Figure 1: Overview of the HL7 Reference Information Model.

The HL7 Reference Information model was constructed as a high-level generic UML object model representing physical Entities (people, places, things) which can play different Roles with respect to different scopes (e.g., person – patient of a hospital, person – subject of a trial, chemical – ingredient in a drug, anatomical part – structure, compound – moiety, etc.). Entities (in a role) can participate in defined ways (participation type) in Acts (e.g., patient – subject of observation or treatment, researcher – author of experimental data, etc.) Acts can relate to other Acts via ActRelationships of types to indicate decomposition (sub-activity), reason, cause-effect, preconditions, etc. A rich set of relationship types between Entities (Role.classCode) and between Entity (in a role) and Act (Participation.type) and between Acts (ActRelationship.type) ensures that rich semantic structures can be expressed. The combination of a high-level abstract model and the rich relationships has allowed the RIM to be applied across dozens of domains (e.g., laboratory, pharmacy, public health, pharmacovigilance, administration and finance).

A Single Model for Particular Information and Universal Knowledge

What makes the HL7 RIM relevant for ontology design is that, in addition to expressions over particular instances, the RIM also allows writing expressions over universals (Figure 2). One can describe Acts in general, mention preconditions, risks and outcomes in general, and speak of Entity universals (determiner = KIND). The Role class “instance (of kind)” links between a particular physical Entity as the instance of a universal Entity. Likewise the ActRelationship of type “instance-of” connects the particular Act with the universal Act which it realizes. Likewise among the Act and Entity universals, specialization relationships and roles (respectively) can be used to express generalization hierarchies. This flexibility for example would allow the BioPax users to continue to treat their knowledge as database records, but at the same time declare that they refer to universals, classes of molecules, kinds of interactions, rather than incorrectly specifying them as particular instances.35

Pragmatics, Meaning and Action

In addition to semantics of universals and particulars, the HL7 RIM expressed pragmatics, i.e., it represents requests, intentions, promises, goals, etc. as Acts with a common model for the content and represents the illocutionary force in the form of the mood code. This is what makes the HL7 RIM uniquely suitable for health care and research data models, because it can explicitly distinguish between standard operating procedures, specific plans, requests, and finally factual statements. From this follows that facts are represented as statements (one could call them “opinions”) which is particularly useful when considering that many large scientific knowledge bases (e.g., pathway databases) contain the result of inferences and expert opinions, therefore all factual statements in these databases are usually traced back to their originators (curators, literature, etc.) or even detailed description of experiment. The HL7 RIM is particularly designed for this purpose of attributing facts to authors, supporting evidence and experimental context.

While the HL7 RIM has been adopted for a world-wide standardization of biomedical information (e.g., U.S., Canada, UK, and Germany) in such domains as healthcare, public health, and among product regulators, and while the caBIG based its clinical trials data model (CTDM) on the HL7 RIM,30 the application of the HL7 RIM to bioinformatics data and knowledge is relatively new. The Clinical Genomics Special interest group within HL7 has applied the RIM to genomics investigations, however, the point of view taken in that activity is to consider all clinical genomics phenomena essentially laboratory results. This approach had been heavily criticized by some colleagues and had led to the understanding as if the RIM itself was unable to represent biological reality. The present proposal is to rectify this misunderstanding by demonstrating how the HL7 RIM is designed to represent the biomedical reality, especially structural biology (through its physical Entity and Role relationship structures) but also biological processes (as Acts.) We have described the approach previously[i] and validated it in our prior work in integrating biological pathway network databases as described below.