biological process inflammatory response and the CYBB gene. The drug that is returned is folic acid. A review of the associated metadata from Comparative Toxicogenomics Database22 establishes an interaction between folic acid and methotrexate. A subsequent Google search of “folic acid” and “chronic granulomatosis” identifies several sources(e.g., 44) that support the use of folic acid to reduce the side effects of methotrexate in the treatment of a subtype of chronic granulomatosis, namely, granulomatosis with polyangiitis.

Figure 2.

Figure 2.

Open in a new tab

A graphical representation of a complex cypher query of the ROBOKOP KG. The query (available in the user guide at robokopkg.renci.org) looks for putative rare disease treatments among treatments for biologically-related common diseases, i.e., common diseases that share both a gene and a biological process with the input rare disease.

2.8. ROBOKOP Queries of the ROBOKOP KG.

In addition to direct cypher queries of the ROBOKOP KG, the ROBOKOP KG can be used as a data provider to a series of specific or general query UIs, including the application ROBOKOP, which applies a complex scoring and ranking algorithm to the answer subgraphs that are returned to users.4

Case Example 4: Mechanistic Explanations for Unexpected Clinical Observations.

A recent query that was posed to ROBOKOP was driven by a real-world clinical observation. Specifically, an experimental trial of inhaled isopropyl alcohol, suggested by literature evidence,45,46 was found to be effective in the treatment of cyclic vomiting syndrome. The medical care team involved in this case wanted to identify the potential mechanism of action for isopropyl alcohol, so the following query was posed to ROBOKOP via the ROBOKOP UI: why is isopropyl alcohol effective in the treatment of cyclic vomiting disorder? The machine question was in the form of a graph query aimed at finding a linear path that starts with isopropyl alcohol and ends with nausea, while traversing a specified set of node types in between, as follows:

isopropyl alcohol -> gene -> biological process or activity -> cell -> anatomical entity -> nausea.

Multiple ranked answer subgraphs are returned in response to the query, with each subgraph representing a different path through the ROBOKOP KG. The answer subgraphs are scored and ranked using an algorithm that factors in the number of supporting PubMed publications, as well as indirect support provided by literature co-occurrence of all pairwise sets of terms in the path.4 One of the top answer subgraphs for this query, shown in Figure 3, suggests that isopropyl alcohol may be acting via the Fos gene to alter the response of hepatocytes to toxic substances and indirectly decrease nausea. The medical care team is now actively exploring this hypothesized mechanism of action.

Figure 3.

Figure 3.

Open in a new tab

One of the top answer subgraphs returned by ROBOKOP in response to the following query: isopropyl alcohol -> gene -> biological process or activity -> cell -> anatomical entity -> nausea. Of note, users can interact with the graph and explore the supporting publications, which can be accessed by clicking on an edge. Users can also click on an edge to determine the provenance of predicate assertions between two nodes and associated metadata or click on a node to retrieve metadata, as shown here for “response to toxic substance”.

3. CONCLUSION

ROBOKOP is currently supporting knowledge discovery as part of the Translator program, but we anticipate that the wide range of knowledge integrated into the ROBOKOP KG will have broad applicability outside of the Translator program, as our approach of rational data integration, modularity, and well-designed interfaces allows innumerable applications to be built from ROBOKOP components.

The modular design of ROBOKOP allows us to release the ROBOKOP KG independent of the reasoning, inference, and other algorithms that interpret the dataset and are accessible via the ROBOKOP application. We are hosting a read-only version of the database at robokopkg.renci.org and regularly updating it with new data sources. We are also provisioning the full database as an export with which users can easily create their own local instance. The user guide, located at the ROBOKOP KG website, contains further information about the database, sample queries, similarity searches, and other information.

ACKNOWLEDGMENTS

This work was supported by the National Center for Advancing Translational Sciences, National Institutes of Health [grant number OT2TR002514 to A.T.]. The authors thank Matt Brush and Chris Mungall for creation of the BioLink data model, Eric Deutch and David Koslicki for contributions to the Translator KG API specification, Richard Bruskiewich for advice on services and useful discussions, and Matt Might and Will Byrd for feedback on the ROBOKOP KG and the isopropyl alcohol use case.

ABBREVIATIONS

API

application programming interface

IBS

irritable bowel syndrome

KG

knowledge graph

KGB

knowledge graph builder

NGA

nondeterministic finite automaton

ROBOKOP

Reasoning Over Biomedical Objects linked in Knowledge-Oriented Pathways

UI

user interface

Footnotes

Supporting Information

ROBOKOP KG, including downloadable copies and example queries, is available at http://robokopkg.renci.org/. ROBOKOP KGB is openly available under the MIT software license from https://github.com/NCATS-Gamma/robokop-interfaces.

The authors declare no competing financial interest.

REFERENCES

  • (1).Singhal A. Introducing the Knowledge Graph: things, not strings https://www.blog.google/products/search/introducing-knowledge-graph-things-not/ (accessed Jul 13, 2019). [Google Scholar]
  • (2).Wilcke X; Bloem P; De Boer V The Knowledge Graph as the Default Data Model for Learning on Heterogeneous Knowledge. Data Science 2017, 1 (1-2), 39–57. [Google Scholar]
  • (3).Collarana D; Galkin M; Traverso-Ribon I; Lange C; Vidal M-E; Auer S Semantic Data Integration for Knowledge Graph Construction at Query Time. In 2017 IEEE 11th International Conference on Semantic Computing (ICSC); IEEE, 2017; pp 109–116. [Google Scholar]
  • (4).Morton K; Wang P; Bizon C; Cox S; Balhoff J; Kebede Y; Fecho K; Tropsha A ROBOKOP: An Abstraction Layer and User Interface for Knowledge Graphs to Support Question Answering. Bioinformatics 2019. 10.1093/bioinformatics/btz604. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (5).Biomedical Data Translator Consortium. Toward A Universal Biomedical Data Translator. Clin. Transl. Sci 2019, 12 (2), 86–90. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (6).biolink-model https://biolink.github.io/biolink-model/ (accessed Jul 13, 2019).
  • (7).Ashburner M; Ball CA; Blake JA; Botstein D; Butler H; Cherry JM; Davis AP; Dolinski K; Dwight SS; Eppig JT; et al. Gene Ontology: Tool for the Unification of Biology. The Gene Ontology Consortium. Nat. Genet 2000, 25 (1), 25–29. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (8).Zaveri A; Dastgheib S; Wu C; Whetzel T; Verborgh R; Avillach P; Korodi G; Terryn R; Jagodnik K; Assis P; et al. smartAPI: Towards a More Intelligent Network of Web APIs. In The Semantic Web; Springer International Publishing, 2017; pp 154–169. [Google Scholar]
  • (9).smartBag; Github. [Google Scholar]
  • (10).Emonet V; Malic A; Zaveri A; Grigoriu A; Dumontier M Data2Services: Enabling Automated Conversion of Data to Services, 2018. 10.6084/m9.figshare.7345868.v1. [DOI] [Google Scholar]
  • (11).Chen B; Dong X; Jiao D; Wang H; Zhu Q; Ding Y; Wild DJ Chem2Bio2RDF: A Semantic Framework for Linking and Data Mining Chemogenomic and Systems Chemical Biology Data. BMC Bioinformatics 2010, 11, 255. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (12).Himmelstein DS; Baranzini SE Heterogeneous Network Edge Prediction: A Data Integration Approach to Prioritize Disease-Associated Genes. PLoS Comput. Biol 2015, 11 (7), e1004259. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (13).Womack F; McClelland J; Koslicki D Leveraging Distributed Biomedical Knowledge Sources to Discover Novel Uses for Known Drugs. bioRxiv, 2019, 765305. 10.1101/765305. [DOI] [Google Scholar]
  • (14).Povey S; Lovering R; Bruford E; Wright M; Lush M; Wain H The HUGO Gene Nomenclature Committee (HGNC). Hum. Genet 2001, 109 (6), 678–680. [DOI] [PubMed] [Google Scholar]
  • (15).Mondo Disease Ontology http://www.obofoundry.org/ontology/mondo.html (accessed Jul 13, 2019).
  • (16).Chambers J; Davies M; Gaulton A; Hersey A; Velankar S; Petryszak R; Hastings J; Bellis L; McGlinchey S; Overington JP UniChem: A Unified Chemical Structure Cross-Referencing and Identifier Tracking System. J. Cheminform 2013, 5 (1), 3. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (17).Medical Subject Headings - Home Page. 2019.
  • (18).ICD-11 https://icd.who.int/en (accessed Nov 21, 2019).
  • (19).Wishart DS; Feunang YD; Guo AC; Lo EJ; Marcu A; Grant JR; 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 (D1), D1074–D1082. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (20).Ursu O; Holmes J; Knockel J; Bologa CG; Yang JJ; Mathias SL; Nelson SJ; Oprea TI DrugCentral: Online Drug Compendium. Nucleic Acids Res. 2017, 45 (D1), D932–D939. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (21).Banda JM; Evans L; Vanguri RS; Tatonetti NP; Ryan PB; Shah NH A Curated and Standardized Adverse Drug Event Resource to Accelerate Drug Safety Research. Sci Data 2016, 3, 160026. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (22).Davis AP; Grondin CJ; Johnson RJ; Sciaky D; McMorran R; Wiegers J; Wiegers TC; Mattingly CJ The Comparative Toxicogenomics Database: Update 2019. Nucleic Acids Res. 2019, 47 (D1), D948–D954. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (23).Kim S; Chen J; Cheng T; Gindulyte A; He J; He S; Li Q; Shoemaker BA; Thiessen PA; Yu B; et al. PubChem 2019 Update: Improved Access to Chemical Data. Nucleic Acids Res. 2019, 47 (D1), D1102–D1109. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (24).Mi H; Muruganujan A; Ebert D; Huang X; Thomas PD PANTHER Version 14: More Genomes, a New PANTHER GO-Slim and Improvements in Enrichment Analysis Tools. Nucleic Acids Res. 2019, 47 (D1), D419–D426. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (25).Gaulton A; Hersey A; Nowotka M; Bento AP; Chambers J; Mendez D; Mutowo P; Atkinson F; Bellis LJ; Cibrián-Uhalte E; et al. The ChEMBL Database in 2017. Nucleic Acids Res. 2017, 45 (D1), D945–D954. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (26).Hastings J; Owen G; Dekker A; Ennis M; Kale N; Muthukrishnan V; Turner S; Swainston N; Mendes P; Steinbeck C ChEBI in 2016: Improved Services and an Expanding Collection of Metabolites. Nucleic Acids Res. 2016, 44 (D1), D1214–D1219. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (27).MyChem.info | The same high-performance BioThings API for chemicals and drugs https://mychem.info/ (accessed Jul 13, 2019).
  • (28).Biolink-Api; Github. [Google Scholar]
  • (29).Köhler S; Carmody L; Vasilevsky N; Jacobsen JOB; Danis D; Gourdine J-P; Gargano M; Harris NL; Matentzoglu N; McMurry JA; et al. Expansion of the Human Phenotype Ontology (HPO) Knowledge Base and Resources. Nucleic Acids Res. 2019, 47 (D1), D1018–D1027. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (30).The Gene Ontology Consortium. The Gene Ontology Resource: 20 Years and Still GOing Strong. Nucleic Acids Res. 2019, 47 (D1), D330–D338. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (31).Binns D; Dimmer E; Huntley R; Barrell D; O’Donovan C; Apweiler R QuickGO: A Web-Based Tool for Gene Ontology Searching. Bioinformatics 2009, 25 (22), 3045–3046. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (32).Carbon S; Ireland A; Mungall CJ; Shu S; Marshall B; Lewis S; AmiGO Hub; Web Presence Working Group. AmiGO: Online Access to Ontology and Annotation Data. Bioinformatics 2009, 25 (2), 288–289. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (33).Nguyen D-T; Mathias S; Bologa C; Brunak S; Fernandez N; Gaulton A; Hersey A; Holmes J; Jensen LJ; Karlsson A; et al. Pharos: Collating Protein Information to Shed Light on the Druggable Genome. Nucleic Acids Res. 2017, 45 (D1), D995–D1002. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (34).Pawliczek P; Patel RY; Ashmore LR; Jackson AR; Bizon C; Nelson T; Powell B; Freimuth RR; Strande N; Shah N; et al. ClinGen Allele Registry Links Information about Genetic Variants. Hum. Mutat 2018, 39 (11), 1690–1701. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (35).Landrum MJ; Lee JM; Benson M; Brown GR; Chao C; Chitipiralla S; Gu B; Hart J; Hoffman D; Jang W; et al. ClinVar: Improving Access to Variant Interpretations and Supporting Evidence. Nucleic Acids Res. 2018, 46 (D1), D1062–D1067. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (36).Buniello A; MacArthur JAL; Cerezo M; Harris LW; Hayhurst J; Malangone C; McMahon A; Morales J; Mountjoy E; Sollis E; et al. The NHGRI-EBI GWAS Catalog of Published Genome-Wide Association Studies, Targeted Arrays and Summary Statistics 2019. Nucleic Acids Res. 2019, 47 (D1), D1005–D1012. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (37).Kanehisa M; Goto S KEGG: Kyoto Encyclopedia of Genes and Genomes. Nucleic Acids Res. 2000, 28 (1), 27–30. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (38).Xin J; Mark A; Afrasiabi C; Tsueng G; Juchler M; Gopal N; Stupp GS; Putman TE; Ainscough BJ; Griffith OL; et al. High-Performance Web Services for Querying Gene and Variant Annotation. Genome Biol. 2016, 17 (1), 91. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (39).Zerbino DR; Achuthan P; Akanni W; Amode MR; Barrell D; Bhai J; Billis K; Cummins C; Gall A; Girón CG; et al. Ensembl 2018. Nucleic Acids Res. 2018, 46 (D1), D754–D761. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (40).Wishart DS; Tzur D; Knox C; Eisner R; Guo AC; Young N; Cheng D; Jewell K; Arndt D; Sawhney S; et al. HMDB: The Human Metabolome Database. Nucleic Acids Res. 2007, 35 (Database issue), D521–D526. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (41).UniProt Consortium. UniProt: A Worldwide Hub of Protein Knowledge. Nucleic Acids Res. 2019, 47 (D1), D506–D515. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (42).Callahan A; Cruz-Toledo J; Ansell P; Dumontier M Bio2RDF Release 2: Improved Coverage, Interoperability and Provenance of Life Science Linked Data. In The Semantic Web: Semantics and Big Data; Springer Berlin Heidelberg, 2013; pp 200–212. [Google Scholar]
  • (43).Cypher Query Language Developer Guides & Tutorials https://neo4j.com/developer/cypher-query-language/ (accessed Jul 13, 2019).
  • (44).Granulomatosis with polyangiitis (GPA) https://www.arthritis.org/about-arthritis/types/granulomatosis-with-polyangiitis/ (accessed Aug 13, 2019).
  • (45).Wang SM; Hofstadter MB; Kain ZN An Alternative Method to Alleviate Postoperative Nausea and Vomiting in Children. J. Clin. Anesth 1999, 11 (3), 231–234. [DOI] [PubMed] [Google Scholar]
  • (46).Pellegrini J; DeLoge J; Bennett J; Kelly J Comparison of Inhalation of Isopropyl Alcohol vs Promethazine in the Treatment of Postoperative Nausea and Vomiting (PONV) in Patients Identified as at High Risk for Developing PONV. AANA J. 2009, 77 (4), 293–299. [PubMed] [Google Scholar]

RESOURCES