For materials database integration, we introduce a semantic web technology into a polymer database, called PoLyInfo. A resource data framework (RDF) was used to create a semantic description of the polymer formation process (polymerization). The polymerization correlates polymers with their source monomers, and then the monomers in the PoLyInfo RDF were conceptually linked to those in other chemical substance databases, such as Nikkaji. Although common ontology was not used in the PoLyInfo and Nikkaji RDFs, 94.3% of the monomers in PoLyInfo were assigned a Nikkaji substance ID, and the established information path provided probable polymerization information for monomers in the Nikkaji RDF.
A semantic web technology using resource data framework (RDF) [
Copyright © 2019 for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).
1 https://www.nims.go.jp/eng/index.html, 2 https://www.jst.go.jp/EN/ (JST)2 to NIMS in 2003, the database has been unconcerned with semantic web technology, that is, the database has been functioning as an unlinked data source. Due to the skillful identification of polymer naming in academic articles, PoLyInfo has proved expensive owing to an increase in polymer data that are not easily obtainable from other chemical databases. On the other hand, the uniqueness of PoLyInfo means it contains few common terms in comparison to other related elements, resulting in a technical barrier to the linked data. From a survey of PoLyInfo aiming to realize linked data, we deduced that the most important and linkable term in the database is the “monomer name”. The grounds of this deduction were established as follows: - Polymers are essentially synthesized from monomers. - The synthesis (polymerization path) represents the primary information for polymer developers. - There are several monomer databases linked to life science with RDF.
Fortunately, as PoLyInfo has managed polymer names, polymerization paths,
monomer names with original identification numbers (IDs), and a protocol between these
IDs, we were able to design an RDF-connected graph for PoLyInfo, as shown in Figure
1, without utilizing frontier tools [
The triple graph in Fig. 1 shows an example of an industrially important polymer, “polystyrene”, and its semantic links can be described as follows: 1) The polymer with composition unit identified by CU010001 has the name “Polystyrene” (CU010001 ns2:label “Polyethylene”@en.). 2) CU010001 has a polymerization path with the ID number J000002 (CU010001 ns1:pHasPolymerizationPath J000002.). 3) J000002 has the name of “Addition polymerization” (J000002 ns2:label “Addition polymerization”@en.). 4) J000002 has source monomers with ID numbers M0301021 and M0101001 (J000002 ns1:pHasMonomer M0301021, M0101001.). 5) M0301021 has the name “Buta1,3-diene”, and M0101001 has the name “Ethene” (M0301021 ns2:label “Buta1,3-diene”@en. M0101001 ns2:label “Ethene”@en.). Consequently, the polymer CU010001 is correlated to the monomers M0301021 and M0101001.
In this study, we performed data linking with an organic chemical database created
by JST, called Nikkaji (Japan Chemical Substance Dictionary) [
To link the PoLyInfo RDF to the Nikkaji RDF, we introduced conceptual linking,
based on Simple Knowledge Organization System (SKOS) specifications [
6) M0301021 (Buta1,3-diene) and M0101001 (Ethene) of PoLyInfo closely match J4.043F and J1.939I of Nikkaji, respectively. The corresponding RDF triples in SKOS can be expressed by: @prefix skos: <http://www.w3.org/2004/02/skos/core#> . ns1: M0301021 skos:closeMatch nikkaji: J4.043F . ns1: M0101001 skos:closeMatch nikkaji: J1.939I .
As these entities may not be chemically identical (e.g., they may exhibit a difference in purity), we concluded that the “skos:closeMatch” was better than “skos:exactMatch” and “owl:sameAs”. Based on the designed RDF in Fig. 1, we made triples for the polymerization process in PoLyInfo, as listed in Table 1.
This table indicates that 15,884 polymers of 2) had information concerning the polymerization path, and 35,018 monomers of 4) were related to these polymerizations. As PoLyInfo has 17,825 monomers while the RDF for skos:closeMatch of 6) was 16,809 triples, this shows that 94.3% of monomers in PoLyInfo were assigned to Nikkaji substance ID. The total number, 227,482 triples, was not significant. However, we also performed a demonstration using a query written in SPARQL. The crossover task was to compile a list of polymers in PoLyInfo, which could be synthesized from the monomers in Nikkaji, and the response provided 39,907 polymers. This result indicates that the PoLyInfo RDF was successfully linked to the Nikkaji RDF.
Although we showed the potential of the use of linked data technology between PoLyInfo and Nikkaji, we still encountered a number of problems intrinsic to polymer science. In PoLyInfo, we normalized the polymer name to International Union of Pure and Applied Chemistry (IUPAC)6 nomenclature. Unfortunately, the international rule sometimes conflicts with the RDF format of W3C. An example of this conflict can be seen in a triple for polymerization ID of J0018355: <https://polymer.nims.go.jp/rdf/J0018355> ns1:pHasMonomer <https://polymer.nims.go.jp/rdf/M4000864> ; ns2:label "Addition_polymerization_of_9,9,9',9',9",9"-hexahexyl-7-(4-vinylphenyl)-2,2':7',2"-terfluorene"@en .
The triples indicate that J0018355 with source monomer M4000864 has a label of “Addition polymerization of 9,9,9',9',9",9"-hexahexyl-7-(4-vinylphenyl)-2,2':7',2"-terfluorene”. However, the IUPAC monomer name M4000864, including double quotes (") without an escape code, clearly conflicts with the RDF format (the escape code is undefined in IUPAC nomenclature), resulting in a syntax error in uploading process. 4
To establish linked data for the polymer database PoLyInfo, we designed RDF triples implementable in a polymerization process. The triples describing polymerization from source monomers finally bridged the entities present in the other chemical substance databases, such as Nikkaji. By using the 227,482 triples created for the purposes of this study, we found 39,907 polymers that could be synthesized from monomers registered in Nikkaji. In some cases, we found an unexpected conflict between the International Union of Pure and Applied Chemistry (IUPAC) nomenclature and the RDF format recommended by W3C.
This study was supported by the Cabinet Office, Government of Japan, Cross-ministerial Strategic Innovation Promotion Program (SIP), “Technologies for Smart Bioindustry and Agriculture” (funding agency: NARO). 5