Pistoia Alliance, a not-for-profit consortium in life sciences and healthcare, encourages collaborative efforts to deliver data-driven value by leading ontology projects aimed at ensuring alignment and interoperability across the pharmaceutical, life-science and healthcare industries. The Identification of Medicinal Products Ontology (IDMP-O) was created to ontologically represent the IDMP ISO standards. More ontologies are currently under development by Pistoia Alliance members. To ensure the utility of those ontologies, we hypothesized that the Basic Formal Ontology (BFO) could enable interoperability across them. After examinations of IDMP-O terms, its mid-level Object Management Group (OMG) terms, and BFO terms, we initiated efforts of mapping IDMP-O/OMG to BFO. Preliminary results and issues encountered are discussed here. Challenges reside in the nature and purpose of IDMP-O, which is mostly representing informational specifications for regulatory purposes, as well as the use of OMG:Role, which is far broader than BFO:role in semantics. Pistoia Alliance is committed to continuously evaluate BFO and to establish a connection between BFO and IDMP-O to enable interoperability between its ontologies.
In this paper, we share first results in evaluating the adoption of the Basic Formal Ontology
(BFO) [
Here, we focus on analyzing the IDMP Ontology® (IDMP-O) [
BFO is a widely adopted top-level ontology and an ISO standard (ISO/IEC 21838-2), which is used as a top-level ontological architecture for more than 600 projects. As such, BFO serves as a baseline for information sharing practices, and enables coherent interoperability of heterogeneous data. To meet Pistoia’s goal of delivering data driven value ontologies such as IDMP-O, Pharmaceutical CMC Process, and Clinical Operation (ClinOps) need to be interoperable. The purpose of BFO, and of other top-level ontologies, is to provide a small, shared set of terms composing a top-level hierarchy that other ontologies can reuse for downward population. In the application considered for this paper, the hypothesis is that BFO would enable interoperability across Pistoia ontologies. More specifically, a mapping from IDMP-O to resources in the BFO ecosystem would allow for IDMP-O to be semantically integrated with other ontologies which are already part of Pistoia Alliance products and adopt BFO.
At the time of writing, development of other Pistoia ontologies was at its beginning, but
version 1.1 of IDMP-O was already released [
After the assessment phase, we started to identify possible relations between classes in IDMP-O and in the BFO ecosystem. Next sections discuss our development of this process, as well as issues encountered. IDMP-O was created with the intent of faithfully representing the terminology created in the IDMP standards, which, as we mentioned earlier, are primarily designed to serve as common standards for improved data sharing during regulatory processes, such as pharmacovigilance, submission, and global supply chain. In the rest of this paper, we motivate a process of mapping of IDMP-O into classes developed in the BFO ecosystem, we describe first steps achieved in identifying such mappings, and we discuss currently open questions and next steps to take.
The project involved members of Pistoia Alliance, developed over several months and included multiple steps of consultation with technology partners of Pistoia Alliance as well as of the BFO ontology communities. As a first step, we reviewed and identified the main classes used in IDMP-O and their use case as motivated by the adopted ISO standards. We compared IDMP-O and BFO terms, both manually and using the Lexical OWL Ontology Matcher (LOOM) algorithm implemented on BioPortal as an open-source software. In the second phase of the project, we started identifying possible mappings between IDMP-O and classes in the BFO community. While developing these mappings, we shared our preliminary results with some of the IDMP-O primary developers. When the issues we faced touched foundational issues in BFO, as for example is in the case of mapping OMG:Role which we describe in the next section, we also reached out to other members of the BFO community.
During the second phase of development, we considered two possible methods to adopt in
order to connect IDMP-O and BFO. The first includes a complete alignment of IDMP-O to BFO,
which would include a subsumption of IDMP-O classes under terms in the BFO hierarchy, as
well as a reconstruction of the definitions introduced by IDMP-O to reflect the new hierarchy.
The second possible method is the creation of mapping triples from BFO and IDMP-O. A
mapping triple is a <s, p, o> triple, where the subject ‘s’ and predicate ‘p’ are terms from two
different ontologies, and the predicate ‘p’ is a relation appropriately connecting the two [
Among the different strategies for creating mappings, we evaluated the adoption of the
method of translation definitions, already employed by Michael Gruninger and associates as
part of the COLORE project [
pharmaceutical product medicinal product matter process specification for the qualitative and
quantitative composition of a medicinal product in the dose form approved for administration in line with the regulated product
information specification for a pharmaceutical
product or combination of pharmaceutical products that may be administered to human beings (or animals) for treating or preventing disease, with the aim/purpose of making a medical diagnosis or to restore, correct or modify physiological functions something that has mass and occupies space by virtue of having
volume structured set of activities involving various enterprise entities designed and organized for a given purpose (IDMP) - clauses 3.1.28
and 9.7, Figure 12 Resourced from ISO 11239:2012 Health
informatics
Identification of medicinal products (IDMP) - clause 3.1.21
Resourced from ISO 11239:2012 Health
informatics
Identification of medicinal products (IDMP) - clause 3.1.17
Resourced from ISO 11238:2018 Health
informatics
Identification of medicinal products (IDMP) - clause 3.41 Resourced from ISO 18629-11:2005(en) Industrial automation systems and integration - Process specification language - clause 3.1.21
To faithfully represent the mentioned IDMP standards, the main classes used in IDMP-O
refer to specifications of medicinal products based on the IDMP standards’ purpose of
“facilitating the reliable exchange of medicinal product information” [
A preliminary comparison of BFO and IDMP-O reveals four overlapping terms with the same label but different semantics: process, role, site, and object. Process p is a process means: p is an occurrent that structured set of activities has some temporal proper part and for some involving various enterprise time t, p has some material entity as entities designed and organized participant at t for a given purpose
Object b is a role means: b is a realizable entity & b named specific behavior of exists because there is some single bearer something participating in a that is in some special physical, social, or particular context* （OMG institutional set of circumstances in which term) this bearer does not have to be & b is not such that, if it ceases to exist, then the physical make-up of the bearer is thereby changed. b is a site means: b is a three-dimensional place, setting, or context in immaterial entity whose boundaries either which something is situated or (1) (partially or wholly) coincide with the to which something is, or may boundaries of one or more material entities be, bound* （OMG term) or (2) have locations determined in relation to some entity an object is a material entity which manifests anything perceivable causal unity & is of a type instances of which conceivable.* (OMG term) are maximal relative to the sort of causal unity manifested or
The comparison of definitions, axioms and editor notes of the above four terms revealed that IDMP-O:process is related (SKOS:closematch) to OBI:planned process2, which is itself a subclass of BFO:process. Therefore, IDMP-O:process is a subclass (or SKOS:narrowerMatch) of BFO:process. The OMG term object is adapted from the ISO 1087 Terminology work and terminology science standard clause 3.1.1. The class OMG:object is defined so broadly that it is almost equivalent to OWL:thing. An OMG:Site is a subclass of OMG:Role, while a BFO:site is 2 OBI refers to Ontology for Biomedical Investigations [17]. The definition of planned process in OBI is “A process that realizes a plan which is the concretization of a plan specification”. not a BFO:role. In this paper, we focus on discussing OMG:Role as a first example in aligning BFO and OMG/IDMP-O.
OMG:Role, which refers to the specific behavior that entities have in a particular context when they play such a role, provides contextual information for resources in the ontology. “Ingredient” is an example of an OMG role. Relevantly for our discussion, a substance can be one type of ingredient in one medicinal product, but another in a different product. Therefore, IDMP-O decided to model ingredients as roles played by substances. These roles, in the IDMPO case, are specified in a regulatory approved specification of the products. As such, “being an ingredient” appears to be an intimately contextual feature, not directly belonging to the physical features of the entity which happens to be an ingredient.
The debate over roles in formal ontology is wide and has produced discussions over many different semantic understandings of roles [18–20] and specifically over BFO:role [19]. OMG:Role has subclasses such as “agent role”, “functional role”, “process role”, “site”, “structural role”, and “undergoer”. IDMP-O mostly expanded from “functional role”, by adding for example, “ingredient”, “medication”, “moiety role” and “product role”. Not all of these subclasses seem to fit under the umbrella of BFO:role. For example, it is not clear whether OMG:site should be mapped into a BFO:site, which is not a subclass of BFO:role. The best possible solution is to implement a disjunctive mapping for the different subclasses of OMG:role, where each of them is mapped into a different class in the BFO or CCO hierarchies. However, the usage and the defined relations within OMG may complicate the mappings.
The differences between OMG:Role and BFO:role are exemplified by the fact that an
OMG:Role doesn’t have any restriction on the type of resources it can be OMG:isPlayedBy,
while a BFO:role is restricted by the BFO:inheres in relation that only allows a BFO:independent
continuant (non-spatial region) as its range. For example, a BFO:role could be OMG:played by
subclasses of OMG:Matter. On the other hand, specifications in IDMP-O can play roles. But a
specification, as we have just discussed, is not a subclass of OMG:matter. A specification rather
seems to be, in BFO terms, a type of generically dependent continuant (GDC), and more
specifically a type of information content entity (ICE), as introduced in CCO [21]. CCO is a
family of ontologies composing a mid-level architecture [
The issue can be solved in two ways. The first is to allow a GDC to bear a BFO:role. The second is to introduce a cognate notion of roles to be adopted for representing contextual information about GDCs. Roles for GDCs seem to have a broader use even outside of the IDMPO use case described in this paper. For example, a certain ICE could have the role of being a credential for a certain institution, or a password could have the role of being accepted to provide access to a certain system but not to another. Such addition would then be beneficial to broader use cases.
IDMP-O:substance also seems to fall under the scope of ICEs. An IDMP-O:substance seems to be a type of ICE which specifies some IDMP-O:Matter. As already argued, IDMP-O:Matter is equivalent to BFO:material entity. This would mean that an IDMP-O:substance is equivalent to a CCO:information content entity which CCO:is about some BFO:material entity. Table 2 shows the first tentative mappings achieved as a result of part 2 of the project. IDMP-O:process is mapped into a subclass of BFO:process, even though the definition mentions “structured set of activities”, which could make it equivalent to a CCO:act or CCO:planned act.
Again, the evaluation of a more precise mapping in this case relies on further discussion with Pistoia Alliance stakeholders and with a closer investigation of the use of IDMP-O:process subclasses. Notice that, pending the issues described above, an OMG:Role is for the moment only a SKOS:relatedMatch with a BFO:role. Although our preference is for creating mappings which are able to connect terms through logical axioms in OWL, we used SKOS mappings as a provisional tool for providing human-readable relations between them, as well as for guiding future developments. We are currently also evaluating other efforts in automating or semiautomating mappings steps. At the time of writing, we were not able to identify any suitable match for subclasses of OMG:Role such as IDMP-O:ingredient.
In the section above we identified preliminary potential mappings between core classes in IDMP-O, OMG, and classes in the BFO ecosystem in an attempt to evaluate BFO as a framework to enable the interoperability of IDMP-O and other Pistoia ontologies. This preliminary effort also identified standing issues in creating such mappings. One unresolved problem we have identified is the axiomatization of BFO:role, which doesn’t allow for GDCs to bear roles. This means that, if an OMG:Specification is indeed a GDC, then it is not allowed to have a BFO:role, which impedes an equivalence mapping between OMG:Role and BFO:role. Moreover, some of the subclasses of OMG:role may not be subclasses of BFO:role at all.
Take as an example the design pattern of IDMP-O:ingredient [20] to further present the problem with mapping OMG:role. In IDMP-O, a “product composition” defines a “pharmaceutical product”, and a “product composition” has as ingredient “ingredient”. The IDMP-O “has ingredient” relation is a subclass of OMG:manifests, which “indicates a role that realizes, displays, or shows something, typically in some context”. A “pharmaceutical product” comprises some “substance” and a “substance” plays the role of “ingredient”. In this way, IDMPO provides information about the substance playing an ingredient role in the context of a product composition. Figure 1 shows a preliminary rendition in equivalent BFO terms representing the above pattern. IDMP-O:defines is provisionally mapped into CCO:describes, IDMP-O:has ingredient has been provisionally mapped into CCO:prescribes, and IDMPO:comprises has been provisionally mapped into BFO:has continuant part. The OMG:plays role object property doesn’t seem to be equivalent to BFO:inheres in due to the “substance” being a “specification”. This “plays role” relation has been substituted by a new relation, “GDC-has role”, that connects GDCs to roles. The exact mappings for these object properties are a subject for future discussions and cannot be exhaustively discussed in this place.
Some terms defined in the original IDMP ISO standards are ambiguous or circular. In order
to remove any ambiguity and ensure a consistent implementation of the standards across the
industry, Pistoia Alliance IDMP-O project members as well as the primary ontology developers
from EDMC went through lengthy discussions for the past two years to develop a more precise
semantic understanding of those terms and to create the current version of IDMP-O. The
agreement reached was to treat most of the terms as referring to the documentation and reports
created when assessing regulatory compliance, for example, in the case of pharmacovigilance,
or medicinal product traceability for global supply chain integrity [
For these reasons, many of the classes in IDMP-O would be subclasses of BFO:GDC, and the large amount of contextual information stored in some of these regulations, as we have seen, requires roles to be related to such entities. The issue can be solved by allowing for a GDC, such as a specification, to play a role. This solution can be implemented as native to IDMP-O or other Pistoia Alliance ontologies, or can be put forward as recommendations to the BFO community. As highlighted in this paper, there are multiple use cases for representing contextual information for a GDC. These range from the IDMP-O use cases discussed in this work and exemplified in Figure 1 to the representation of relations between GDCs or parts of the same GDC, as it is the case for example for parts of a document.
In many cases, these relations may be reduced to other models which can be created using existing BFO resources. Nevertheless, if only as a shortcut relation, allowing GDCs to bear roles seems to be an extremely powerful tool for ontological representation, which greatly simplifies models for use cases which require representing complex information entities. Discussion with members of the BFO development team revealed that this was already an open issue in the community. Introducing a way to connect GDCs to roles, or introducing a new class for representing the specific types of roles borne by GDCs, would allow for BFO-based ontologies to specify these new relations and to represent a wider array of use cases.
Although ISO standards for identification of medicinal products provide definitions and terminologies, not all of these are directly usable to develop an ontology. These definitions can be circular, vague, or even missing. A consistent interoperable implementation of the IDMP ISO standards across companies and stakeholders is the motivation behind the creation of the IDMPO ontology, since an ontological approach can clearly define and remove the ambiguity surrounding the terminology employed in ISO standards. IDMP-O is developed as part of efforts initiated by Pistoia Alliance alongside other life sciences, health care, and pharmaceutical industry ontologies, for example, the Pharmaceutical Process CMC Ontology and Clinical Operations Ontology. Both projects are based on the IOF-core ontology and the BFO ecosystem, therefore evaluating BFO for interoperability across those ontologies is critical. After evaluating the semantics of core IDMP-O terms and the higher level OMG terms, we came to the conclusion that although IDMP-O cannot be easily aligned to BFO, establishing mappings from IDMPO/OMG to BFO and viceversa is indeed possible. Many of the terms in IDMP-O refer to documentations and prescriptions such as OMG:Specification, which are types of GDC in BFO. The first major obstacle we encountered is that IDMP-O makes use of an OMG:Role that is far broader than BFO:role in semantics. Furthermore, specifications in IDMP-O can play roles, but a GDC in BFO is not allowed to bear roles. The authors are working with developers in the BFO community to establish the IDMP-O use case, in order to potentially expand the use of BFO:role so that GDCs can be connected with roles.
The main purpose of IDMP-O is to faithfully represent IDMP standards for regulatory purposes. A lot of terms in IDMP-O are represented as specifications of the regulated medicinal product information. IDMP-O does connect specifications with physical entities via the “specifies” relation. Pistoia Alliance is furthermore developing other BFO-based ontological tools which focus on the R&D phase in a pharmaceutical product life cycle. The R&D world utilizes many BFO-based OBO foundry ontologies, whose primary focus is to represent physical entities, and not the regulations surrounding them. Connecting R&D data regarding physical processes and entities to regulatory data will only be feasible by adopting a common semantic understanding. Evaluating BFO as a top-level ontology to achieve these objectives and establishing a BFO-IDMP-O connection is a continuous effort to which Pistoia Alliance is committed, whose first steps we have presented in this paper.
The authors wish to acknowledge the insights gained from comments received by Pawel Garbacz from EDM Council, Elisa Kendall from Object Management Group, Dr. Christian Baber, the Chief Portfolio Officer from Pistoia Alliance, and John Beverley, assistant professor from University at Buffalo, SUNY and Co-director of the National Center for Ontological Research. [17] A. Bandrowski, R. Brinkman, M. Brochhausen, M.H. Brush, B. Bug, M.C. Chibucos, K.
Clancy, M. Courtot, D. Derom, M. Dumontier, L. Fan, J. Fostel, G. Fragoso, F. Gibson, A. Gonzalez-Beltran, M.A. Haendel, Y. He, M. Heiskanen, T. Hernandez-Boussard, M. Jensen, Y. Lin, A.L. Lister, P. Lord, J. Malone, E. Manduchi, M. McGee, N. Morrison, J.A. Overton, H. Parkinson, B. Peters, P. Rocca-Serra, A. Ruttenberg, S.-A. Sansone, R.H. Scheuermann, D. Schober, B. Smith, L.N. Soldatova, C.J. Stoeckert, C.F. Taylor, C. Torniai, J.A. Turner, R. Vita, P.L. Whetzel, J. Zheng, The Ontology for Biomedical Investigations, PLOS ONE 11 (2016) e0154556. https://doi.org/10.1371/journal.pone.0154556. [18] F. Loebe, Abstract vs. social roles --Towards a general theoretical account of roles, Appl.
Ontol. 2 (2007) 127–158. [19] N. Guarino, BFO and DOLCE: So Far, So Close..., Cosm. TAXIS 4 (2017) 10–18. [20] R. Mizoguchi, S. Borgo, YAMATO: Yet-another more advanced top-level ontology, Appl.
Ontol. 17 (2022) 211–232. https://doi.org/10.3233/AO-210257. [21] M. Jensen, G. De Colle, S. Kindya, C. More, A.P. Cox, J. Beverley, The Common Core
Ontologies, Preprint (n.d.). https://arxiv.org/abs/2404.17758. [22] E. Kendall, Pattern: Representing Ingredients - APPROVED - Pistoia Alliance IDMP Ontology - Confluence, (2023). https://diledmcouncil.atlassian.net/wiki/spaces/IDMP/pages/7246664/Pattern+Representing+Ingred ients+-+APPROVED (accessed May 1, 2024).