Increasing the nanopublication recall with a BridgeDb Identi er Mapping Service

Abstract

The volume of literature in the life sciences is continuously growing and keeping up with it is a problem. While review articles and databases help us by summarizing vast amounts of research, dissemination of core research outcomes is still mostly restricted to scholarly journal. Nanopublications have been proposed as a solution to capture scienti c statements. This led to a 2010 proposal to serialize nanopubs in the Resource Description Framework (RDF) and in 2016 to an international network of nanopublication servers. However, RDF has a limitation that the Internationalized Resource Identi er (IRI) for resources does not have to be normalized and unique. To overcome this issue, the Open PHACTS project developed an Identi er Mapping Service and an approach called scienti c lenses for mapping of equivalent IRIs. We here demonstrate the application of this approach to improve the recall of nanopublications from the international network.

Keywords

• nanopublications identi er BridgeDb

-

The volume of literature in the life sciences is continuously growing and keeping up with literature is a problem [ 1 ]. While literature reviews and databases summarize vast amounts of research, dissemination of core research outcomes is still mostly restricted to the written word. Nanopublications have been proposed to capture scienti c statements with provenance to the origin of that statement [ 2 ]. This led to a 2010 proposal to serialize nanopubs in the Resource Description Framework (RDF) [ 3 ]. Kuhn et al. introduced in 2016 an international network of connected servers to host nanopublications, with initial data sets from DisGeNET [ 4 ], neXtProt [ 5 ], and others [ 6 ]. Nanopublications for WikiPathways were added later [ 7 ]. The network currently hosts about 10 million nanopublications [ 8 ].

However, RDF has a limitation that the Internationalized Resource Identi er (IRI) for resources does not have meaning and do not have to be unique. This causes an in nite number of possible IRIs for the same resource. And because when creating RDF content, one is not meant to reuse domain names not under your control and one often wishes to make resource IRIs dereferenceable, this is exactly what we see in practise: di erent data sets use di erent IRIs for the same gene, protein, or metabolite.

The Open PHACTS project has had the same problem when starting to link di erent pharmacology data sets [ 9 ], including ChEMBL [ 10 ] and WikiPathways [ 11 ]. To overcome this issue, it developed an Identi er Mapping Service (IMS) based on BridgeDb [ 12 ] and the scienti c lenses approach that allowed mapping of equivalent IRIs [ 13,14 ]. The alternative, of course, is to normalize IRIs in data sets before the integration, e.g. with identi ers.org [ 15 ], but removes exibility to change the level of equivalence depending on the data analysis done [ 13,14 ]. The IMS implemented the identi er mapping as integral part of the Open PHACTS Linked Data API, hiding this need to map equivalent IRIs when querying the underlying data sets.

Therefore, the assumption is that if we use an IMS as outlined here with appropriate loaded scienti c lenses, we will nd more nanopubs for a particular biological entity. To test this hypothesis, we searched nanopubs with information about a set of genes on the international network of nanopublication servers. 2

Methods

To implement our work ow, an R Markdown document was developed to perform the various steps detailed below. The full document is available at https:// github.com/egonw/swat4hcls2018/. It uses a few R packages and introduces a helper function to simplify searching nanopublications.

Genes As representative data set we took two pathways from the list of most viewed WikiPathways (https://www.wikipathways.org/index.php/Special: PopularPathwaysPage). Pathway WP241 was selected, about the human one carbon metabolism [ 16 ], with mostly NCBI Gene identi ers. WP2059 was selected as a second example, with predominantly Ensembl gene identi ers [ 17 ]. The rrdf package was used to query all genes in this pathways from the SPARQL endpoint [ 18 ].

Nanopublication Server As source of nanopubs we took a nearby server from the international network of mirroring servers, hosted by the Institute for Data Sciences at Maastricht University (http://graphdb.dumontierlab.com/ repositories/nanopubs) [ 6 ]. To simplify the interaction with the server, we took advantage of an online running instance of grlc (grlc.io) [ 19 ] that wraps the nanopublication server API with an OpenAPI [ 20 ]: https://github.com/ peta-pico/nanopub-api.

BridgeDb Identi er Mapping Server For the IRI mapping, we used the Docker image of the IMS developed by the Open PHACTS project (https: //hub.docker.com/r/openphacts/identitymappingservice/) [ 12,13,9 ]. This was recently repurposed by Ehrhart et al. for gene-variant mappings [ 21 ]. The IMS is started and loaded with IRI mapping data as explained in [ 21 ]: 1. the user starts to Docker image, and then 2. loads the identi er mapping les into the IMS instance using a loading script. The actual mapping les were generated by J. Melius based on data from Ensembl 87 (see http://bridgedb.org/data/ linksets/current/HomoSapiens/. The source code can be found at https: //github.com/BiGCAT-UM/EnsemblLinksetsCreator). These linksets map Ensembl identi ers with NCBI Gene, HGNC, and others.

Data Analysis The R Markdown notebook integrates the three aforementioned approaches into a single analysis. To test the hypothesis, it rst retrieves the genes from two WikiPathways and for each gene it searches for nanopubs. This is done by looking up equivalent gene IRIs using the IMS server and then for each gene IRI search for nanopublications. It then counts only the original IRI and for all equivalent IRIs and reports the di erences. 3

Results

With the R Markdown notebook we searched for nanopublications for two popular WikiPathways, WP241 and WP2059. The rst has mostly NCBI Gene identi ers and the second mostly Ensembl Gene identi ers. The script reports that NCBI Gene identi ers return the most nanopublications when searching the full nanopublication network, indicating nanopublication data sets prefer to use NCBI Gene identi er-based IRIs. This observation a ects the number of additional nanopubs found via equivalent IRIs: for WP241 we indeed nd a high number of found pathways when using only the IRI returned by the WikiPathways SPARQL endpoint and a lower number for WP2059. For WP241 we get on average 464 nanopublications (min: 10, max: 1000, median: 288), while for WP2059 we retrieve on average 21 nanopublications (min: 0, max:1000, median: 1). The count is currently capped at 1000 nanopublications, imposed by the grlc API wrapping around the nanopublication server, which explains this artifact.

As hypothesized, using equivalent IRIs, as returned by the IMS, will retrieve additional nanopublications. The number of equivalent IRIs by the IMS is different for both pathways. For WP241 it returns on average 9 IRIs (min: 7, max: 23, median: 7) and for WP2059 it returns on average 11 IRIs (min: 7, max: 29, median: 10). The returned IRIs are a mix of mappings to other database sources and di erent IRI patterns for the same database identi er.

Indeed, with these additional IRIs, more nanopublications are found on the international nanopublication network. Furthermore, when the additional IRIs include NCBI Gene identi er-based IRIs, we nd a higher number of additional nanopublications. This observation is similar for that observed when only using the original NCBI Gene identi er-based IRI, as explained above. Therefore, we nd fewer additional nanopublications for WP241 which used predominantly NCBI Gene identi ers: on average it nd 17 additional nanopublications (min: 0, max: 230, median: 3). However, for WP2059 which used predominantly Ensembl identi ers we nd a much higher number of additional nanopublications, on average 44 (min: 0, max: 1099, median: 0).

Discussion

It goes without discussion that the results show that we indeed nd more nanopublications about a certain gene. However, there are some aspects that must be noted. First, the enrichment is only as good as the completeness and quality of the gene-gene identi er mapping link sets. The link sets used in this study are biased towards equivalent IRIs based on Ensembl and NCBI Gene identi ers. If nanopublication data sets use other gene identi ers, these will still not be found. Besides this completeness issue, the author had the impression that some mappings were missing, something will be explored. A more elaborate analysis is planned, involving more pathways and more identi er sources.

Another e ect of this completeness aspect is that the link sets used only cover gene-gene identi er mappings. However, the scienti c lenses approach also allows gene-RNA and gene-protein mappings, under a lens that equates genes and proteins. This is particularly relevant for WikiPathways, where gene and proteins are frequently used as equivalent. The ability to load such additional link sets and the feature of the IMS to turn on and o the lenses, would allow returning even more nanopublications. 5

Conclusion

The results show that using an IRI mapping service increases the recall when searching nanopublication, overcoming the problem that nanopublications do not (and should not need to) normalize IRIs. This paper demonstrates this with the application of a locally installed BridgeDb IMS service and an R script in combination with online nanopublication services.

References

1. Pain , E. : How to keep up with the scienti c literature . Science (November 2016 )

2. Mons , B. , Velterop , J.: Nano-Publication in the e-Science Era . In: Proceedings of the Workshop on Semantic Web Applications in Scienti c Discourse (SWASD 2009 ), collocated with the 8th International Semantic Web Conference (ISWC2009) , Washington DC , USA, October 26 , 2009 . Volume 523 ., CEUR Workshop Proceedings ( October 2009 )

3. Groth , P. , Gibson , A. , Velterop , J.: The anatomy of a nanopublication . Information Services and Use 30 ( 1 ) ( January 2010 ) 51 { 56

4. Queralt-Rosinach , N. , Kuhn , T. , Chichester , C. , Dumontier , M. , Sanz , F. , Furlong , L.I. : Publishing DisGeNET as nanopublications . Semantic Web 7 ( 5 ) ( June 2016 ) 519 { 528

5. Chichester , C. , Gaudet , P. , Karch , O. , Groth , P. , Lane , L. , Bairoch , A. , Mons , B. , Loizou , A. : Querying neXtProt nanopublications and their value for insights on sequence variants and tissue expression . Web Semantics: Science, Services and Agents on the World Wide Web 29 ( December 2014 ) 3 { 11

6. Kuhn , T. , Chichester , C. , Krauthammer , M. , Queralt-Rosinach , N. , Verborgh , R. , Giannakopoulos , G. , Ngonga Ngomo , A.C. , Viglianti , R. , Dumontier , M. : Decentralized provenance-aware publishing with nanopublications . PeerJ Computer Science 2 ( August 2016 ) e78+

7. Kuhn , T. , Willighagen , E. , Evelo , C. , Queralt-Rosinach , N. , Centeno , E. , Furlong , L.I. : Reliable Granular References to Changing Linked Data . In d'Amato, C. , Fernandez , M. , Tamma , V. , Lecue , F. , Cudr-Mauroux , P. , Sequeda , J. , Lange , C. , He in , J., eds.: The Semantic Web ISWC 2017 . Volume 10587 . Springer International Publishing, Cham ( 2017 ) 436 { 451

8. Kuhn , T. , Meroo-Peuela , A. , Malix , A. , Poelen , J. , Hurlbert , A. , Centeno , E. , Furlong , L.I. , Queralt-Rosinach , N. , Chichester , C. , Banda , J. , Willighagen , E. , Ehrhart , F. , Evelo , C. , Malas , T. , Dumontier , M. : Nanopublications: A Growing Resource of Provenance-Centric Scienti c Linked Data . In: Proceedings of IEEE eScience 2018 , arXiv. org (September 2018 )

9. Williams , A.J. , Harland , L. , Groth , P. , Pettifer , S. , Chichester , C. , Willighagen , E.L. , Evelo , C.T. , Blomberg , N. , Ecker , G. , Goble , C. , Mons , B. : Open PHACTS : semantic interoperability for drug discovery . Drug Discovery Today 17 ( 21 - 22 ) ( November 2012 ) 1188 { 1198

10. Willighagen , E.L. , Waagmeester , A. , Spjuth , O. , Ansell , P. , Williams , A.J. , Tkachenko , V. , Hastings , J. , Chen , B. , Wild , D.J.: The ChEMBL database as linked open data . Journal of Cheminformatics 5 ( 1 ) ( 2013 ) 23

11. Waagmeester , A. , Kutmon , M. , Riutta , A. , Miller , R. , Willighagen , E.L. , Evelo , C.T. , Pico , A.R. : Using the Semantic Web for Rapid Integration of WikiPathways with Other Biological Online Data Resources . PLOS Computational Biology 12 ( 6 ) ( June 2016 ) e1004989

12. Van Iersel , M.P. , Pico , A.R. , Kelder , T. , Gao , J. , Ho , I. , Hanspers , K. , Conklin , B.R. , Evelo , C.T.: The BridgeDb framework: standardized access to gene, protein and metabolite identi er mapping services . BMC Bioinformatics 11 ( 1 ) ( January 2010 ) 5 +

13. Brenninkmeijer , C. , Evelo , C. , Goble , C. , Gray , A.J.G. , Groth , P. , Pettifer , S. , Stevens , R. , William , A.J. , Willighagen , E.L. : Scienti c Lenses over Linked Data: An Approach to Support Task Speci c Views of the Data. A Vision . In: Linked Science 2012 - Tackling Big Data . ( 2012 )

14. Batchelor , C. , Brenninkmeijer , C.Y.A. , Chichester , C. , Davies , M. , Digles , D. , Dunlop , I. , Evelo , C.T. , Gaulton , A. , Goble , C. , Gray , A.J.G. , Groth , P. , Harland , L. , Karapetyan , K. , Loizou , A. , Overington , J.P. , Pettifer , S. , Steele , J. , Stevens , R. , Tkachenko , V. , Waagmeester , A. , Williams , A. , Willighagen , E.L. : Scienti c Lenses to Support Multiple Views over Linked Chemistry Data . In Mika, P., Tudorache , T. , Bernstein , A. , Welty , C. , Knoblock , C. , Vrandei , D. , Groth , P. , Noy , N. , Janowicz , K. , Goble , C., eds.: The Semantic Web ISWC 2014 . Volume 8796 . Springer International Publishing, Cham ( 2014 ) 98 { 113

15. Juty , N. , Le Novre , N. , Laibe , C. : Identi ers.org and MIRIAM Registry: community resources to provide persistent identi cation . Nucleic Acids Research 40 ( D1 ) ( January 2012 ) D580{ D586

16. Adriaens , M. , Evelo , E. , Willighagen , E. , Marianthi , F. , Kalafati , Pico, A. , Kelder , T. , Slenter , D. , Hanspers , K. , Txr24, Egoyenechea, Fehrhart, Thakur, G. , Je : One Carbon Metabolism (Homo sapiens)

17. Salomonis , N. , Hanspers , K. , Fehrhart , Pico, A. , Willighagen , E. , Kelder , T. , Mlius , J.: Alzheimers Disease (Homo sapiens)

18. Willighagen , E.: Accessing biological data in R with semantic web technologies . Technical report, PeerJ Inc. (March 2014 )

19. Meroo-Peuela , A. , Hoekstra , R. : grlc Makes GitHub Taste Like Linked Data APIs . In Sack, H., Rizzo , G. , Steinmetz , N. , Mladeni , D. , Auer , S. , Lange , C., eds.: The Semantic Web . Volume 9989 . Springer International Publishing, Cham ( 2016 ) 342 { 353

20. Sferruzza , D. , Rocheteau , J. , Attiogb , C. , Lanoix , A. : Extending OpenAPI 3.0 to Build Web Services from their Speci cation . In: Proceedings of the 14th International Conference on Web Information Systems and Technologies , Seville, Spain, SCITEPRESS - Science and Technology Publications ( 2018 ) 412 { 419

21. Ehrhart , F. , Melius , J. , Cirillo , E. , Kutmon , M. , Willighagen , E.L. , Coort , S.L. , Curfs , L.M. , Evelo , C.T.: Providing gene-to-variant and variant-to-gene database identi er mappings to use with BridgeDb mapping services . F1000Research 7 ( September 2018 ) 1390